Color coordinated image stitching for look-around systems and applications

Abstract

In various examples, one or more color statistics from a ground projection are used to coordinate colors between a reference frame and a target frame representing an environment. The reference frame and the target frame may be projected onto a representation of the ground (e.g., ground plane) of the environment, an overlap region between the projections may be identified, and a portion of each projection that falls in the overlap region may be regarded as a corresponding ground projection. Color statistics (e.g., one or more correlations between mean values, variances, standard deviations, kurtosis, skew, color channels) may be calculated from ground projections (or portions thereof, such as majority clusters) and used to modify the color of the target frame to have updated color statistics that match those color statistics from ground projections of the reference frame, therefore, colors are coordinated across the reference frame and the target frame.

Description

Background Art

[0001] A surround view system (SVS) provides an occupant of the ego-machine with a visualization of the area surrounding the ego-machine. For example, SVS provides the driver and / or occupant with the ability to view the surrounding area (including blind spots where vision is obscured by a portion of the ego-machine and / or other objects in the environment) without having to reposition (e.g., turn their head, lift up from their seat, lean in a certain direction, etc.). This visualization can assist and facilitate various driving maneuvers, such as smoothly entering or exiting a parking spot without striking objects such as curbs, pedestrians or other vehicles or dynamic objects. More and more vehicles, especially luxury brands and new models, are being produced with SVS capabilities.

[0002] Existing SVS typically uses fisheye cameras (usually mounted on the front, left, rear, and right sides of the vehicle body) to perceive the surrounding area from multiple directions. Additional cameras may be included in special cases, such as long trucks or vehicles with trailers. Frames of image data from individual cameras are stitched together using camera parameters to align the frames and a blending technique is used to combine overlapping areas to provide a horizontal 360° surround visualization.

[0003] However, the sensor array of the camera may capture light under conditions that are different from human vision, and the images generated from the sensor data may represent colors differently from the way our eyes perceive colors. In addition, different sensors or different conditions applicable to the same sensor type may produce images of the environment that do not appear to match between images. In some cases, the mismatched colors appear due to the lighting characteristics of the environment. For example, if the sun is on one side of the ego vehicle, if the ego vehicle is entering a parking structure or tunnel, or if the ego vehicle is partially in the shadow, one side of the ego vehicle may be exposed to more light than the other side, so that the camera on the brighter side may produce a different exposure or use a different automatic setting, and therefore may capture a different color from the camera on the other side. Another way that mismatched colors may occur is through image processing, which is typically applied separately to each frame of the image data, resulting in different changes to different images. Example image processing may include gamma correction for improving the color range, exposure compensation, tone mapping, noise reduction, removing bad pixels, applying white balance, applying color correction to remove lens shading artifacts in fisheye images, and / or other. However, because image processing evaluates and operates on each image independently of the other images, image processing often results in mismatched colors in different images of the same environment.

[0004] One prior art technique for coordinating colors between different images of the same environment selects one of the images as a reference image, determines global color statistics for the reference image, and transfers the global color statistics to the other images so that their global color statistics match those of the reference image. However, this technique is generally ineffective at coordinating colors; for example, when different cameras are looking at different objects or different parts of the same object. Applying a stitching algorithm that uses uncoordinated camera images as input results in an uncoordinated stitched image with noticeable color mismatches at the seam where the two images are stitched together, which can be perceived as artifacts. As a result, conventional techniques can generate distracting artifacts in areas of the stitched image that may be important for a driver or autonomous system to safely operate a vehicle. Therefore, there is a need for improved color coordination techniques that improve the visual quality of stitched images. Summary of the invention

[0005] Embodiments of the present disclosure relate to color coordination across multiple camera sensors. More specifically, systems and methods are disclosed in which color statistics of a ground projection from a reference frame of image data are migrated to a target frame of image data.

[0006] Compared to conventional systems (such as those described above), one or more color statistics of ground projections of reference and target frames can be used to coordinate colors between the reference and target frames. At a high level, a reference frame and a target frame can be identified from frames of image data representing overlapping views of an environment (e.g., an environment surrounding an ego object such as an ego vehicle). The reference frame and the target frame can be projected onto a representation of the ground (e.g., a ground plane) of the environment, an overlapping region between these projections can be identified, and the portion of each projection that lands in the overlapping region can be considered as a corresponding ground projection. Instead of calculating color statistics on the entire reference and target frames, one or more color statistics (e.g., statistical moments or properties of one or more color channels) can be calculated for each ground projection (or a portion thereof, such as a majority cluster). As such, the color statistics from the ground projection can be used to modify the color of the target frame to have updated color statistics that match the color statistics of the ground projection from the reference frame. Migrating color statistics from the ground projection rather than from the entire image in this manner improves color coordination and visual quality of the stitched image compared to prior art techniques that migrate global color statistics from the entire image. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] The present system and method for color coordination of stitched images in autonomous systems and applications are described in detail below with reference to the accompanying drawings, in which:

[0008] Figure 1is a schematic diagram illustrating an example data flow through an example surround view system (SVS) according to some embodiments of the present disclosure;

[0009] Figure 2 is a schematic diagram illustrating an example technique for migrating color statistics of a ground projection from a reference frame to a target frame according to some embodiments of the present disclosure;

[0010] Figure 3 is a flow chart illustrating a method for applying color coordination based at least on object detection according to some embodiments of the present disclosure;

[0011] Figure 4 shows visualization of a stitched image before and after applying color coordination using the entire reference frame and target frame to calculate color statistics according to some embodiments of the present disclosure;

[0012] Figure 5 shows visualization of a stitched image before and after applying color coordination using ground projections of reference and target frames to calculate color statistics in accordance with some embodiments of the present disclosure;

[0013] Figure 6 shows an example overlap region between two ground projections according to some embodiments of the present disclosure, wherein the amount of pixels in the overlap region belonging to a detected object exceeds a threshold;

[0014] Fig. 7A shows an example overlap region between two ground projections according to some embodiments of the present disclosure, wherein the amount of pixels in the overlap region belonging to a detected object is less than a threshold, and Figure 7B An example ground projection with clustered pixels is shown;

[0015] Figure 8 is a flow chart illustrating a method for color coordination using a ground projection of a reference frame according to some embodiments of the present disclosure;

[0016] Fig. 9 is a schematic diagram illustrating an example data flow of an example SVS by applying color coordination to process image data according to some embodiments of the present disclosure;

[0017] Fig.10 is a flow chart illustrating a method for applying color coordination to process image data according to some embodiments of the present disclosure;

[0018] Fig.11A and Fig. 11B An example of image stitching according to some embodiments of the present disclosure is shown, and Fig. 11C is a schematic diagram illustrating an example data flow through an example image stitching system;

[0019] Fig. 12A is an illustration of an example autonomous vehicle according to some embodiments of the present disclosure;

[0020] Fig. 12B According to some embodiments of the present disclosure Fig. 12A Examples of autonomous vehicle camera positions and fields of view;

[0021] Fig. 12C According to some embodiments of the present disclosure Fig. 12A a block diagram of an example system architecture for an example autonomous vehicle;

[0022] Fig.12D According to some embodiments of the present disclosure, a method for Fig. 12A System diagram of an example of communication between autonomous vehicles;

[0023] Fig.13 is a block diagram of an example computing device suitable for implementing some embodiments of the present disclosure; and

[0024] Fig.14 is a block diagram of an example data center suitable for implementing some embodiments of the present disclosure DETAILED DESCRIPTION

[0025] Systems and methods related to color coordination of stitched images are disclosed. For example, systems and methods for migrating color statistics of a ground projection from a reference frame of image data to a target frame of image data are disclosed. The present technology can be used to visualize the environment around an ego object such as a vehicle, a robot, and / or other types of objects in a system such as a parking visualization system, a surround view system (SVS), and / or other systems.

[0026] Although the present disclosure may be related to an example autonomous vehicle 1200 (referred to alternately herein as “vehicle 1200 ” or “ego machine 1200 ”), the example autonomous vehicle 1200 may be referred to herein as “ego vehicle 1200 ” or “ego machine 1200 ”. FIG. 12A to FIG. 12DThe present disclosure may be described with respect to image stitching for a surround view system, but this is not intended to be limiting. For example, the systems and methods described herein may be used by, but are not limited to, non-autonomous vehicles, semi-autonomous vehicles (e.g., in one or more adaptive driver assistance systems (ADAS)), manned and unmanned robots or robotic platforms, warehouse vehicles, off-road vehicles, vehicles coupled to one or more trailers, flying boats, ships, space shuttles, emergency response vehicles, motorcycles, electric or motorized bicycles, aircraft, engineering vehicles, underwater vehicles, drones, and / or other vehicle types. Furthermore, while the present disclosure may be described with respect to image stitching for a surround view system, this is not intended to be limiting, and the systems and methods described herein may be used for augmented reality, virtual reality, mixed reality, robotics, security and surveillance, autonomous or semi-autonomous machine applications, and / or any other technology space where image stitching may be used.

[0027] At a high level, color coordination can be applied to frames of image data captured using different cameras and / or representing overlapping views of an environment (e.g., around an ego object such as an ego vehicle) at substantially the same time or time slice. More specifically, a reference frame and a target frame can be identified from the frames (e.g., based on the direction of ego motion, active viewpoint, operator gaze, brightest color, patch uniformity, and / or other factors), and the ground projections of the reference and target frames can be used to transfer color statistics (e.g., statistical moments or properties of one or more color channels, such as mean, variance, standard deviation, kurtosis, skew, and / or one or more correlations between color channels) from the ground projection of the reference frame to the target frame by modifying the color of the target frame to have the same color statistics as the ground projection of the reference frame (or a portion thereof). In some embodiments, after the color statistics from the reference frame are transferred to an adjacent target frame, the process is repeated using the updated target frame as a new reference frame and transferring the color statistics of the ground projection from the new reference frame to the next adjacent target frame. In some embodiments involving 360° surround view visualization (or at least partial surround view visualization generated using stitched images from two or more image sensors), the process can be repeated to loop the frames, and any known computer vision loop closure technique can be applied to close the loop with the original reference frame, and / or any known global optimization technique can be applied to improve color coordination.

[0028] In some implementations, color coordination can be applied in the image processing pipeline, and at any stage in the pipeline. For example, in some embodiments including multiple cameras that capture different frames of image data, each camera can be controlled by its own capture algorithm that runs independently of those running for the other cameras. The image data frames can be processed using various types of image processing to generate processed image data frames, and the processed image data frames can be stitched together to form stitched image data (e.g., 360° look-around visualization, panorama, etc.). Depending on the embodiment, color coordination can be applied to the image data, processed image data, deprocessed image data, or stitched image data.

[0029] With respect to the de-processed image data, in some embodiments, some or all of the image processing used to generate the processed image data frames may be reversed to convert the processed image data frames into de-processed image data in a common (reference) light space of linear light, and color coordination may be applied to the de-processed image data in the reference light space. More specifically, each camera may be associated with a capture algorithm that determines one or more capture configuration parameters used to control the image sensor in the camera and perform image processing on the captured image data. Example capture configuration parameters include sensor exposure time, analog-to-digital gain, optoelectronic transfer function (OETF) curve (also known as gamma curve), tone mapping curve, automatic white balance (AWB) correction coefficients, and lens shading profile, to name a few non-limiting examples. The image sensor may capture image data (e.g., pixel values) in one or more color channels that, prior to image processing, are in a linear light space - meaning that the values ​​from the image sensor are proportional to the illuminance in the captured scene. Various types of image processing (e.g., gamma correction for improving color range, exposure compensation, tone mapping, noise reduction, removing bad pixels, applying white balance, applying color correction to remove lens shading artifacts in fisheye images, etc.) may be applied to the captured image data, which may involve one or more non-linear transformations. Thus, the image processing may transform the captured image data from a linear light space to a non-linear space.

[0030] Thus, the image processing may be inverted using the capture configuration parameters for each frame to transform the processed image data back into a linear light space (or some other reference light space, such as a reference light space that better approximates the illumination of the scene than the processed image data), and color coordination may be applied in the reference (e.g., linear) light space. Multiple stages of image processing may be inverted in reverse order (e.g., if tone mapping is applied at the end of the image processing pipeline, then inverse tone mapping may be applied first using an inverse curve). In some cases, one or more stages of image processing may be omitted from the inverse image processing (e.g., because it may not be possible or practical to invert some types of image processing (e.g., denoising or sharpening). As such, inverse image processing may be applied to transform the processed image data into de-processed image data in a reference light space, and color coordination may be applied to the de-processed image data in the reference light space. Applying color coordination across different frames in a common (reference) or linear light space may often improve the results of color coordination compared to applying color coordination across different frames in different or non-linear light spaces, because the common and linear light space is more conducive to transferring color statistics between frames.

[0031] After the colors on the de-processed image data have been coordinated, the corresponding image processing can be reapplied to the coordinated image data using the corresponding capture configuration parameters. In some embodiments, the re-application of a particular type of image processing can use a combination of capture configuration parameters that average, approximate, or otherwise combine capture configuration parameters used to generate different frames in a set of frames representing a common time slice. For example, assume that four different tone mapping curves are applied (and inverted) to four different frames of image data. To reapply tone mapping after color coordination, the four different tone mapping curves can be averaged, and the resulting average tone mapping curve can be applied to each frame of the coordinated image data.

[0032] In some embodiments that coordinate colors between a set of frames, a reference frame can be selected from the set in various ways. For example, a reference frame (e.g., an image) can be selected to correspond to the direction of ego motion (e.g., using an image from a forward-facing camera when moving forward), an active viewport (e.g., using an image from a camera pointing in the direction of the active viewport), or an operator gaze (e.g., using an image from a camera pointing in the direction of the operator's gaze pointing at the ego object). In some embodiments, images in which the brightest hue is not white can be invalidated from use as reference frames. For example, suppose an ego vehicle drives past a bright object such as a red fence. Applying auto white balance to an image of the red fence can lead to incorrect results because auto white balance typically assumes that the brightest hue is white, so it may not be desirable to copy color statistics from that image. In some embodiments, each image can be decomposed into multiple patches, one or more color statistics can be calculated for each patch, and the image with the most uniform patch (e.g., based on a measure of similarity of one or more color statistics thereof) can be selected as a reference frame. Selecting the image with the most uniform patch can indicate fewer objects and therefore more reliable and representative color information to transfer to other images. These are just a few examples, and other ways of selecting reference frames may be implemented within the scope of the present disclosure.

[0033] In some embodiments, a state machine can be used to determine whether and how to migrate color statistics from a reference frame to an adjacent target frame. More specifically, for each set of frames captured at substantially the same time or otherwise representing the same time slice, a reference frame and an adjacent target frame can be selected. Image data from each frame can be projected onto a three-dimensional (3D) representation of the ground (e.g., ground plane) of the environment. Some embodiments model the geometry of the vehicle's surroundings as a 3D bowl shape that includes a circular ground plane for an inner portion of the bowl that is connected to an outer bowl represented as a curved surface that rises from the ground plane to a height or slope that increases in proportion to the distance from the center of the bowl. In some such embodiments, image data from each frame is projected onto the 3D bowl, and portions of the image data projected onto the outer bowl are discarded, leaving portions of the circular inner ground projections from each frame (e.g., wedges). Regions of overlapping image data between the ground projections of the reference frame and the target frame can be identified, and it can be determined whether any pixels in the overlapping region belong to a detected object. For example, object detection may be performed using two-dimensional (2D) object detection (e.g., from an image) or 3D object detection (e.g., from a 3D point cloud detected from an image, LiDAR, or RADAR), a representation of the detected object (e.g., an object or a segmentation mask) may be projected onto a ground plane, and each pixel in the overlap region may be compared to a corresponding pixel of the projected representation of the detected object (e.g., the projected object and / or one or more segmentation masks generated from a reference frame and / or a target frame) to determine whether the pixel belongs to the detected object. If it is determined that there is no detected object in the overlap region (or less than or equal to a threshold number of detected object pixels, e.g., zero), the color statistics may be migrated from the ground projection of the reference frame to the target frame. If it is determined that there are more than a specified threshold number or percentage of pixels (e.g., 50%) that belong to the detected object in the overlap region, the ground projection of the reference frame from a previous time slice may be used to migrate the color statistics to the target frame. If there are less than a specified threshold number or percentage of points or pixels, those points or pixels can be removed from the ground projection for the target frame and the source frame, the remaining pixels from each ground projection can be clustered (e.g., using k-means clustering), and the majority cluster from each ground projection can be used to migrate the color statistics of the majority cluster of the ground projection from the reference frame to the target image.

[0034] In some embodiments, transferring color statistics from a reference frame of image data (e.g., a ground projection or a majority cluster thereof) to a target frame of image data (e.g., an entire target image) involves color statistics of both the ground projection of the reference frame (or a majority cluster thereof) and the ground projection of the target frame (or a majority cluster thereof). For example, for each pixel in the target frame (e.g., and each color channel), the mean color of the ground projection of the reference frame (or a majority cluster thereof) can be subtracted from the color of the pixel (e.g., to represent color variance or per-channel color variance), the color variance scaled by the ratio of the standard deviation of the color of the pixel in the ground projection of the reference frame (or a majority cluster thereof) to the standard deviation of the color of the pixel in the ground projection of the target frame (or a majority cluster thereof), and added to the mean color of the ground projection of the reference frame (or a majority cluster thereof). In another example, a color covariance matrix representing correlations between color channels can be transferred from the ground projection of the reference frame (or a majority cluster thereof) to the target frame. Migrating the color covariance matrix can produce better color migration than the prior art, for example, in scenarios where hue differences between frames are due to corresponding views being illuminated with varying illuminants, resulting in different auto-white balance coefficients. In general, any color statistics can be calculated and migrated using any suitable color space (e.g., RGB, YCbCr, IPT, CIELAB).

[0035] In contrast to conventional systems that migrate color statistics to the entire target frame, in some embodiments, the reference frame and the target frame of image data may be divided into multiple segments or columns, and the color statistics may be migrated from the segment or column to the adjacent segment or column. More specifically, the color statistics may be migrated from one segment or column (or its ground projection) to another segment or column within the reference frame, and then from the last segment or column of the reference frame to the first segment or column of the target frame at the boundary between frames. In this manner, the color statistics may be gradually migrated from the reference frame (e.g., its ground projection) to the target frame. In some embodiments involving a stitched 360° surround view visualization, the color statistics may be gradually migrated from the reference segment to the target segment corresponding to a ring around the stitched 360° surround view visualization. Migrating the color statistics in this manner may result in better color migration and prevent exposure above or below the stitched 360° surround view visualization. In some embodiments, to achieve smooth transitions, weighting may be applied more heavily at the boundaries between frames during color migration than at the center of the frame, which may result in the center of the frame representing more realistic image data for each camera with smooth transitions at the boundaries of the frames.

[0036] In some embodiments involving stitched 360° surround visualization, color statistics can be migrated from frame to adjacent frame, effectively cycling through image space (whether clockwise or counterclockwise). In some cases, migrating color statistics from frame to frame may result in some finite error, so that migrating the color statistics to subsequent frames acts as an accumulated error. As such, in some embodiments, any known computer vision loop closure technique may be performed to reduce the accumulated error. Additionally or alternatively, a global optimization may be performed to improve color coordination, such as applying one or more geometric or photometric transforms and cycling until one or more component quality scores are above or equal to a threshold, as described in U.S. patent application Ser. No. 17 / 139,587, filed on Dec. 31, 2020.

[0037] Various aspects of the present color coordination techniques provide numerous advantages and benefits over the prior art. For example, migrating color statistics from ground projections rather than from the entire image effectively improves color coordination of stitched images and thus improves visual quality of stitched images, compared to prior art techniques that migrate global color statistics from the entire image. More specifically, global color statistics of an image including an object effectively represent different lighting conditions in the same image, since objects are often illuminated differently from the ground. In contrast, since different ground projections are located in the same plane (ground plane), ground projections will generally represent more consistent lighting, and thus using color statistics from ground projections is generally more reliable and includes more representative color information to migrate to other images. In addition, using color statistics from ground projections of each frame, rather than using global color statistics of the entire frame, reduces computational requirements and thus reduces latency compared to the prior art. In addition, some embodiments facilitate improved parallelization by performing certain computations in parallel (e.g., computing ground projections for each frame, computing each overlap region between each pair of adjacent frames, computing color statistics from each ground projection or majority clusters), thereby further reducing latency compared to the prior art. Embodiments that utilize color statistics from previous time slices avoid the need to calculate color statistics for the current time slice, thereby further reducing latency compared to prior art techniques. Thus, the techniques described herein may be used to improve the visual quality of stitched images, reduce latency in generating coordinated stitched images, and thus facilitate safe operation of self-machines.

[0038] refer to Figure 1 , Figure 1is a schematic diagram illustrating an example data flow through an example surround view system (SVS) 100 according to some embodiments of the present disclosure. It should be understood that this and other arrangements described herein are set forth by way of example only. Other arrangements and elements (e.g., machines, interfaces, functions, sequences, functional groupings, etc.) may be used in addition to or in place of the arrangements and elements shown, and some elements may be omitted entirely. Further, many of the elements described herein are functional entities that may be implemented as discrete or distributed components or in combination with other components, and in any suitable combination and location. The various functions described herein as being performed by an entity may be performed by hardware, firmware, and / or software. For example, the various functions may be implemented by a processor executing instructions stored in a memory. In some embodiments, the same may be used. Figures 12A-12D Example of autonomous vehicle 1200, Fig.13 The example computing device 1300 of , and / or Fig.14 The systems, methods, and processes described herein may be implemented using components, features, and / or functionality similar to the example data center 1400.

[0039] At a high level, the example surround view system 100 generates a visualization 155 of a 3D environment (e.g., an environment around an ego object such as a vehicle) based on frames of image data 102 of the environment. As the ego object or ego actor navigates through the 3D environment, the frames of image data 102 may be viewed by the ego object or ego actor (e.g., FIG. 12A to FIG. 12D 1200 ). The image processing module 105 may process frames of the image data 102 using any number and type of known image processing techniques to generate frames of processed image data 110. The stitching module 120 may stitch the frames of the processed image data 110 into stitched image data 125 (e.g., a 360° look-around visualization, a panorama). The color coordinator 130 may coordinate colors across portions of the stitched image data 125 from different frames of the processed image data 110 to generate coordinated image data 150 (e.g., a coordinated 360° look-around visualization, a coordinated panorama). The presentation module 160 may cause a visualization 165 of at least a portion of the coordinated image data 150 to be presented (e.g., on a monitor visible to an occupant or operator of the ego-object or ego-actor). In some embodiments, rendering module 160 projects coordinated image data 150 onto a 3D representation of a 3D environment (e.g., a 3D bowl that models the 3D environment), renders a view of the projected coordinated image data from the perspective of a virtual camera, and causes the rendered view to be presented as visualization 165.

[0040] In general, color coordination may be performed on image data 102, processed image data 110, and / or stitched image data 125 from any number and type of cameras (e.g., one or more cameras 101), such as described below with respect to Figures 12A-12D Cameras 101 may include one or more cameras of self-objects or self-actors, such as, FIG. 12A to FIG. 12D The stereo camera 1268, wide angle camera 1270 (e.g., fisheye camera), infrared camera 1272, surround camera 1274 (e.g., 360° camera), and / or long-range and / or mid-range camera degrees 1298 of the ego vehicle 1200, and the camera 101 can be used to generate image data 102 of the 3D environment around the ego object or ego actor. In embodiments using multiple cameras, the multiple cameras can view a common area of ​​the 3D environment with overlapping portions of their respective fields of view, so that the image data 102 (e.g., images) from different cameras represent the common area. Each frame of the image data 102 can be processed by the image processing module 105 using any number or type of image processing techniques (e.g., gamma correction, exposure compensation, tone mapping, noise reduction, removal of bad pixels, application of white balance, application of color correction to remove lens shading artifacts in fisheye images) to generate a frame of processed image data 110. Frames of processed image data 110 may be stitched together by stitching module 120 (e.g., using any known stitching techniques and / or as described with respect to Fig.10 The splicing module 1006 is described in more detail). Figure 1 In the implementation shown in , color coordinator 130 applies color coordination to stitched image data 125, but this need not be the case. For example, color coordination may be applied to frames of processed image data 110 generated by image processing module 105 to generate coordinated frames of image data, and the coordinated frames of image data may be stitched together by stitching module 120. In general, color coordination may be performed at any stage of the image processing pipeline, depending on the implementation.

[0041] At a high level, the color coordinator 130 may include a frame identification module 135 that identifies or receives identification of a reference frame (e.g., an image from a particular camera or a portion of a stitched image) of image data from which color statistics should be migrated and a target frame (e.g., another image from a different camera or another portion of a stitched image) of image data that should be modified so that its color statistics match the color statistics of a ground projection of the reference frame (or portion thereof). Given the reference frame and the target frame, a projection module 140 of the color coordinator 130 may generate ground projections of the reference frame and the target frame, for example, by projecting the reference frame and the target frame onto corresponding portions of a ground plane of a 3D environment (e.g., a 3D bowl that models the 3D environment) represented by the reference frame and the target frame, and identifying a portion of each projection that falls in an overlap region between the two projections as the ground projection for each frame. The color coordination module 145 of the color coordinator 130 may calculate one or more reference color statistics (e.g., any statistical moment or characteristic of one or more color channels, such as mean, variance, standard deviation, kurtosis, skew, and / or correlation between color channels) for pixels in a ground projection of a reference frame (or a portion thereof), and migrate the one or more reference color statistics to the target frame by modifying the color of the target frame to match the one or more reference color statistics. In some embodiments where there is a series of overlapping frames or a loop of overlapping frames (e.g., a panoramic or 360° look-around visualization), the color coordinator 130 may repeat this process in consecutive (e.g., overlapping) pairs of frames in any direction to coordinate colors. In some embodiments involving 360° look-around visualization, the color coordinator 130 may perform any known computer vision loop closure techniques and / or global optimization to improve color coordination.

[0042] The frame identification module 135 may identify the reference frame in various ways. For example, in embodiments where the camera 101 is used to generate frames (e.g., frames of the image data 102, frames of the processed image data 110, different portions of the stitched image data 125) representing different views of the 3D environment around the ego object (e.g., a vehicle), the frame identification module 135 may select one of the frames as the reference frame corresponding to the direction of the ego motion (e.g., using image data corresponding to one of the cameras 101 that faces forward when moving forward), the active viewport viewing the 3D environment (e.g., using image data corresponding to one of the cameras 101 that points in the direction of the active viewport displayed on a monitor visible to an occupant or operator of the ego object), or the operator's gaze (e.g., using image data corresponding to one of the cameras 101 that points in the direction of the operator's gaze pointing at the ego object). In embodiments where color coordination is applied after stitching together different portions of the stitched image, the frame identification module 135 may determine which pixels of the stitched image come from which camera, and the frame identification module 135 may identify the reference frame and the target frame as pixels of the stitched image from the corresponding cameras. In some embodiments, the frame identification module 135 may determine not to select a frame in which the brightest hue is not white, in which case the frame identification module 135 may identify an alternative frame (e.g., an adjacent frame) as a reference frame. In some embodiments, the processed image data 110 generated by the image processing module 105 may identify multiple patches for each frame and represent color statistics calculated by the image processing module 105 for each patch, and the frame identification module 135 may select the frame with the most uniform patch (e.g., based on a measure of similarity of its color statistics) from a set of candidate frames (e.g., representing the same time slice) as the reference frame. These are just a few examples, and other ways of selecting a reference frame may be implemented within the scope of the present disclosure.

[0043] The projection module 140 may generate ground projections of the reference frame and the target frame. Typically, the reference frame and the target frame may each include a 2D representation of the 3D environment, and the projection module 140 may project each frame (or a portion thereof) onto a representation of the ground (e.g., a ground plane) of the 3D environment. In some embodiments, the projection module 140 may limit the ground projection to within a specific radius (e.g., three meters) of the self-object. For example, some embodiments may model the geometry of the 3D environment as a 3D bowl including a circular ground plane centered on the self-object and an outer bowl represented as a curved surface rising from the ground plane, and the projection module 140 may project each frame (or a portion thereof) onto the ground plane of the 3D bowl. In some embodiments, the projection module 140 may identify the portion of each projection that falls in the overlapping region between the two projections as the corresponding ground projection for each frame.

[0044] At a high level, the color modification module 145 may compute and transfer color statistics from the ground projection of a reference frame to a target frame. Figure 2 2 is a diagram illustrating an example technique for migrating reference color statistics 218 from a ground projection 215 of a reference frame 210 (or a portion thereof) to a target frame 220 according to some embodiments of the present disclosure. In this example, the reference frame 210 includes the ground projection 215, and the target frame 220 includes the ground projection 225. The reference color statistics 218 may be calculated from some or all pixels belonging to the ground projection 215 of the reference frame 210, and the initial color statistics 228 may be calculated from some or all pixels belonging to the ground projection 225 of the target frame 220. Figure 2 In the example shown in FIG. 2 , reference color statistics 218 from reference frame 210 and initial color statistics 228 from target frame 220 include per-channel means and standard deviations in RGB space. Example color statistics migration 230 shows an example technique for migrating reference color statistics 218 from ground projection 215 of reference frame 210 to target frame 220. Using example color statistics migration 230, each pixel R of target frame 220 is i , G i , B i The input color of is modified using the color statistics of each ground projection to become the output color R o , G o , B o , where ⊙ represents an element-by-element multiplication. This is just an example color statistics transfer technique, and other techniques may be implemented within the scope of the present disclosure. For example, a color covariance matrix representing the correlation between color channels may be transferred from a ground projection of a reference frame (or a majority cluster thereof) to a target frame. In general, color statistics may be calculated using any suitable color space (e.g., RGB, YCbCr, IPT, CIELAB).

[0045] Reference now Figure 3 , each block of the method 300 described herein includes a computing process that can be performed using any combination of hardware, firmware, and / or software. For example, the various functions can be implemented by a processor executing instructions stored in a memory. The method can also be embodied as computer-usable instructions stored on a computer storage medium. To name a few, the method can be provided by a stand-alone application, a service or a hosted service (standalone or in combination with other hosted services), or a plug-in for another product. In addition, the method 300 can be provided by Figure 1 The method is implemented by the color coordinator 130 of the example surround view system 100. However, this method may additionally or alternatively be performed by any one system or any combination of systems, including but not limited to the systems described herein.

[0046] Figure 3is a flow chart illustrating a method 300 for applying color coordination based at least on object detection according to some embodiments of the present disclosure. The method 300 may be used to coordinate colors within a set of consecutive frames of image data representing a 3D environment in consecutive time slices. Each set may include two or more frames of image data. In a non-limiting example, each set includes four fisheye images and / or a stitched image of four fisheye images generated using four cameras pointing in different directions (e.g., front, right, back, left) in the 3D environment. Any suitable technique may be used to select reference frames and target frames 302 (e.g., from a set of frames representing each time slice), such as those described herein (e.g., with respect to Figure 1 In embodiments where color coordination is applied after stitching into different portions of a stitched image, a determination may be made as to which pixels of the stitched image came from which camera, and thus, the reference frame and the target frame may be identified as pixels of the stitched image from corresponding cameras.

[0047] At block B 306 , the method 300 includes projecting the reference frame and the target frame 302 onto corresponding portions of the representation of the floor of the 3D environment using camera calibration parameters 304 representing the positions and orientations of the corresponding cameras in the 3D environment. Figure 1 , the projection module 140 may project each frame onto a 3D bowl that models the 3D environment to generate a projected 3D bowl representation of image data from the frame, discard a portion of the projected 3D bowl representation that corresponds to the outer bowl, and retain a portion of the projected 3D bowl representation that corresponds to the ground plane of the 3D environment.

[0048] At block B308, method 300 includes calculating an overlap area between the projections of the reference frame and the target frame. Figure 1 , the projection module 140 can identify the portion of each projection that falls in the overlapping region between the two projections as the corresponding ground projection for each frame, and use the portion from each frame that falls in the overlapping region as the ground projection for that frame. The ground projections can be represented as corresponding 2D (e.g., top-down) or 3D representations of the environment that overlap each other, forming a 2D or 3D overlapping region in which image data from the reference frame and the target frame are projected onto the same point (e.g., pixel, voxel). Taking the overlapping region in a 2D top-down representation of a 3D environment as an example, any given pixel in the overlapping region can represent different values ​​for different ground projections.

[0049] At decision block B310, the method 300 includes determining whether there are any detected objects represented in the overlapping region (e.g., based on objects or / and segmentation masks corresponding to either or both of the overlapping ground projections). For example, object detection may be performed using two-dimensional (2D) object detection (e.g., on the reference and target frames 302) or 3D object detection (e.g., on 3D projections of the reference and target frames 302, on 3D point clouds detected by LiDAR or RADAR), and Figure 1 The projection module 140 may project a representation of a detected object (e.g., one or more objects and / or segmentation masks) onto an overlap region (e.g., of a ground plane). In this manner, the projection module 140 may determine whether each point or pixel in the overlap region belongs to a detected object by determining whether a corresponding (e.g., pixel) value from one or more projected representations of the detected object (e.g., a projected object and / or one or more segmentation masks generated from a reference frame, from a target frame, from either a reference frame or a target frame, or from both a reference frame and a target frame) is part of the detected object. If it is determined that there is no detected object represented in the overlap region (or less than or equal to a threshold number of detected object pixels), the method 300 proceeds to block B312. If it is determined that there is a detected object represented in the overlap region, the method 300 proceeds to decision block B314.

[0050] At block B312, method 300 includes migrating color statistics from the reference frame to the target frame. Figure 1 , the color coordinator 130 may calculate and transfer color statistics from the ground projection of the reference frame to the target frame. In some embodiments where there is a series of overlapping frames or a ring of overlapping frames (e.g., a panoramic or 360° look-around visualization), the color coordinator 130 may repeat this process across consecutive pairs of frames (e.g., overlapping frames) in any direction to coordinate colors.

[0051] As an illustration, Figure 4 shows visualization of a stitched image before (410) and after (450) applying color coordination using the entire reference frame and target frame to calculate color statistics, according to some embodiments of the present disclosure, and Figure 5 Visualization of a stitched image before (510) and after (550) applying color coordination using ground projections of reference and target frames to calculate color statistics in accordance with some embodiments of the present disclosure is shown. Figure 4 , the uncoordinated stitched image 410 includes four frames (a front frame 415, a right frame 420, a rear frame 425, and a left frame 430), and the coordinated stitched image 450 includes four frames (a coordinated front frame 455, a coordinated right frame 460, a coordinated rear frame 465, and a coordinated left frame 470). Figure 4In each case using color statistics calculated over the entire reference frame and target frame for each pair of frames, colors may be coordinated by migrating color statistics clockwise around the uncoordinated stitched image 410 (e.g., from the front frame 415 to the right frame 420 to generate the coordinated right frame 460, from the coordinated right frame 460 to the back frame 425 to generate the coordinated back frame 465, and from the coordinated back frame 465 to the left frame 430 to generate the coordinated left frame 470). As a result, the colors of the coordinated stitched image 450 match better than the colors of the uncoordinated stitched image 410. In another implementation, Figure 5 represents the color coordination of the uncoordinated stitched image 510 using color statistics calculated on the ground projections of the reference and target frames, rather than calculating color statistics on the entire reference and target frames. Figure 5 The color ratio at the boundary between the coordinated frames 555, 560, 565 and 570 of the coordinated stitched image 550 is Figure 4 The colors in the coordinated stitched image 450 match better.

[0052] Back to Figure 3 If it is determined at decision block B310 that there is a detected object represented in the overlapped region, the method 300 proceeds to decision block B314, which includes: determining whether the count or percentage of points (e.g., pixels) (object points or pixels) in the overlapped region and belonging to the detected object exceeds a certain threshold (e.g., percentage, number of pixels). Taking the overlapped region between two 2D top-down ground projections (one ground projection of the reference frame and one ground projection of the target frame) as an example, Figure 1 The color coordinator 130 may determine whether each pixel in the overlap region belongs to the detected object by determining whether a corresponding pixel value from a projected representation of the detected object (e.g., a projected object and / or segmentation mask generated from a reference frame and / or a target frame) is part of the detected object, determining a count or percentage of pixels that belong to the detected object (e.g., based on an overlapped projected mask corresponding to either, one, or both of the reference frame and / or the target frame, and determining whether the count or percentage exceeds a specified threshold (e.g., 50% of pixels in the overlap region). If it is determined that the count or percentage of points (object points or pixels) that are in the overlap region and belong to the detected object exceeds the threshold, the method 300 proceeds to block B316. Otherwise, the method 300 proceeds to block B318.

[0053] If it is determined at decision block B314 that the count or percentage of points that are in the overlap region and belong to the detected object exceeds a specified threshold, the method 300 proceeds to block B316, which includes: migrating the color statistics from (the ground projection of) the reference frame from the previous time slice to the target frame from the current time slice. Figure 1 , the color coordinator 130 may calculate (or access previously calculated) reference color statistics for the ground projection of the reference frame from the previous time slice, and migrate the reference color statistics to the target frame of the current time slice.

[0054] As an illustration, Figure 6 An example overlap region between two ground projections is shown, wherein the amount of pixels belonging to a detected object in the overlap region exceeds a threshold value, according to some embodiments of the present disclosure. Figure 6 , the uncoordinated stitched image 610 includes a front frame 620 and a left frame 630. The front frame 620 and the left frame 630 are projected onto the ground plane of a 3D bowl that models the 3D environment, and the overlapping portions are taken to produce portions of these projections of ground projection 640 and ground projection 650, respectively. In this example, most of the pixels in both ground projection 640 and ground projection 650 represent neighboring vehicles, so a corresponding object mask can be projected and used to determine that the number of pixels in the overlapping area corresponding to ground projection 640 and ground projection 650 exceeds a certain threshold (e.g., more than 50% of the pixels). As a result, a determination can be made not to use either the front frame 620 or the left frame 630 as a reference frame. In some cases (e.g., in Figure 1 In some embodiments, the frame identification module 135 is configured to select a reference frame corresponding to the direction of self-motion, but it is determined that the reference frame is not to be used, the reference frame from the previous time slice can be used to migrate the color statistics to the target frame of the current time slice. By way of explanation, in some embodiments, during relatively short durations (such as the duration between consecutive time slices), it can be assumed that the surrounding environment is unlikely to change significantly, so relying on color statistics from the previous time slice should be possible. In addition, using the reference frame from the previous time slice in cases like this helps maintain the temporal stability of color coordination over a set of consecutive frames.

[0055] return Figure 3 If, at decision block B314, it is determined that the count or percentage of points (e.g., pixels) that are in the overlapped region and that belong to the detected object does not exceed a specified threshold (e.g., meaning that there may be some object pixels and some ground pixels), then method 300 proceeds to block B318, which includes: identifying common points (e.g., pixels) shared by both the overlapping ground projections of the reference frame and the target frame and passing the common pixels from each overlapping ground projection (e.g., its majority cluster) to block B312 for calculation and migration of color statistics. For example, with respect to Figure 1, the color coordinator 130 may use the projected objects and / or the segmentation mask to remove points (e.g., pixels) belonging to the detected objects from the corresponding ground projections, cluster the remaining points (e.g., using k-means clustering), and identify a majority cluster of the remaining points. Because block B318 occurs after determining that there are at least some points belonging to the detected objects but less than a specified threshold (e.g., 50%), an assumption may be made that the majority cluster from each ground projection is formed by common pixels belonging to the ground. In another example, the color coordinator 130 may compare the ground projections using some measure of image quality or similarity (e.g., by generating a structural similarity or an “SSIM” quality map representing an SSIM value of structural similarity for each pixel), may identify points (e.g., pixels) having a measure of image quality or similarity above a specified threshold, and may use those points (or their majority clusters) as common pixels.

[0056] By showing that Fig. 7A An example overlap region between two ground projections is shown, wherein the amount of pixels in the overlap region and belonging to a detected object is less than a threshold value, according to some embodiments of the present disclosure. Fig. 7A , an uncoordinated stitched image 710 includes a front frame 720 and a right frame 730. The front frame 720 and the left frame 730 are projected onto a ground plane of a 3D bowl that models a 3D environment, and the overlapping portions are taken to produce portions of these projections of a ground projection 740 and a ground projection 750, respectively. In this example, the overlapping region between the ground projection 740 and the ground projection 750 includes less than a certain threshold number or count of pixels belonging to the detected object. As a result, the pixels in each of the ground projection 740 and the ground projection 750 can be clustered (e.g., using k-means clustering). Figure 7B An example ground projection with clustered pixels is shown in accordance with some embodiments of the present disclosure. Figure 7B , a ground projection 740 is shown on the left, and a clustering result 760 is shown on the right. In this example, the clustering result 760 includes two clusters, a majority cluster 770 and a minority cluster 780.

[0057] Therefore, and returning to Figure 3 , common pixels identified from each ground projection (e.g., majority clusters from each ground projection) may be identified at block B318. Method 300 may then return to block B312, which includes: migrating color statistics of common pixels (e.g., majority clusters) of the ground projections of the reference frame to the target frame. Figure 1, the color coordinator 130 may use the identified pixels from each ground projection to calculate relevant color statistics for the reference frame and the target frame. As such, the color coordinator 130 may calculate color statistics for the majority clusters of the ground projections of the reference frame and the target frame, and use these color statistics to modify the color of the target frame to match the color statistics for the majority cluster of the ground projection of the reference frame. The method 300 may be repeated to coordinate colors between a set of frames representing a given time slice and / or for consecutive sets of frames representing consecutive time slices.

[0058] Now refer to Figure 8 , each block of the method 800 described herein includes a computing process that can be performed using any combination of hardware, firmware, and / or software. For example, the various functions can be implemented by a processor executing instructions stored in a memory. The method can also be embodied as computer-usable instructions stored on a computer storage medium. To name a few, the method can be provided by a stand-alone application, a service or a hosted service (standalone or in combination with other hosted services), or a plug-in for another product. In addition, the method 800 can be provided by Figure 1 The method is implemented by the color coordinator 130 of the example surround view system 100. However, this method may additionally or alternatively be performed by any one system or any combination of systems, including but not limited to the systems described herein.

[0059] Figure 8 8 is a flow chart illustrating a method 800 for color coordination using a ground projection of a reference frame according to some embodiments of the present disclosure. At block B802, the method 800 includes determining a reference frame and a target frame from frames of image data representing overlapping views of an environment surrounding an ego object. For example, Figure 1 , the frame identification module 135 may be provided with a set of consecutive frames of image data representing the 3D environment in consecutive time slices, and the frame identification module 135 may identify, from the set of frames for each time slice, a reference frame (e.g., an image from a particular camera or a portion of a stitched image) from which color statistics should be migrated, and a target frame (e.g., another image from a different camera or another portion of a stitched image) that should be modified so that its color statistics match the color statistics of a ground projection of the reference frame (or a portion thereof). The reference frames (e.g., images) may be selected to correspond to the direction of ego motion (e.g., using an image from a forward-facing camera when moving forward), an active viewport (e.g., using an image from a camera pointing in the direction of the active viewport), an operator gaze (e.g., using an image from a camera pointing in the direction of the operator's gaze pointing at the ego object), and / or other means. A frame having a view that overlaps with a selected reference frame may be selected as a target frame.

[0060] At block B804, method 800 includes generating a ground projection of a reference frame. Figure 1, the projection module 140 can project the reference frame and the target frame onto corresponding portions of a ground plane of a 3D environment (e.g., a 3D bowl that models the 3D environment) represented by the reference frame and the target frame, identify an overlapping area between these projections, and designate the portion of the projection of the reference frame that falls in the overlapping area as a ground projection of the reference frame.

[0061] At block B806, method 800 includes calculating reference color statistics from a ground projection of a reference frame. Figure 1 The color coordination module 145 may calculate one or more reference color statistics (e.g., mean, standard deviation, per-color channel mean, per-color variance, correlation between color channels) for pixels in the ground projection of the reference frame (or a portion thereof, such as a majority cluster), and may calculate one or more initial color statistics for pixels in the ground projection of the target frame.

[0062] At block B808, method 800 includes migrating reference color statistics from the ground projection of the reference frame to the target frame. Figure 1 , color coordination module 145 may migrate one or more reference color statistics from the ground projection of the reference frame to the target frame by modifying the colors of the target frame to match the reference color statistics (e.g., scaled by a ratio of color statistics of pixels from the ground projection of the reference frame to color statistics of pixels of the ground projection of the target frame).

[0063] Now turn to Fig. 9 , Fig. 9 is a schematic diagram illustrating an example data flow of an example surround view system 900 by applying color coordination to de-processed image data according to some embodiments of the present disclosure. At a high level, the example surround view system 900 generates a visualization 965 of a 3D environment (e.g., a 3D environment around an ego object such as a vehicle) based on frames of captured image data 902 of the environment. As the ego object or ego actor navigates through the 3D environment, the ego object or ego actor (e.g., FIG. 12A to FIG. 12DOne or more cameras 901 of an autonomous vehicle 1200 (of an autonomous vehicle 1200) capture frames of captured image data 902. Image processing module 905 may process frames of captured image data 902 using any number and type of known image processing techniques to generate frames of processed image data 910. Image processing inversion module 915 may generate de-processed image data 920 by inverting one or more stages or types of image processing applied by image processing module 905. In scenarios where de-processed image data 920 includes values ​​(e.g., pixel values) generated by inverting one or more stages or types of image processing, de-processed image data 920 may be considered inverted image data. Color coordinator 930 may coordinate colors across frames of de-processed image data 920 to generate frames of coordinated image data 935. Image reprocessing module 940 may generate frames of coordinated processed image data 945 by reapplying one or more stages or types of image processing inverted by image processing inversion module 915. The stitching module 950 may stitch the frames of the coordinated image data 945 into stitched image data 955 (e.g., a 360° look-around visualization, a panorama), and the rendering module 960 may cause a visualization 965 (e.g., on a monitor visible to an occupant or operator of the ego-object or ego-actor) to be rendered as a representation of at least a portion of the stitched image data 955. In some embodiments, the rendering module 960 projects the stitched image data 955 onto a 3D representation of a 3D environment (e.g., a 3D bowl modeling the 3D environment), renders a view of the projected image data from the perspective of a virtual camera, and causes the rendered view to be rendered as the visualization 965.

[0064] The example camera 901 includes an image sensor (e.g., a complementary metal oxide semiconductor or "CMOS" sensor) that captures pixel values ​​in one or more color channels. The camera 901 may include or be associated with a capture algorithm that determines one or more capture configuration parameters (e.g., sensor exposure time, analog gain, OETF or gamma curve, tone mapping curve, AWB correction coefficients, lens shading profile, etc.). Capture configuration parameters such as sensor exposure time and analog gain may be used to control the image sensor and generate a frame of captured image data 902 that may be in a linear light space, meaning that the values ​​captured by the image sensor are proportional to the illumination in the captured scene.

[0065] The image processing module 905 may process the frames of captured image data 902 using any number and types of known image processing techniques and capture configuration parameters (e.g., OETF or gamma curves, tone mapping curves, AWB correction coefficients, lens shading profiles, etc.) to generate frames of processed image data 910. More specifically, because the color space of the image sensor may not correspond to the color space of the downstream display, for example, due to different color gamuts (e.g., sensors typically have wider color gamuts) or spectral responses, the image processing module 905 may apply color space correction (CSC), which may include a linear transformation from the sensor color space to the display color space (e.g., using a 3x3 CSC matrix that transforms the sensor RBG space to the display RBG space, to use an example color space). The CSC may include white balance correction (AWB), which attempts to remove unrealistic color casts based on the estimated temperature of the illuminants in the scene. Additionally or alternatively, the image processing module 905 may apply tone mapping by applying a non-linear curve to the luminance data (e.g., to enhance shadows and compress highlights to approximate how our eyes work), which may be used to map a high dynamic range representation to a compressed representation (e.g., to facilitate downstream processing). These are just a few examples, and other types of image processing may be applied additionally or alternatively.

[0066] Since each of the one or more cameras 901 may have its own capture algorithm that runs independently of those running for the other cameras 901, different capture configuration parameters may be used to generate frames of processed image data 910 for different cameras 901. Thus, different frames of processed image data 910 may have been generated using different types of non-linear transformations, and the frames of processed image data 910 may not be in the same light space, and may not be in a linear light space. However, by reversing the image processing performed on each frame of processed image data 910, the frames may be converted back into a common reference space of linear light (or a common reference space that better approximates the illumination of the scene than processed image data 920).

[0067] Accordingly, image processing inversion module 915 may generate deprocessed image data 920 by inverting one or more stages or types of image processing applied by image processing module 905 to the corresponding frame of captured image data 902. To invert any particular type or stage of image processing, image processing inversion module 915 may access the corresponding capture configuration parameters that were applied to generate processed image data 910, invert the capture configuration parameters, and apply the inverted capture configuration parameters to processed image data 910 to generate deprocessed image data 920. Image processing inversion module 915 may invert multiple stages or types of image processing in the reverse order to that applied by image processing module 905 (e.g., if image processing module 905 applies tone mapping as the last stage, image processing inversion module 915 may first apply the inverse tone mapping by applying the inverse of the tone mapping curve to processed image data 910). Some stages of the image processing applied by the image processing module 905 may be omitted from the inverse image processing applied by the image processing inversion module 915 (e.g., because it may be impossible or impractical to invert some types of image processing (e.g., denoising or sharpening). As such, the image processing applied by the image processing module 905 may be inverted using the corresponding capture configuration parameters for each frame to transform each frame of processed image data 920 from a non-linear light space back into a (substantially) linear light space, or otherwise migrate multiple frames of processed image data 920 into a common reference light space.

[0068] Color coordinator 930 may use any of the techniques described herein to coordinate colors across frames of deprocessed image data 920 to generate frames of coordinated image data 935. For example, color coordinator 930 may correspond to color coordinator 130 and perform corresponding functions to generate frames of coordinated image data 935. Applying color coordination across different frames in a common (reference) space or (substantially) linear light space tends to improve the results of color coordination compared to applying color coordination across different frames in different or non-linear light spaces, because the common and linear light space is more conducive to transferring color statistics between frames.

[0069] Image reprocessing module 940 may reapply one or more stages of image processing reversed by image processing reverse module 915 to generate a frame of coordinated processed image data 945. Fig. 9, the image reprocessing module 940 is shown as a separate component from the image processing module 905, but this need not be the case, as the image reprocessing module 940 may include, correspond to, or utilize some or all of the image processing modules 905. In some embodiments, a particular frame of the coordinated processed image data 945 may be processed using the same capture configuration parameters (e.g., a particular tone mapping curve) used by the image processing module 905, and may be inverted by the image processing inversion module 915. In some embodiments, the image reprocessing module 940 may reapply a particular stage or type of image processing by determining and applying a combined capture configuration parameter that averages, approximates, or otherwise combines the capture configuration parameters used to generate different frames in the set of frames representing a common time slice. For example, assume that four different tone mapping curves are applied (and inverted) to four different frames of the captured image data 902. To reapply tone mapping, the four different tone mapping curves may be averaged, and the resulting average tone mapping curve may be applied to each respective frame of the coordinated processed image data 945. As such, image reprocessing module 940 may generate a frame of coordinated processed image data 945 by reapplying one or more stages of image processing applied by image processing module 905 (eg, in the same order as image processing module 905). Fig. 9 An example implementation is shown in which the color harmonizer 930 is applied before the image reprocessing module 940, but in some embodiments, the color harmonizer 930 may be applied after the image reprocessing module 940. As such, frames of harmonized processed image data 945 may be stitched together by a stitching module 950, and representations thereof may be rendered by a rendering module 960.

[0070] See now Fig.10 , each block of the method 1000 described herein includes a computing process that can be performed using any combination of hardware, firmware, and / or software. For example, the various functions can be implemented by a processor executing instructions stored in a memory. The method can also be embodied as computer-usable instructions stored on a computer storage medium. To name a few, the method can be provided by a stand-alone application, a service or a hosted service (standalone or in combination with another hosted service), or a plug-in for another product. In addition, the method 1000 can be provided by Fig. 9 The example surround view system 900 is implemented. However, this method may additionally or alternatively be performed by any one system or any combination of systems, including but not limited to the systems described herein.

[0071] Fig.101 is a flow chart illustrating a method 1000 for applying color coordination to processed image data according to some embodiments of the present disclosure. At block B1002, the method 1000 includes: processing the captured image data into processed image data using a particular type of image processing. For example, with respect to Fig. 9 , the image processing module 905 may process the frames of captured image data 902 using any number and type of known image processing techniques to generate frames of processed image data 910. Example image processing may include gamma correction for improving color range, exposure compensation, tone mapping, noise reduction, removing bad pixels, applying white balance, applying color correction to remove lens shading artifacts in fisheye images, and / or the like.

[0072] At block B1004, method 1000 includes converting the processed image data into unprocessed image data based at least on applying the inverse of the particular type of image processing. Fig. 9 , the image processing inversion module 915 may generate a frame of de-processed image data 920 by inverting one or more stages of image processing (e.g., gamma correction for improving color range, exposure compensation, tone mapping, AWB, CSC) applied by the image processing module 905 to the corresponding frame of the captured image data 902. To invert any particular type or stage of image processing, the image processing inversion module 915 may access the corresponding capture configuration parameters (e.g., OETF or gamma curve, tone mapping curve, AWB correction coefficients, CSC matrix, lens shading profile) used to generate the processed image data 910, and the image processing inversion module 915 may invert the capture configuration parameters and apply the inverted capture configuration parameters to the processed image data 910 to generate the de-processed image data 920.

[0073] At block B1006, method 1000 includes generating coordinated image data by applying color coordination to the frames of de-processed image data. For example, Fig. 9 Color Coordinator 930 or Figure 1 The color coordinator 130 may coordinate colors across frames of the processed image data 920 to generate frames of coordinated image data 935. More specifically, the color coordinator 930 may include: a frame identification module (e.g., corresponding to Figure 1 Frame identification module 135), which identifies or receives the identification of the reference frame and the target frame to be processed image data 920; projection module (e.g., corresponding to Figure 1 A projection module 140) that generates a ground projection of the reference frame and the target frame; and a color coordination module (e.g., corresponding to Figure 1A color coordination module 145) is provided that computes one or more reference color statistics (eg, any statistical moments or characteristics of one or more color channels) and transfers them from the ground projection of the reference frame to the target frame.

[0074] At block B1008, method 1000 includes processing the coordinated image data based at least on reapplying the particular type of image processing. Fig. 9 , image reprocessing module 940 may generate frames of coordinated processed image data 945 by reapplying one or more stages or types of image processing that were inverted by image processing inversion module 915. In some embodiments, image reprocessing module 940 may reapply a particular stage or type of image processing by determining and applying a combined capture configuration parameter that averages, approximates, or otherwise combines capture configuration parameters used to generate different frames in a set of frames representing a common time slice (e.g., an average tone mapping curve that averages different tone mapping curves originally applied to different frames of captured image data 902 by image processing module 905).

[0075] Example Image Stitching Technique

[0076] FIG. 11A to FIG. 11B An example of image stitching according to some embodiments of the present disclosure is shown. Fig.11A An example of a method that can be used by an autonomous vehicle (e.g., Figures 12A-12D 1200) or other system or device. In this example, the cameras have different but partially overlapping fields of view so that the images can be stitched together without gap filling or additional image data generation, but it should be understood that there may be other situations where the cameras do not produce images with at least partially overlapping views (which may require such a task). In this example, each camera may have various intrinsic or extrinsic values ​​that may affect the appearance of the captured images, where those values ​​may relate to field of view, optical center, focal length, or camera pose, and other such options.

[0077] As mentioned, it may be desirable to generate a single, consistent view of the surrounding environment based at least in part on the captured images 1130. This may include, for example, generating Fig. 11BThe composite image 1150 shown. As shown, whether it is a complete 360° view or at least a portion of the angular range, such a composite image can provide a single, consistent representation of the environment. Such a view can be presented as a single view showing the entire image 1150, or portions of the image can be displayed at different times, where the view can be controlled by the user. In this example, an angular alpha map can be used, rather than, for example, a vertical alpha map, at least to avoid the alpha discontinuity problem observed in various traditional stitching techniques. Blended area color continuity can be improved by extending the blended area to hide color discontinuities around corners or other such features.

[0078] Fig. 11C 1 is a schematic diagram illustrating an example data flow through an example image stitching system 1100 according to some embodiments of the present disclosure. In this example, a set of camera images 1102 may be received as input and may be obtained from a given device or system (e.g., Figures 12A-12D In one embodiment, multiple cameras associated with or positioned relative to a given environment (eg, autonomous vehicle 1200) are received, as well as other such options. In at least one embodiment, the images represent two or more different views of the environment that overlap at most. The images may represent a complete or partial view of a scene, location, or environment. The image data may include "live" data that is streamed or transmitted shortly after the image is captured, or may include image data that was previously captured and provided in a more offline manner.

[0079] In this example, the camera images may be provided to a view generator 1104, which may use these input camera images 1102 to generate an output view or composite image or video stream for presentation via at least one display 1124 (e.g., a monitor, projector, or wearable display, among other such options). In this example, the camera images 1102 may be provided to a stitching module 1106, which will attempt to stitch the multiple images together to generate a composite representation. The stitching module may use any of a variety of different stitching or compositing algorithms, such as various blending or other image manipulation techniques that may be performed. In at least one embodiment, the stitching module may use various intrinsic and extrinsic parameters of the cameras, at least to some extent, whose values ​​are known and available from a camera database 1108 or other such location, in order to appropriately allocate image data for synthesis. This may include, for example, information such as the relative pose or orientation of the different cameras, so that at least an initial stitching position and orientation may be determined for each image. It should be understood that for a view generator system that receives a sequence of images or a stream of video frame data, the images or frames to be synthesized may be images or frames corresponding to or captured at approximately the same point in time, at least to the extent that such capture may be synchronized or otherwise represent the same time slice.

[0080] The camera calibration parameters can be used to map the camera view into the stitching space, such as can correspond to a top-down view or a "bowl" view in the projection space. This projection can be used to identify any overlapping areas between adjacent cameras, where one or more blending algorithms can be used to blend at least some of the pixels to make the stitching less visible or less obvious. The stitching module 1106 can use the values ​​of one or more stitching parameters, such as can be stored in the parameter database 1110 or other such location. The values ​​of these parameters can determine aspects of how the component images are stitched together, such as can involve weighting or positions and other such aspects for blending. Example stitching parameters include, but are not limited to, blending methods (e.g., alpha blending or multi-band blending), blending width, blending alpha map shape (e.g., angle-based or vertical-based), seam types (e.g., diagonal seams, vertical seams, or horizontal seams), and seam locations. The stitching module 1106 can use the values ​​of these various parameters together with the input component images to generate a composite image or stitched image 1112 that provides a single representation of the environment, scene, or location at a point in time (such as the "current" point in time), thereby accounting for some amount of latency in transmission and processing.

[0081] If the scene is not coordinated between the cameras, it may be desirable to use a larger blending weight or radius (e.g., 200) to provide a smoother transition between the data from the images. For example, if it is a highly structured scene with many buildings and edges, it may be desirable to use a smaller blending weight (e.g., 2) to avoid ghosting and other artifacts due to misalignment between the cameras. Single-band blending may be used, where images are blended in only one band (or subset) of multiple bands. In at least one embodiment, the component images are decomposed into different bands or components, and different blending weights may be used for each of these bands or components.

[0082] In an attempt to provide a stitched image of high subjective quality, a certain amount of processing of the stitched image may be performed in an attempt to assess the quality, and use the results of that assessment to make any changes to stitching parameters that may be desirable to improve the quality, at least where the determined quality is below a target value or threshold or determination. In this example, both the stitched image 1112 from the stitching module 1106 and the component images 1114 used to generate the stitched image may be used for the quality assessment determination. In some embodiments, the component images may correspond directly to the input camera images 1102, while in some embodiments, the component images may have had at least some amount of processing performed on them, such as to reduce variations in brightness, color, or contrast, or to reduce the presence of noise or remove image artifacts, among other such options. In at least one embodiment, removing or reducing the presence of image artifacts in the individual component images prior to stitching may result in a higher quality stitched image.

[0083] In this example, stitched image 1112 and component images 1114 may be processed to produce image data that better provides one or more specific types of comparisons. In at least one embodiment, this may include utilizing a high pass filter 1116 (or edge or feature detector) on the image to enhance or identify edges or other significant features in the image. In some embodiments, quality assessment 1122 may be performed using any known technique, and if the quality assessment measure is below a specified threshold, the image data may be optimized in a loop, for example, by applying one or more geometric or photometric transformations and feeding the transformed image data back to stitching module 1106.

[0084] The systems and methods described herein may be used by, but are not limited to, non-autonomous vehicles, semi-autonomous vehicles (e.g., in one or more adaptive driver assistance systems (ADAS)), manned and unmanned robots or robotic platforms, warehouse vehicles, off-road vehicles, vehicles coupled to one or more trailers, flying boats, boats, shuttles, emergency response vehicles, motorcycles, electric or motorized bicycles, aircraft, engineering vehicles, underwater vehicles, drones, and / or other vehicle types. Further, the systems and methods described herein may be used for a variety of purposes, including, by way of example and not limitation, for machine control, machine motion, machine driving, synthetic data generation, model training, perception, augmented reality, virtual reality, mixed reality, robotics, safety and surveillance, simulation and digital twins, autonomous or semi-autonomous machine applications, deep learning, environmental simulation, object or actor simulation and / or digital twins, data center processing, conversational AI, light transport simulation (e.g., ray tracing, path tracing, etc.), collaborative content creation of 3D assets, cloud computing, and / or any other suitable application.

[0085] The disclosed embodiments may be included in a variety of different systems, such as automotive systems (e.g., control systems for autonomous or semi-autonomous machines, perception systems for autonomous or semi-autonomous machines), systems implemented using robots, aerial systems, medical systems, boating systems, smart area monitoring systems, systems for performing deep learning operations, systems for performing simulation operations, systems for performing digital twin operations, systems implemented using edge devices, systems including one or more virtual machines (VMs), systems for performing synthetic data generation operations, systems implemented at least in part in a data center, systems for performing conversational AI operations, systems for performing light transport simulations, systems for performing collaborative content creation of 3D assets, systems implemented at least in part using cloud computing resources, and / or other types of systems.

[0086] Example Autonomous Vehicle

[0087] Fig. 12Ais an illustration of an exemplary autonomous vehicle 1200 according to some embodiments of the present disclosure. Autonomous vehicle 1200 (alternatively referred to herein as "vehicle 1200") may include, but is not limited to, a passenger vehicle such as a car, a truck, a bus, a first response vehicle, a shuttle, an electric or motorized bicycle, a motorcycle, a fire truck, a police vehicle, an ambulance, a boat, a construction vehicle, an underwater vessel, a robotic vehicle, a drone, an airplane, a vehicle coupled to a trailer (e.g., a semi-tractor-trailer truck for hauling cargo), and / or another type of vehicle (e.g., a vehicle that is unmanned and / or accommodates one or more passengers). Autonomous vehicles are generally described in terms of automation levels defined by the National Highway Traffic Safety Administration (NHTSA), a division of the U.S. Department of Transportation, and the Society of Automotive Engineers (SAE) “Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles” (Standard No. J3016-201806, issued on June 15, 2018, Standard No. J3016-201609, issued on September 30, 2016, and previous and future versions of the standard). The vehicle 1200 may be capable of implementing one or more of Levels 3-5 of autonomous driving levels. For example, depending on the embodiment, the vehicle 1200 may be capable of driver assistance (Level 1), partial automation (Level 2), conditional automation (Level 3), high automation (Level 4), and / or full automation (Level 5). The term “autonomous” as used herein may include any and / or all types of autonomy of the vehicle 1200 or other machines, such as full autonomy, high autonomy, conditional autonomy, partial autonomy, providing assisted autonomy, semi-autonomous, primarily autonomous, or other designations.

[0088] The vehicle 1200 may include components such as a chassis, a body, wheels (e.g., 2, 4, 6, 8, 18, etc.), tires, axles, and other components of the vehicle. The vehicle 1200 may include a propulsion system 1250, such as an internal combustion engine, a hybrid power plant, an all-electric engine, and / or another type of propulsion system. The propulsion system 1250 may be connected to a drive train of the vehicle 1200, which may include a transmission, to achieve propulsion of the vehicle 1200. The propulsion system 1250 may be controlled in response to receiving a signal from a throttle / accelerator 1252.

[0089] A steering system 1254, which may include a steering wheel, may be used to steer the vehicle 1200 (e.g., along a desired path or route) when the propulsion system 1250 is operating (e.g., when the vehicle is in motion). The steering system 1254 may receive signals from a steering actuator 1256. For fully automated (Level 5) functionality, a steering wheel may be optional.

[0090] Brake sensor system 1246 may be used to operate vehicle brakes in response to receiving signals from brake actuator 1248 and / or brake sensors.

[0091] May include one or more system on chip (SoC) 1204 ( Fig. 12C ) and / or one or more controllers 1236 of one or more GPUs may provide signals (e.g., representing commands) to one or more components and / or systems of the vehicle 1200. For example, the one or more controllers may send signals to operate vehicle brakes via one or more brake actuators 1248, to operate the steering system 1254 via one or more steering actuators 1256, and to operate the propulsion system 1250 via one or more throttles / accelerators 1252. The one or more controllers 1236 may include one or more onboard (e.g., integrated) computing devices (e.g., supercomputers) that process sensor signals and output operating commands (e.g., signals representing commands) to enable autonomous driving and / or assist a human driver in driving the vehicle 1200. The one or more controllers 1236 may include a first controller 1236 for autonomous driving functions, a second controller 1236 for functional safety functions, a third controller 1236 for artificial intelligence functions (e.g., computer vision), a fourth controller 1236 for infotainment functions, a fifth controller 1236 for redundancy in emergency situations, and / or other controllers. In some examples, a single controller 1236 may handle two or more of the above functions, two or more controllers 1236 may handle a single function, and / or any combination thereof.

[0092] The one or more controllers 1236 may provide signals for controlling one or more components and / or systems of the vehicle 1200 in response to sensor data (e.g., sensor input) received from one or more sensors. The sensor data may be received from, for example and without limitation, a global navigation satellite system ("GNSS") sensor 1258 (e.g., a global positioning system sensor), a RADAR sensor 1260, an ultrasonic sensor 1262, a LIDAR sensor 1264, an inertial measurement unit (IMU) sensor 1266 (e.g., an accelerometer, a gyroscope, a magnetic compass, a magnetometer, etc.), a microphone 1296, a stereo camera 1268, a wide angle camera 1270 (e.g., a fisheye camera), an infrared camera 1272, a surround camera 1274 (e.g., a 360° camera), a long-range and / or mid-range camera 1298, a speed sensor 1244 (e.g., for measuring the velocity of the vehicle 1200), a vibration sensor 1242, a steering sensor 1240, a brake sensor (e.g., as part of a brake sensor system 1246), and / or other sensor types.

[0093] One or more of the controllers 1236 may receive input (e.g., represented by input data) from the instrument cluster 1232 of the vehicle 1200 and provide output (e.g., represented by output data, display data, etc.) via a human machine interface (HMI) display 1234, an audible annunciator, a speaker, and / or via other components of the vehicle 1200. These outputs may include information such as vehicle speed, velocity, time, map data (e.g., Fig. 12C The HMI display 1234 may include information such as a high-definition (“HD”) map 1222 of the vehicle 1200 , location data (e.g., the location of the vehicle 1200 on the map), directions, locations of other vehicles (e.g., an occupancy grid), information about objects and states of objects as sensed by the controller 1236 , and the like. For example, the HMI display 1234 may display information about the presence of one or more objects (e.g., street signs, warning signs, traffic light changes, etc.) and / or information about driving maneuvers that the vehicle has made, is making, or will make (e.g., changing lanes now, leaving 34B in two miles, etc.).

[0094] The vehicle 1200 also includes a network interface 1224 that can communicate via one or more networks using one or more wireless antennas 1226 and / or a modem. For example, the network interface 1224 can be capable of communicating via Long Term Evolution (“LTE”), Wideband Code Division Multiple Access (“WCDMA”), Universal Mobile Telecommunications System (“UMTS”), Global System for Mobile Communications (“GSM”), IMT-CDMA Multi-Carrier (“CDMA2000”), etc. The one or more wireless antennas 1226 can also enable communication between objects in the environment (e.g., vehicles, mobile devices, etc.) using one or more local area networks such as Bluetooth, Bluetooth Low Energy (“LE”), Z-wave, ZigBee, etc. and / or one or more low power wide area networks (“LPWAN”) such as LoRaWAN, SigFox, etc.

[0095] Fig. 12B For use according to some embodiments of the present disclosure Fig. 12A 1200. The cameras and respective fields of view are an example embodiment and are not intended to be limiting. For example, additional and / or replaceable cameras may be included and / or the cameras may be located at different locations on the vehicle 1200.

[0096] The camera type used for the camera may include, but is not limited to, a digital camera that may be suitable for use with components and / or systems of the vehicle 1200. The camera may operate at an automotive safety integrity level (ASIL) B and / or at another ASIL. The camera type may have any image capture rate, such as 60 frames per second (fps), 120fps, 240fps, etc., depending on the embodiment. The camera may be capable of using a rolling shutter, a global shutter, another type of shutter, or a combination thereof. In some examples, the color filter array may include a red-white-white-white (RCCC) color filter array, a red-white-white-blue (RCCB) color filter array, a red-blue-green-white (RBGC) color filter array, a Foveon X3 color filter array, a Bayer sensor (RGGB) color filter array, a monochrome sensor color filter array, and / or another type of color filter array. In some embodiments, a clear pixel camera such as a camera with an RCCC, RCCB, and / or RBGC color filter array may be used in an effort to improve light sensitivity.

[0097] In some examples, one or more of the cameras can be used to perform advanced driver assistance system (ADAS) functions (e.g., as part of a redundant or fail-safe design). For example, a multi-function monocular camera can be installed to provide functions including lane departure warning, traffic sign assistance, and intelligent headlight control. One or more of the cameras (e.g., all of the cameras) can simultaneously record and provide image data (e.g., video).

[0098] One or more of the cameras may be mounted in a mounting assembly such as a custom designed (three-dimensional ("3D") printed) assembly to cut out stray light and reflections from within the car (e.g., reflections from the dashboard reflected in the windshield mirror) that may interfere with the camera's image data capture capabilities. With respect to the wing mirror mounting assembly, the wing mirror assembly may be custom 3D printed so that the camera mounting plate matches the shape of the wing mirror. In some examples, one or more cameras may be integrated into the wing mirror. For the side view cameras, one or more cameras may also be integrated into the four pillars at each corner of the cab.

[0099] A camera (e.g., a front-facing camera) having a field of view that includes a portion of the environment in front of the vehicle 1200 can be used for surround view to help identify the forward path and obstacles, as well as assist in providing information critical to generating an occupancy grid and / or determining a preferred vehicle path with the help of one or more controllers 1236 and / or control SoCs. The front-facing camera can be used to perform many of the same ADAS functions as LIDAR, including emergency braking, pedestrian detection, and collision avoidance. The front-facing camera can also be used for ADAS functions and systems, including lane departure warning ("LDW"), autonomous cruise control ("ACC"), and / or other functions such as traffic sign recognition.

[0100] A variety of cameras may be used in the front-facing configuration, including, for example, a monocular camera platform including a complementary metal oxide semiconductor ("CMOS") color imager. Another example may be a wide-angle camera 1270, which may be used to sense objects (e.g., pedestrians, intersection traffic, or bicycles) entering the field of view from the periphery. Fig. 12B Only one wide-angle camera is shown in the figure, but there may be any number (including zero) of wide-angle cameras 1270 on the vehicle 1200. In addition, long-range cameras 1298 (e.g., a long-view stereo camera pair) may be used for depth-based object detection, especially for objects for which a neural network has not been trained. Long-range cameras 1298 may also be used for object detection and classification and basic object tracking.

[0101] Any number of stereo cameras 1268 may also be included in the front configuration. In at least one embodiment, one or more stereo cameras 1268 may include an integrated control unit including a scalable processing unit that may provide a multi-core microprocessor and programmable logic ("FPGA") with an integrated controller area network ("CAN") or Ethernet interface on a single chip. Such a unit can be used to generate a 3D map of the vehicle environment, including distance estimates for all points in the image. Alternative stereo cameras 1268 may include a compact stereo vision sensor that may include two camera lenses (one on the left and one on the right) and an image processing chip that may measure the distance from the vehicle to the target object and use the generated information (e.g., metadata) to activate autonomous emergency braking and lane departure warning functions. Other types of stereo cameras 1268 may be used in addition to or alternatively to those described herein.

[0102] A camera (e.g., a side view camera) having a field of view that includes portions of the environment to the sides of the vehicle 1200 can be used for surround viewing, providing information used to create and update an occupancy grid and generate side impact collision warnings. Fig. 12B The four surround cameras 1274 shown in FIG. 1274 may be placed on the vehicle 1200. The surround cameras 1274 may include a wide-angle camera 1270, a fisheye camera, a 360° camera, and / or the like. For example, the four fisheye cameras may be placed in front, behind, and on the sides of the vehicle. In an alternative arrangement, the vehicle may use three surround cameras 1274 (e.g., left, right, and rear), and may utilize one or more other cameras (e.g., a forward-facing camera) as a fourth surround camera.

[0103] A camera having a field of view that includes a portion of the environment behind the vehicle 1200 (e.g., a rear view camera) can be used to assist with parking, surround view, rear collision warning, and creating and updating an occupancy grid. A variety of cameras can be used, including but not limited to cameras that are also suitable as front-facing cameras as described herein (e.g., long-range and / or mid-range cameras 1298, stereo cameras 1268, infrared cameras 1272, etc.).

[0104] Fig. 12C For use according to some embodiments of the present disclosure Fig. 12ABlock diagram of an example system architecture of an example autonomous vehicle 1200. It should be understood that this arrangement and other arrangements described herein are set forth merely as examples. Other arrangements and elements (e.g., machines, interfaces, functions, sequences, functional groupings, etc.) may be used in addition to or in place of those shown, and some elements may be omitted entirely. Further, many of the elements described herein are functional entities that may be implemented as discrete or distributed components or in combination with other components, and in any appropriate combination and location. The various functions described herein as being performed by an entity may be implemented by hardware, firmware, and / or software. For example, the various functions may be implemented by a processor executing instructions stored in a memory.

[0105] Fig. 12C Each of the components, features, and systems of the vehicle 1200 is illustrated as being connected via a bus 1202. The bus 1202 may include a controller area network (CAN) data interface (alternatively, referred to herein as a "CAN bus"). The CAN may be a network internal to the vehicle 1200 that assists in controlling various features and functions of the vehicle 1200, such as the actuation of brakes, acceleration, braking, steering, windshield wipers, and the like. The CAN bus may be configured to have tens or even hundreds of nodes, each with its own unique identifier (e.g., CAN ID). The CAN bus may be read to find steering wheel angle, ground speed, engine speed per minute (RPM), button position, and / or other vehicle status indicators. The CAN bus may be ASIL B compliant.

[0106] Although bus 1202 is described as a CAN bus here, this is not intended to be restrictive. For example, in addition to or alternatively to the CAN bus, FlexRay and / or Ethernet can be used. In addition, although bus 1202 is represented by a single line, this is not intended to be restrictive. For example, there can be any number of buses 1202, which may include one or more CAN buses, one or more FlexRay buses, one or more Ethernet buses, and / or one or more other types of buses using different protocols. In some examples, two or more buses 1202 can be used to perform different functions, and / or can be used for redundancy. For example, a first bus 1202 can be used for a collision avoidance function, and a second bus 1202 can be used for drive control. In any example, each bus 1202 can communicate with any component of the vehicle 1200, and two or more buses 1202 can communicate with the same component. In some examples, each SoC 1204, each controller 1236, and / or each computer within the vehicle may have access to the same input data (e.g., input from sensors of the vehicle 1200) and may be connected to a common bus such as a CAN bus.

[0107] The vehicle 1200 may include one or more controllers 1236, such as those described herein. Fig. 12A Those controllers described. Controller 1236 can be used for a variety of functions. Controller 1236 can be coupled to any other different components and systems of vehicle 1200, and can be used for control of vehicle 1200, artificial intelligence of vehicle 1200, infotainment for vehicle 1200, and / or the like.

[0108] The vehicle 1200 may include one or more system on chip (SoC) 1204. The SoC 1204 may include a CPU 1206, a GPU 1208, a processor 1210, a cache 1212, an accelerator 1214, a data store 1216, and / or other components and features not shown. The SoC 1204 may be used to control the vehicle 1200 in a variety of platforms and systems. For example, one or more SoCs 1204 may be combined with an HD map 1222 in a system (e.g., a system of the vehicle 1200), and the HD map may be downloaded from one or more servers (e.g., a server) via a network interface 1224. Fig.12D one or more servers 1278) to obtain map refreshes and / or updates.

[0109] CPU 1206 may include a CPU cluster or CPU complex (alternatively, referred to herein as "CCPLEX"). CPU 1206 may include multiple cores and / or L2 caches. For example, in some embodiments, CPU 1206 may include eight cores in a coherent multiprocessor configuration. In some embodiments, CPU 1206 may include four dual-core clusters, each of which has a dedicated L2 cache (e.g., 2MB L2 cache). CPU 1206 (e.g., CCPLEX) may be configured to support simultaneous cluster operations, so that any combination of clusters of CPU 1206 can be active at any given time.

[0110] CPU 1206 may implement power management capabilities including one or more of the following features: each hardware block may be automatically clock gated when idle to save dynamic power; each core clock may be gated when the core is not actively executing instructions due to the execution of WFI / WFE instructions; each core may be independently power gated; each core cluster may be independently clock gated when all cores are clock gated or power gated; and / or each core cluster may be independently power gated when all cores are power gated. CPU 1206 may further implement an enhanced algorithm for managing power states, in which allowed power states and expected wake-up times are specified, and hardware / microcode determines the optimal power state to enter for the core, cluster, and CCPLEX. The processing core may support a simplified power state entry sequence in software, with the work being offloaded to the microcode.

[0111] GPU 1208 may include an integrated GPU (alternatively, referred to herein as an "iGPU"). GPU 1208 may be programmable and efficient for parallel workloads. In some examples, GPU 1208 may use an enhanced tensor instruction set. GPU 1208 may include one or more streaming microprocessors, each of which may include an L1 cache (e.g., an L1 cache with at least 96KB storage capacity), and two or more of these streaming microprocessors may share an L2 cache (e.g., an L2 cache with 512KB storage capacity). In some embodiments, GPU 1208 may include at least eight streaming microprocessors. GPU 1208 may use a computing application programming interface (API). In addition, GPU 1208 may use one or more parallel computing platforms and / or programming models (e.g., NVIDIA's CUDA).

[0112] In the case of automotive and embedded use, GPU 1208 can be power optimized to achieve optimal performance. For example, GPU 1208 can be manufactured on fin field effect transistors (FinFETs). However, this is not intended to be limiting, and GPU 1208 can be manufactured using other semiconductor manufacturing processes. Each streaming microprocessor can merge several mixed precision processing cores divided into multiple blocks. For example and without limitation, 64 PF32 cores and 32 PF64 cores can be divided into four processing blocks. In such an example, each processing block can be allocated 16 FP32 cores, 8 FP64 cores, 16 INT32 cores, two mixed precision NVIDIA tensor cores for deep learning matrix arithmetic, L0 instruction cache, warp scheduler, dispatch unit and / or 64KB register file. In addition, the streaming microprocessor may include independent parallel integer and floating point data paths to provide efficient execution of workloads using a mix of computation and addressing computations. The streaming microprocessor may include independent thread scheduling capabilities to allow for finer-grained synchronization and cooperation between parallel threads. The streaming microprocessor may include a combined L1 data cache and shared memory unit to improve performance while simplifying programming.

[0113] GPU 1208 can include high bandwidth memory (HBM) and / or a 16GB HBM2 memory subsystem that provides a peak memory bandwidth of approximately 900 GB / s in some examples. In some examples, synchronous graphics random access memory (SGRAM), such as fifth generation graphics double data rate synchronous random access memory (GDDR5), can be used in addition to or in lieu of HBM memory.

[0114] The GPU 1208 may include unified memory technology that includes access counters to allow memory pages to be more accurately migrated to the processor that accesses them most frequently, thereby improving the efficiency of memory ranges shared between processors. In some examples, address translation service (ATS) support may be used to allow the GPU 1208 to directly access the CPU 1206 page table. In such an example, when the GPU 1208 memory management unit (MMU) experiences a miss, an address translation request may be transmitted to the CPU 1206. In response, the CPU 1206 may look up the virtual-physical mapping for the address in its page table and transmit the translation back to the GPU 1208. In this way, unified memory technology may allow a single unified virtual address space to be used for memory of both the CPU 1206 and the GPU 1208, thereby simplifying GPU 1208 programming and porting applications to the GPU 1208.

[0115] In addition, GPU 1208 may include access counters that can track how often GPU 1208 accesses the memory of other processors. Access counters can help ensure that memory pages are moved to the physical memory of the processor that accesses them most frequently.

[0116] SoC 1204 may include any number of caches 1212, including those described herein. For example, cache 1212 may include an L3 cache available to both CPU 1206 and GPU 1208 (e.g., connected to both CPU 1206 and GPU 1208). Cache 1212 may include a write-back cache that may track the state of a line, for example, by using a cache coherence protocol (e.g., MEI, MESI, MSI, etc.). Depending on the embodiment, the L3 cache may include 4MB or more, although smaller cache sizes may also be used.

[0117] The SoC 1204 may include an arithmetic logic unit (ALU) that may be utilized in performing any of a variety of tasks or operations with respect to the vehicle 1200, such as processing a DNN. Additionally, the SoC 1204 may include a floating point unit (FPU) (or other math coprocessor or digital coprocessor type) for performing mathematical operations within the system. For example, the SoC 104 may include one or more FPUs integrated as execution units within the CPU 1206 and / or GPU 1208.

[0118] SoC 1204 may include one or more accelerators 1214 (e.g., hardware accelerators, software accelerators, or a combination thereof). For example, SoC 1204 may include a hardware accelerator cluster, which may include optimized hardware accelerators and / or large on-chip memory. The large on-chip memory (e.g., 4MB SRAM) may enable the hardware accelerator cluster to accelerate neural networks and other calculations. The hardware accelerator cluster may be used to supplement GPU 1208 and offload some tasks of GPU 1208 (e.g., free up more cycles of GPU 1208 for performing other tasks). As an example, accelerator 1214 may be used for targeted workloads (e.g., perception, convolutional neural networks (CNNs), etc.) that are sufficiently stable to be easily controlled for acceleration. When used herein, the term "CNN" may include all types of CNNs, including region-based or regional convolutional neural networks (RCNNs) and fast RCNNs (e.g., for object detection).

[0119] Accelerator 1214 (e.g., a hardware accelerator cluster) may include a deep learning accelerator (DLA). DLA may include one or more tensor processing units (TPUs) that may be configured to provide an additional 10 trillion operations per second for deep learning applications and reasoning. TPU may be an accelerator configured to perform image processing functions (e.g., for CNN, RCNN, etc.) and optimized for performing image processing functions. DLA may be further optimized for a specific set of neural network types and floating point operations and reasoning. The design of DLA may provide higher performance per millimeter than a general-purpose GPU, and far exceeds the performance of a CPU. TPU may perform several functions, including a single instance convolution function, support, for example, INT8, INT16, and FP16 data types for both features and weights, and post-processor functions.

[0120] The DLA can quickly and efficiently execute neural networks, particularly CNNs, on processed or unprocessed data for any of a wide variety of functions, such as, but not limited to: a CNN for object recognition and detection using data from a camera sensor; a CNN for distance estimation using data from a camera sensor; a CNN for emergency vehicle detection and identification and detection using data from a microphone; a CNN for facial recognition and vehicle owner identification using data from a camera sensor; and / or a CNN for safety and / or security-related events.

[0121] The DLA can perform any function of the GPU 1208, and by using an inference accelerator, for example, the designer can target any function to either the DLA or the GPU 1208. For example, the designer can focus the processing of CNNs and floating point operations on the DLA, and leave other functions to the GPU 1208 and / or other accelerators 1214.

[0122] The accelerator 1214 (e.g., a hardware accelerator cluster) may include a programmable vision accelerator (PVA), which may be alternatively referred to as a computer vision accelerator herein. The PVA may be designed and configured to accelerate computer vision algorithms for advanced driver assistance systems (ADAS), autonomous driving, and / or augmented reality (AR) and / or virtual reality (VR) applications. The PVA may provide a balance between performance and flexibility. For example, each PVA may include, for example and without limitation, any number of reduced instruction set computer (RISC) cores, direct memory access (DMA), and / or any number of vector processors.

[0123] The RISC core can interact with an image sensor (e.g., an image sensor of any camera described herein), an image signal processor, and / or the like. Each of these RISC cores can include any amount of memory. Depending on the embodiment, the RISC core can use any of a number of protocols. In some examples, the RISC core can execute a real-time operating system (RTOS). The RISC core can be implemented using one or more integrated circuit devices, application specific integrated circuits (ASICs), and / or storage devices. For example, the RISC core can include an instruction cache and / or a tightly coupled RAM.

[0124] The DMA may enable components of the PVA to access system memory independently of the CPU 1206. The DMA may support any number of features used to provide optimizations for the PVA, including, but not limited to, support for multi-dimensional addressing and / or circular addressing. In some examples, the DMA may support addressing in up to six or more dimensions, which may include block width, block height, block depth, horizontal block stepping, vertical block stepping, and / or depth stepping.

[0125] The vector processor can be a programmable processor that can be designed to efficiently and flexibly perform programming for computer vision algorithms and provide signal processing capabilities. In some examples, the PVA can include a PVA core and two vector processing subsystem partitions. The PVA core can include a processor subsystem, one or more DMA engines (e.g., two DMA engines), and / or other peripherals. The vector processing subsystem can operate as the main processing engine of the PVA, and can include a vector processing unit (VPU), an instruction cache, and / or a vector memory (e.g., VMEM). The VPU core can include a digital signal processor, such as, for example, a single instruction multiple data (SIMD), a very long instruction word (VLIW) digital signal processor. The combination of SIMD and VLIW can enhance throughput and rate.

[0126] Each of the vector processors may include an instruction cache and may be coupled to a dedicated memory. As a result, in some examples, each of the vector processors may be configured to execute independently of the other vector processors. In other examples, the vector processors included in a particular PVA may be configured to employ data parallelism. For example, in some embodiments, multiple vector processors included in a single PVA may execute the same computer vision algorithm, but on different regions of an image. In other examples, the vector processors included in a particular PVA may execute different computer vision algorithms simultaneously on the same image, or even execute different algorithms on sequence images or portions of an image. Among other things, any number of PVAs may be included in a hardware accelerator cluster, and any number of vector processors may be included in each of these PVAs. In addition, the PVA may include additional error correction code (ECC) memory to enhance overall system security.

[0127] The accelerator 1214 (e.g., a hardware accelerator cluster) may include an on-chip computer vision network and SRAM to provide high bandwidth, low latency SRAM for the accelerator 1214. In some examples, the on-chip memory may include at least 4MB of SRAM consisting of, for example and not limited to, eight field-configurable memory blocks that can be accessed by both the PVA and the DLA. Each pair of memory blocks may include an advanced peripheral bus (APB) interface, configuration circuitry, a controller, and a multiplexer. Any type of memory may be used. The PVA and DLA may access the memory via a backbone that provides high-speed memory access to the PVA and DLA. The backbone may include an on-chip computer vision network that interconnects the PVA and DLA to the memory (e.g., using APB).

[0128] The on-chip computer vision network may include an interface that determines that both the PVA and DLA provide ready and valid signals before transmitting any control signals / addresses / data. Such an interface may provide separate phases and separate channels for transmitting control signals / addresses / data, as well as burst communication for continuous data transmission. This type of interface may comply with ISO 26262 or IEC 61508 standards, but other standards and protocols may also be used.

[0129] In some examples, SoC 1204 may include a real-time ray tracing hardware accelerator such as described in U.S. Patent Application No. 16 / 101,232 filed on August 10, 2018. The real-time ray tracing hardware accelerator may be used to quickly and efficiently determine the position and range of objects (e.g., within a world model) to generate real-time visualization simulations for RADAR signal interpretation, for sound propagation synthesis and / or analysis, for SONAR system simulations, for general wave propagation simulations, for comparison with LIDAR data for the purpose of positioning and / or other functions, and / or for other purposes. In some embodiments, one or more tree traversal units (TTUs) may be used to perform one or more ray tracing related operations.

[0130] The accelerator 1214 (e.g., a hardware accelerator cluster) has a wide range of uses in autonomous driving. The PVA can be a programmable visual accelerator that can be used in key processing stages in ADAS and autonomous vehicles. The capabilities of the PVA are a good match for algorithmic domains that require predictable processing, low power, and low latency. In other words, the PVA performs well on semi-dense or dense rule computations, and even on small data sets that require predictable runtimes with low latency and low power. Therefore, in the context of a platform for autonomous vehicles, the PVA is designed to run classic computer vision algorithms because they are efficient at object detection and integer math.

[0131] For example, according to one embodiment of the technology, PVA is used to perform computer stereo vision. In some examples, algorithms based on semi-global matching can be used, but this is not intended to be limiting. Many applications for level 3-5 autonomous driving require instant motion estimation / stereo matching (e.g., structure from self-motion, pedestrian recognition, lane detection, etc.). PVA can perform computer stereo vision functions on input from two monocular cameras.

[0132] In some examples, PVA can be used to perform dense optical flow. According to the process raw RADAR data (e.g., using 4D fast Fourier transform) to provide processed RADAR. In other examples, PVA is used for time-of-flight depth processing, which is for example by processing raw time-of-flight data to provide processed time-of-flight data.

[0133] DLA can be used to run any type of network to enhance control and driving safety, including, for example, a neural network that outputs a confidence measure for each object detection. Such confidence values ​​can be interpreted as probabilities, or as providing a relative "weight" of each detection compared to other detections. The confidence value enables the system to make further decisions about which detections should be considered true positive detections rather than false positive detections. For example, the system can set a threshold for confidence and only consider detections that exceed the threshold as true positive detections. In an automatic emergency braking (AEB) system, a false positive detection can cause the vehicle to automatically perform emergency braking, which is obviously undesirable. Therefore, only the most confident detections should be considered triggers for AEB. DLA can run a neural network for regressing confidence values. The neural network may take as its input at least some subset of parameters, such as bounding box dimensions, a ground plane estimate obtained (e.g., from another subsystem), an inertial measurement unit (IMU) sensor 1266 output related to the orientation and distance of the vehicle 1200, a 3D position estimate of an object obtained from the neural network and / or other sensors (e.g., a LIDAR sensor 1264 or a RADAR sensor 1260), etc.

[0134] SoC 1204 may include one or more data stores 1216 (e.g., memory). Data store 1216 may be on-chip memory of SoC 1204 that may store neural networks to be executed on the GPU and / or DLA. In some examples, data store 1216 may be large enough to store multiple instances of a neural network for redundancy and safety. Data store 1212 may include an L2 or L3 cache 1212. References to data store 1216 may include references to memory associated with a PVA, DLA, and / or other accelerator 1214 as described herein.

[0135] SoC 1204 may include one or more processors 1210 (e.g., embedded processors). Processor 1210 may include a boot and power management processor, which may be a dedicated processor and subsystem for handling boot power and management functions and related safety implementations. The boot and power management processor may be part of the SoC 1204 boot sequence and may provide runtime power management services. The boot power and management processor may provide clock and voltage programming, auxiliary system low power state transitions, SoC 1204 thermal and temperature sensor management, and / or SoC 1204 power state management. Each temperature sensor may be implemented as a ring oscillator whose output frequency is proportional to the temperature, and SoC 1204 may use the ring oscillator to detect the temperature of CPU 1206, GPU 1208, and / or accelerator 1214. If it is determined that the temperature exceeds a threshold, the boot and power management processor may enter a temperature fault routine and place SoC 1204 in a lower power state and / or place vehicle 1200 in a driver safety parking mode (e.g., parking vehicle 1200 safely).

[0136] Processor 1210 may also include a set of embedded processors that can be used as an audio processing engine. The audio processing engine can be an audio subsystem that allows full hardware support for multi-channel audio through multiple interfaces and a wide range of flexible audio I / O interfaces. In some examples, the audio processing engine is a dedicated processor core with a digital signal processor with dedicated RAM.

[0137] The processor 1210 may also include an always-on-processor engine that can provide the necessary hardware features to support low-power sensor management and wake-up use cases. The always-on-processor engine may include a processor core, tightly coupled RAM, supporting peripherals (such as timers and interrupt controllers), various I / O controller peripherals, and routing logic.

[0138] The processor 1210 may also include a safety cluster engine that includes a dedicated processor subsystem that handles safety management of automotive applications. The safety cluster engine may include two or more processor cores, tightly coupled RAM, supporting peripherals (e.g., timers, interrupt controllers, etc.), and / or routing logic. In safety mode, the two or more cores may operate in lockstep mode and act as a single core with comparison logic to detect any differences between their operations.

[0139] Processor 1210 may also include a real-time camera engine, which may include a dedicated processor subsystem for handling real-time camera management.

[0140] Processor 1210 may also include a high dynamic range signal processor, which may include an image signal processor, which is a hardware engine that is part of the camera processing pipeline.

[0141] The processor 1210 may include a video image compositer, which may be a processing block (e.g., implemented on a microprocessor) that implements video post-processing functions required for a video playback application to produce a final image for a player window. The video image compositer may perform lens distortion correction for the wide-angle camera 1270, the surround camera 1274, and / or for an in-cab surveillance camera sensor. The in-cab surveillance camera sensor is preferably monitored by a neural network running on another instance of the advanced SoC, configured to recognize in-cab events and respond accordingly. The in-cab system may perform lip reading to activate mobile phone service and place a call, dictate an email, change a vehicle destination, activate or change the vehicle's infotainment system and settings, or provide voice-activated web surfing. Certain functions are only available to the driver when the vehicle is operating in autonomous mode and are disabled in other circumstances.

[0142] The video image compositer may include enhanced temporal noise reduction for both spatial and temporal noise reduction. For example, in the case of motion in the video, the noise reduction appropriately weights the spatial information and reduces the weight of information provided by adjacent frames. In the case where an image or portion of an image does not include motion, the temporal noise reduction performed by the video image compositer may use information from previous images to reduce noise in the current image.

[0143] The video image compositor may also be configured to perform stereoscopic rectification on the input stereoscopic footage frames. The video image compositor may further be used for user interface composition when the operating system desktop is in use and the GPU 1208 does not need to continuously render new surfaces. Even when the GPU 1208 is powered on and active for 3D rendering, the video image compositor may be used to offload the GPU 1208 to improve performance and responsiveness.

[0144] SoC 1204 may also include a mobile industry processor interface (MIPI) camera serial interface for receiving video and input from a camera, a high-speed interface, and / or a video input block that may be used for camera and related pixel input functions. SoC 1204 may also include an input / output controller that may be controlled by software and may be used to receive I / O signals that are not committed to a specific role.

[0145] The SoC 1204 may also include a wide range of peripheral interfaces to enable communication with peripherals, audio codecs, power management, and / or other devices. The SoC 1204 may be used to process data from cameras (connected via Gigabit multimedia serial links and Ethernet), sensors (e.g., LIDAR sensor 1264, RADAR sensor 1260, etc., which may be connected via Ethernet), data from the bus 1202 (e.g., speed of the vehicle 1200, steering wheel position, etc.), data from the GNSS sensor 1258 (connected via Ethernet or CAN bus). The SoC 1204 may also include dedicated high-performance mass storage controllers, which may include their own DMA engines, and which may be used to free the CPU 1206 from routine data management tasks.

[0146] SoC 1204 can be an end-to-end platform with a flexible architecture that spans automation levels 3-5, thereby providing a comprehensive functional safety architecture that utilizes and efficiently uses computer vision and ADAS technologies to achieve diversity and redundancy, together with deep learning tools to provide a platform for a flexible and reliable driving software stack. SoC 1204 can be faster, more reliable, and even more energy efficient and space efficient than conventional systems. For example, when combined with CPU 1206, GPU 1208, and data storage 1216, accelerator 1214 can provide a fast and efficient platform for level 3-5 autonomous vehicles.

[0147] The technology thus provides capabilities and functionality that cannot be achieved by conventional systems. For example, computer vision algorithms can be executed on CPUs, which can be configured using high-level programming languages ​​such as the C programming language to execute a wide variety of processing algorithms across a wide variety of visual data. However, CPUs often fail to meet the performance requirements of many computer vision applications, such as those related to, for example, execution time and power consumption. In particular, many CPUs are unable to execute complex object detection algorithms in real time, which is a requirement for in-vehicle ADAS applications and a requirement for practical Level 3-5 autonomous vehicles.

[0148] In contrast to conventional systems, by providing a CPU complex, a GPU complex, and a cluster of hardware accelerators, the techniques described herein allow multiple neural networks to be executed simultaneously and / or sequentially, and the results to be combined together to achieve Level 3-5 autonomous driving functions. For example, a CNN executed on a DLA or dGPU (e.g., GPU 1220) may include text and word recognition, allowing the supercomputer to read and understand traffic signs, including signs for which a neural network has not been specifically trained. The DLA may also include a neural network that can recognize, interpret, and provide semantic understanding of the signs, and transfer that semantic understanding to a path planning module running on the CPU complex.

[0149] As another example, as required for Level 3, 4, or 5 driving, multiple neural networks may be running simultaneously. For example, a warning sign consisting of "Caution: Flashing Lights Indicate Icing Conditions" along with electric lights may be interpreted by several neural networks independently or collectively. The sign itself may be recognized as a traffic sign by a deployed first neural network (e.g., a trained neural network), and the text "Flashing Lights Indicate Icing Conditions" may be interpreted by a deployed second neural network that informs the vehicle's path planning software (preferably executing on a CPU complex) that icing conditions exist when the flashing lights are detected. The flashing lights may be recognized by operating a deployed third neural network over multiple frames that informs the vehicle's path planning software of the presence (or absence) of the flashing lights. All three neural networks may be running simultaneously, for example, within a DLA and / or on GPU 1208.

[0150] In some examples, a CNN for facial recognition and owner recognition can use data from the camera sensor to identify the presence of an authorized driver and / or owner of the vehicle 1200. The always-on sensor processing engine can be used to unlock the vehicle and turn on the lights when the owner approaches the driver's door, and in security mode, disable the vehicle when the owner leaves the vehicle. In this way, the SoC 1204 provides security against theft and / or carjacking.

[0151] In another example, a CNN for emergency vehicle detection and identification can use data from microphone 1296 to detect and identify emergency vehicle sirens. In contrast to conventional systems that use general classifiers to detect sirens and manually extract features, SoC 1204 uses CNN to classify environmental and urban sounds and to classify visual data. In a preferred embodiment, the CNN running on the DLA is trained to identify the relative closing rate of emergency vehicles (e.g., by using the Doppler effect). The CNN can also be trained to identify emergency vehicles specific to the local area in which the vehicle is operating, as identified by the GNSS sensor 1258. Thus, for example, when operating in Europe, the CNN will seek to detect European sirens, and when in the United States, the CNN will seek to identify only North American sirens. Once an emergency vehicle is detected, with the assistance of the ultrasonic sensor 1262, the control program can be used to execute emergency vehicle safety routines to slow the vehicle, drive to the side of the road, stop the vehicle, and / or idle the vehicle until the emergency vehicle passes.

[0152] The vehicle may include a CPU 1218 (e.g., a discrete CPU or dCPU) that may be coupled to the SoC 1204 via a high-speed interconnect (e.g., PCIe). The CPU 1218 may include, for example, an X86 processor. The CPU 1218 may be used to perform any of a variety of functions, including, for example, arbitrating potentially inconsistent results between ADAS sensors and the SoC 1204, and / or monitoring the status and health of the controller 1236 and / or the infotainment SoC 1230.

[0153] The vehicle 1200 may include a GPU 1220 (e.g., a discrete GPU or dGPU) that may be coupled to the SoC 1204 via a high-speed interconnect (e.g., NVIDIA's NVLINK). The GPU 1220 may provide additional artificial intelligence functionality, for example, by executing redundant and / or different neural networks, and may be used to train and / or update the neural network based at least in part on input from sensors of the vehicle 1200 (e.g., sensor data).

[0154] The vehicle 1200 may also include a network interface 1224, which may include one or more wireless antennas 1226 (e.g., one or more wireless antennas for different communication protocols, such as a cellular antenna, a Bluetooth antenna, etc.). The network interface 1224 can be used to enable wireless connections to the cloud (e.g., with a server 1278 and / or other network devices), to other vehicles, and / or to computing devices (e.g., a passenger's client device) via the Internet. In order to communicate with other vehicles, a direct link can be established between the two vehicles, and / or an indirect link can be established (e.g., across a network and through the Internet). The direct link can be provided using a vehicle-to-vehicle communication link. The vehicle-to-vehicle communication link can provide the vehicle 1200 with information about vehicles approaching the vehicle 1200 (e.g., vehicles in front of, to the side of, and / or behind the vehicle 1200). This functionality can be part of the cooperative adaptive cruise control functionality of the vehicle 1200.

[0155] The network interface 1224 may include a SoC that provides modulation and demodulation functions and enables the controller 1236 to communicate over a wireless network. The network interface 1224 may include a radio frequency front end for up-conversion from baseband to radio frequency and down-conversion from radio frequency to baseband. The frequency conversion may be performed by a known process and / or may be performed using a super-heterodyne process. In some examples, the radio frequency front end function may be provided by a separate chip. The network interface may include a wireless function for communicating via LTE, WCDMA, UMTS, GSM, CDMA2000, Bluetooth, Bluetooth LE, Wi-Fi, Z-wave, ZigBee, LoRaWAN, and / or other wireless protocols.

[0156] The vehicle 1200 may also include data storage 1228, which may include off-chip storage (e.g., outside the SoC 1204). The data storage 1228 may include one or more storage elements, including RAM, SRAM, DRAM, VRAM, flash memory, hard disk, and / or other components and / or devices that can store at least one bit of data.

[0157] The vehicle 1200 may also include a GNSS sensor 1258. The GNSS sensor 1258 (e.g., GPS, assisted GPS sensor, differential GPS (DGPS) sensor, etc.) is used to assist in mapping, perception, occupancy grid generation, and / or path planning functions. Any number of GNSS sensors 1258 may be used, including, for example and without limitation, GPS using a USB connector with an Ethernet to serial (RS-232) bridge.

[0158] The vehicle 1200 may also include a RADAR sensor 1260. The RADAR sensor 1260 may be used by the vehicle 1200 for remote vehicle detection even in darkness and / or inclement weather conditions. The RADAR functional safety level may be ASIL B. The RADAR sensor 1260 may use CAN and / or bus 1202 (e.g., to transmit data generated by the RADAR sensor 1260) for control and access to object tracking data, accessing Ethernet in some examples to access raw data. A variety of RADAR sensor types may be used. For example and without limitation, the RADAR sensor 1260 may be suitable for front, rear, and side RADAR use. In some examples, a pulse Doppler RADAR sensor is used.

[0159] The RADAR sensor 1260 may include different configurations, such as long-range with a narrow field of view, short-range with a wide field of view, short-range side coverage, and the like. In some examples, the long-range RADAR may be used for adaptive cruise control functions. The long-range RADAR system may provide a wide field of view (e.g., within a range of 250m) achieved by two or more independent scans. The RADAR sensor 1260 may help distinguish between static and moving objects and may be used by the ADAS system for emergency brake assistance and forward collision warnings. The long-range RADAR sensor may include a single-station multimode RADAR with multiple (e.g., six or more) fixed RADAR antennas and high-speed CAN and FlexRay interfaces. In an example with six antennas, the central four antennas may create a focused beam pattern designed to record the surroundings of the vehicle 1200 at a higher rate with minimal traffic interference from adjacent lanes. The other two antennas may expand the field of view, making it possible to quickly detect vehicles entering or leaving the lane of the vehicle 1200.

[0160] As an example, a medium-range RADAR system may include a range of up to 1260m (front) or 80m (rear) and a field of view of up to 42 degrees (front) or 1250 degrees (rear). A short-range RADAR system may include, but is not limited to, a RADAR sensor designed to be mounted at both ends of the rear bumper. When mounted at both ends of the rear bumper, such a RADAR sensor system may create two beams that continuously monitor the blind spots behind and beside the vehicle.

[0161] Short-range RADAR systems can be used in ADAS systems for blind spot detection and / or lane change assistance.

[0162] The vehicle 1200 may also include ultrasonic sensors 1262. The ultrasonic sensors 1262, which may be placed on the front, rear, and / or sides of the vehicle 1200, may be used for parking assistance and / or creating and updating an occupancy grid. A variety of ultrasonic sensors 1262 may be used, and different ultrasonic sensors 1262 may be used for different detection ranges (e.g., 2.5 m, 4 m). The ultrasonic sensors 1262 may operate at a functional safety level of ASIL B.

[0163] The vehicle 1200 may include a LIDAR sensor 1264. The LIDAR sensor 1264 may be used for object and pedestrian detection, emergency braking, collision avoidance, and / or other functions. The LIDAR sensor 1264 may be of functional safety level ASILB. In some examples, the vehicle 1200 may include multiple LIDAR sensors 1264 (e.g., two, four, six, etc.) that may be used in Ethernet (e.g., to provide data to a Gigabit Ethernet switch).

[0164] In some examples, the LIDAR sensor 1264 may be able to provide a list of objects and their distances for a 360° field of view. Commercially available LIDAR sensors 1264 may have, for example, an advertised range of approximately 1200m, an accuracy of 2cm-3cm, and support for 1200Mbps Ethernet connections. In some examples, one or more non-protruding LIDAR sensors 1264 may be used. In such examples, the LIDAR sensor 1264 may be implemented as a small device that can be embedded in the front, back, side, and / or corner of the vehicle 1200. In such an example, the LIDAR sensor 1264 may provide a field of view of up to 120 degrees horizontally and 35 degrees vertically, with a range of 200m, even for low reflectivity objects. The front-mounted LIDAR sensor 1264 may be configured for a horizontal field of view between 45 and 135 degrees.

[0165] In some examples, LIDAR technologies such as 3D flash LIDAR may also be used. 3D flash LIDAR uses a flash of laser as an emission source to illuminate the vehicle's surroundings up to about 200m. The flash LIDAR unit includes a receiver that records the laser pulse transmission time and the reflected light on each pixel, which in turn corresponds to the range from the vehicle to the object. Flash LIDAR can allow highly accurate and distortion-free images of the surrounding environment to be generated with each laser flash. In some examples, four flash LIDAR sensors can be deployed, one on each side of the vehicle 1200. Available 3D flash LIDAR systems include solid-state 3D staring array LIDAR cameras (e.g., non-scanning LIDAR devices) with no moving parts other than fans. The flash LIDAR device can use 5 nanosecond Class I (eye-safe) laser pulses per frame, and can capture the reflected laser light in the form of a 3D range point cloud and co-registered intensity data. By using flash LIDAR, and because flash LIDAR is a solid-state device with no moving parts, the LIDAR sensor 1264 may be less susceptible to motion blur, vibration, and / or shock.

[0166] The vehicle may also include an IMU sensor 1266. In some examples, the IMU sensor 1266 may be located at the center of the rear axle of the vehicle 1200. The IMU sensor 1266 may include, for example and without limitation, an accelerometer, a magnetometer, a gyroscope, a magnetic compass, and / or other sensor types. In some examples, such as in a six-axis application, the IMU sensor 1266 may include an accelerometer and a gyroscope, while in a nine-axis application, the IMU sensor 1266 may include an accelerometer, a gyroscope, and a magnetometer.

[0167] In some embodiments, the IMU sensor 1266 can be implemented as a miniature high-performance GPS-aided inertial navigation system (GPS / INS) that combines a micro-electromechanical system (MEMS) inertial sensor, a high-sensitivity GPS receiver, and an advanced Kalman filter algorithm to provide estimates of position, velocity, and attitude. Thus, in some examples, the IMU sensor 1266 can enable the vehicle 1200 to estimate heading by directly observing and correlating velocity changes from the GPS to the IMU sensor 1266 without the need for input from a magnetic sensor. In some examples, the IMU sensor 1266 and the GNSS sensor 1258 can be combined into a single integrated unit.

[0168] The vehicle may include microphones 1296 positioned in and / or around the vehicle 1200. The microphones 1296 may be used for, among other things, emergency vehicle detection and identification.

[0169] The vehicle may also include any number of camera types, including stereo cameras 1268, wide-angle cameras 1270, infrared cameras 1272, surround cameras 1274, long-range and / or mid-range cameras 1298, and / or other camera types. These cameras can be used to capture image data around the entire periphery of the vehicle 1200. The type of camera used depends on the embodiment and the requirements of the vehicle 1200, and any combination of camera types can be used to provide the necessary coverage around the vehicle 1200. In addition, the number of cameras can vary depending on the embodiment. For example, the vehicle may include six cameras, seven cameras, ten cameras, twelve cameras, and / or another number of cameras. As an example and not limitation, the cameras can support Gigabit Multimedia Serial Link (GMSL) and / or Gigabit Ethernet. Each of the cameras described herein may be described in detail with respect to Fig. 12A and Fig. 12B Described in more detail.

[0170] The vehicle 1200 may also include a vibration sensor 1242. The vibration sensor 1242 may measure vibrations of components of the vehicle, such as an axle. For example, changes in vibration may indicate changes in the road surface. In another example, when two or more vibration sensors 1242 are used, the difference between the vibrations may be used to determine friction or slip of the road surface (e.g., when there is a vibration difference between a powered drive shaft and a free-spinning shaft).

[0171] The vehicle 1200 may include an ADAS system 1238. In some examples, the ADAS system 1238 may include a SoC. The ADAS system 1238 may include autonomous / adaptive / automatic cruise control (ACC), cooperative adaptive cruise control (CACC), forward collision warning (FCW), automatic emergency braking (AEB), lane departure warning (LDW), lane keeping assist (LKA), blind spot warning (BSW), rear cross traffic warning (RCTW), collision warning system (CWS), lane centering (LC), and / or other features and functions.

[0172] The ACC system may use a RADAR sensor 1260, a LIDAR sensor 1264, and / or a camera. The ACC system may include a longitudinal ACC and / or a lateral ACC. The longitudinal ACC monitors and controls the distance to the vehicle immediately in front of the vehicle 1200, and automatically adjusts the vehicle speed to maintain a safe distance from the vehicle in front. The lateral ACC performs distance keeping and suggests that the vehicle 1200 change lanes when necessary. The lateral ACC is related to other ADAS applications such as LCA and CWS.

[0173] CACC uses information from other vehicles, which can be received from other vehicles indirectly via a wireless link or through a network connection (e.g., through the Internet) via a network interface 1224 and / or a wireless antenna 1226. A direct link can be provided by a vehicle-to-vehicle (V2V) communication link, while an indirect link can be an infrastructure-to-vehicle (I2V) communication link. Typically, the V2V communication concept provides information about the vehicle immediately ahead (e.g., the vehicle immediately ahead of the vehicle 1200 and in the same lane as it), while the I2V communication concept provides information about traffic farther ahead. The CACC system may include either or both of the I2V and V2V information sources. Given information about the vehicle ahead of the vehicle 1200, CACC can be more reliable, and it is possible to improve the smoothness of traffic flow and reduce road congestion.

[0174] The FCW system is designed to alert the driver to hazards so that the driver can take corrective action. The FCW system uses a front camera and / or RADAR sensor 1260 coupled to a dedicated processor, DSP, FPGA and / or ASIC, which is electrically coupled to driver feedback such as a display, speaker and / or vibrating components. The FCW system can provide warnings in the form of, for example, sound, visual warnings, vibrations and / or rapid brake pulses.

[0175] The AEB system detects an impending forward collision with another vehicle or other object and can automatically apply the brakes if the driver does not take corrective action within specified time or distance parameters. The AEB system can use a front camera and / or RADAR sensor 1260 coupled to a dedicated processor, DSP, FPGA and / or ASIC. When the AEB system detects a hazard, it typically first alerts the driver to take corrective action to avoid the collision, and if the driver does not take corrective action, the AEB system can automatically apply the brakes in an effort to prevent or at least mitigate the effects of the predicted collision. The AEB system may include technologies such as dynamic brake support and / or collision approach braking.

[0176] The LDW system provides visual, audible, and / or tactile warnings, such as steering wheel or seat vibrations, to alert the driver when the vehicle 1200 crosses a lane marking. When the driver indicates an intention to leave the lane, by activating a turn signal, the LDW system is not activated. The LDW system may use a front-facing camera coupled to a dedicated processor, DSP, FPGA, and / or ASIC that is electrically coupled to driver feedback such as a display, speaker, and / or vibrating component.

[0177] The LKA system is a variation of the LDW system. If the vehicle 1200 begins to leave the lane, the LKA system provides steering input or braking to correct the vehicle 1200.

[0178] The BSW system detects and warns the driver of vehicles in the car's blind spot. The BSW system can provide visual, auditory and / or tactile alerts to indicate that it is unsafe to merge or change lanes. The system can provide additional warnings when the driver uses a turn signal. The BSW system can use a rear-facing camera and / or RADAR sensor 1260 coupled to a dedicated processor, DSP, FPGA and / or ASIC, which is electrically coupled to driver feedback such as a display, speaker and / or vibration components.

[0179] The RCTW system may provide visual, audible, and / or tactile notifications when an object is detected outside the range of the rear camera while the vehicle 1200 is in reverse. Some RCTW systems include AEB to ensure that the vehicle brakes are applied to avoid a crash. The RCTW system may use one or more rear RADAR sensors 1260 coupled to a dedicated processor, DSP, FPGA, and / or ASIC that is electrically coupled to driver feedback such as a display, speaker, and / or vibrating component.

[0180] Conventional ADAS systems may be prone to false positive results, which may annoy and distract the driver, but are typically not catastrophic because the ADAS system alerts the driver and allows the driver to decide whether the safety condition actually exists and take action accordingly. However, in the autonomous vehicle 1200, in the case of conflicting results, the vehicle 1200 itself must decide whether to pay attention to the results from the main computer or the auxiliary computer (e.g., the first controller 1236 or the second controller 1236). For example, in some embodiments, the ADAS system 1238 can be a backup and / or auxiliary computer for providing perception information to the backup computer rationality module. The backup computer rationality monitor can run redundant and diverse software on hardware components to detect faults in perception and dynamic driving tasks. The output from the ADAS system 1238 can be provided to the supervisory MCU. If the outputs from the main computer and the auxiliary computer conflict, the supervisory MCU must determine how to coordinate the conflict to ensure safe operation.

[0181] In some examples, the master computer can be configured to provide a confidence score to the supervisory MCU, indicating the master computer's confidence in the selected result. If the confidence score exceeds a threshold, the supervisory MCU can follow the direction of the master computer, regardless of whether the auxiliary computer provides conflicting or inconsistent results. In the case where the confidence score does not meet the threshold and the master computer and the auxiliary computer indicate different results (e.g., conflicts), the supervisory MCU can arbitrate between these computers to determine the appropriate result.

[0182] The supervisory MCU may be configured to run a neural network that is trained and configured to determine conditions under which the auxiliary computer provides a false alarm based, at least in part, on outputs from the primary computer and the auxiliary computer. Thus, the neural network in the supervisory MCU may learn when the output of the auxiliary computer may be trusted and when it may not. For example, when the auxiliary computer is a RADAR-based FCW system, the neural network in the supervisory MCU may learn when the FCW system is identifying a metal object that is not actually dangerous, such as a drainage grate or manhole cover that triggers an alarm. Similarly, when the auxiliary computer is a camera-based LDW system, the neural network in the supervisory MCU may learn to ignore the LDW when a cyclist or pedestrian is present and lane departure is actually the safest strategy. In an embodiment that includes a neural network running on the supervisory MCU, the supervisory MCU may include at least one of a DLA or GPU suitable for running the neural network with associated memory. In a preferred embodiment, the supervisory MCU may include a component of SoC 1204 and / or be included as a component of SoC1204.

[0183] In other examples, the ADAS system 1238 can include an auxiliary computer that uses traditional computer vision rules to perform ADAS functions. In this way, the auxiliary computer can use classic computer vision rules (if-then), and the presence of a neural network in the supervisory MCU can improve reliability, safety, and performance. For example, diverse implementations and intentional non-identity make the entire system more fault-tolerant, especially for failures caused by software (or software-hardware interface) functions. For example, if there is a software vulnerability or error in the software running on the main computer and the non-identical software code running on the auxiliary computer provides the same overall result, the supervisory MCU can be more confident that the overall result is correct and that the vulnerability in the software or hardware on the main computer does not cause a substantial error.

[0184] In some examples, the output of the ADAS system 1238 can be fed to the perception block of the main computer and / or the dynamic driving task block of the main computer. For example, if the ADAS system 1238 indicates a forward collision warning due to an object immediately ahead, the perception block can use this information in identifying the object. In other examples, the auxiliary computer can have its own neural network that is trained and thus reduces the risk of false positives as described herein.

[0185] The vehicle 1200 may also include an infotainment SoC 1230 (e.g., an in-vehicle infotainment system (IVI)). Although illustrated and described as an SoC, the infotainment system may not be an SoC and may include two or more discrete components. The infotainment SoC 1230 may include a combination of hardware and software that can be used to provide audio (e.g., music, personal digital assistant, navigation instructions, news, radio, etc.), video (e.g., TV, movies, streaming, etc.), phone (e.g., hands-free calling), network connectivity (e.g., LTE, WiFi, etc.), and / or information services (e.g., navigation system, rear parking assistance, radio data system, vehicle-related information such as fuel level, total distance covered, brake fuel level, oil level, door open / closed, air filter information, etc.) to the vehicle 1200. For example, the infotainment SoC 1230 may include a radio, a disc player, a navigation system, a video player, USB and Bluetooth connectivity, an onboard computer, onboard entertainment, WiFi, steering wheel audio controls, hands-free voice controls, a head-up display (HUD), an HMI display 1234, a telematics device, a control panel (e.g., for controlling and / or interacting with various components, features, and / or systems), and / or other components. The infotainment SoC 1230 may further be used to provide information (e.g., visual and / or auditory) to a user of the vehicle, such as information from an ADAS system 1238, autonomous driving information such as planned vehicle maneuvers, trajectories, surrounding environment information (e.g., intersection information, vehicle information, road information, etc.), and / or other information.

[0186] The infotainment SoC 1230 may include GPU functionality. The infotainment SoC 1230 may communicate with other devices, systems, and / or components of the vehicle 1200 via a bus 1202 (e.g., a CAN bus, Ethernet, etc.). In some examples, the infotainment SoC 1230 may be coupled to a supervisory MCU so that in the event of a failure of a master controller 1236 (e.g., a primary and / or backup computer of the vehicle 1200), the GPU of the infotainment system may perform some self-driving functions. In such an example, the infotainment SoC 1230 may place the vehicle 1200 in a driver-safe parking mode as described herein.

[0187] The vehicle 1200 may also include an instrument cluster 1232 (e.g., a digital instrument panel, an electronic instrument cluster, a digital instrument panel, etc.). The instrument cluster 1232 may include a controller and / or a supercomputer (e.g., a separate controller or a supercomputer). The instrument cluster 1232 may include a set of instruments, such as a speedometer, a fuel level, an oil pressure, a tachometer, an odometer, a turn indicator, a shift position indicator, a seat belt warning light, a parking brake warning light, an engine fault light, an airbag (SRS) system information, lighting controls, safety system controls, navigation information, and the like. In some examples, information may be displayed and / or shared between the infotainment SoC 1230 and the instrument cluster 1232. In other words, the instrument cluster 1232 may be included as part of the infotainment SoC 1230, or vice versa.

[0188] Fig.12D For a cloud-based server and Fig. 12A 1276. System 1276 may include server 1278, network 1290, and vehicles including vehicle 1200. Server 1278 may include multiple GPUs 1284(A)-1284(H) (collectively referred to herein as GPU 1284), PCIe switches 1282(A)-1282(H) (collectively referred to herein as PCIe switches 1282), and / or CPUs 1280(A)-1280(B) (collectively referred to herein as CPU 1280). GPU 1284, CPU 1280, and PCIe switch may be interconnected with a high-speed interconnect and / or PCIe connection 1286 such as, for example and without limitation, NVLink interface 1288 developed by NVIDIA. In some examples, GPU 1284 is connected via NVLink and / or NVSwitch SoC, and GPU 1284 and PCIe switch 1282 are connected via PCIe interconnect. Although eight GPUs 1284, two CPUs 1280, and two PCIe switches are illustrated, this is not intended to be limiting. Depending on the embodiment, each of the servers 1278 may include any number of GPUs 1284, CPUs 1280, and / or PCIe switches. For example, each of the servers 1278 may include eight, sixteen, thirty-two, and / or more GPUs 1284.

[0189] Server 1278 may receive image data over network 1290 and from a vehicle, the image data representing images showing unexpected or changed road conditions, such as recently begun road work. Server 1278 may transmit neural network 1292, updated neural network 1292, and / or map information 1294, including information about traffic and road conditions, over network 1290 and to the vehicle. Updates to map information 1294 may include updates to HD map 1222, such as information about construction sites, potholes, curves, flooding, or other obstacles. In some examples, neural network 1292, updated neural network 1292, and / or map information 1294 may have been generated from new training and / or data received from any number of vehicles in the environment and / or based on experience from training performed at a data center (e.g., using server 1278 and / or other servers).

[0190] Server 1278 can be used to train a machine learning model (e.g., a neural network) based on training data. The training data can be generated by the vehicle, and / or can be generated in a simulation (e.g., using a game engine). In some examples, the training data is labeled (e.g., when the neural network benefits from supervised learning) and / or undergoes other preprocessing, while in other examples, the training data is not labeled and / or preprocessed (e.g., when the neural network does not require supervised learning). Training can be performed according to any one or more categories of machine learning techniques, including but not limited to the following categories: supervised training, semi-supervised training, unsupervised training, self-learning, reinforcement learning, joint learning, transfer learning, feature learning (including main components and cluster analysis), multilinear subspace learning, manifold learning, representation learning (including alternate dictionary learning), rule-based machine learning, anomaly detection, and any variants or combinations thereof. Once the machine learning model is trained, the machine learning model can be used by the vehicle (e.g., transmitted to the vehicle via network 1290), and / or the machine learning model can be used by server 1278 to remotely monitor the vehicle.

[0191] In some examples, server 1278 can receive data from the vehicle and apply the data to the latest real-time neural network for real-time intelligent reasoning. Server 1278 may include a deep learning supercomputer and / or a dedicated AI computer powered by GPU 1284, such as DGX and DGX Station machines developed by NVIDIA. However, in some examples, server 1278 may include a deep learning infrastructure of a data center powered only by CPUs.

[0192] The deep learning infrastructure of server 1278 may be capable of rapid real-time inference, and may use this capability to assess and verify the health of the processors, software, and / or associated hardware in vehicle 1200. For example, the deep learning infrastructure may receive periodic updates from vehicle 1200, such as a sequence of images and / or objects located in the sequence of images that vehicle 1200 has located (e.g., via computer vision and / or other machine learning object classification techniques). The deep learning infrastructure may run its own neural network to identify objects and compare them to the objects identified by vehicle 1200, and if the results do not match and the infrastructure concludes that the AI ​​in vehicle 1200 has failed, server 1278 may transmit a signal to vehicle 1200 instructing the fail-safe computer of vehicle 1200 to take control, notify passengers, and complete a safe parking maneuver.

[0193] For reasoning, server 1278 may include GPU 1284 and one or more programmable reasoning accelerators (e.g., NVIDIA's TensorRT). The combination of GPU-powered servers and reasoning acceleration can make real-time responses possible. In other examples, such as when performance is not so important, CPU, FPGA, and other processor-powered servers can be used for reasoning.

[0194] Example computing device

[0195] Fig.13 13 is a block diagram of an example computing device 1300 suitable for implementing some embodiments of the present disclosure. The computing device 1300 may include an interconnect system 1302 that directly or indirectly couples the following devices: a memory 1304, one or more central processing units (CPUs) 1306, one or more graphics processing units (GPUs) 1308, a communication interface 1310, an input / output (I / O) port 1312, an input / output component 1314, a power supply 1316, one or more presentation components 1318 (e.g., one or more displays), and one or more logic units 1320. In at least one embodiment, one or more computing devices 1300 may include one or more virtual machines (VMs), and / or any of their components may include virtual components (e.g., virtual hardware components). For non-limiting examples, one or more of the GPUs 1308 may include one or more vGPUs, one or more of the CPUs 1306 may include one or more vCPUs, and / or one or more of the logic units 1320 may include one or more virtual logic units. As such, one or more computing devices 1300 may include discrete components (eg, a full GPU dedicated to computing device 1300 ), virtual components (eg, a portion of a GPU dedicated to computing device 1300 ), or a combination thereof.

[0196] although Fig.13 The various blocks of are shown as being connected via interconnect system 1302 using wires, but this is not intended to be limiting and is for clarity only. For example, in some embodiments, presentation component 1318 (such as a display device) may be considered to be I / O component 1314 (e.g., if the display is a touch screen). As another example, CPU 1306 and / or GPU 1308 may include memory (e.g., memory 1304 may represent a storage device in addition to the memory of GPU 1308, CPU 1306, and / or other components). In other words, Fig.13 The computing devices described herein are illustrative only. No distinction is made between such categories as "workstations," "servers," "laptops," "desktop computers," "tablet computers," "client devices," "mobile devices," "handheld devices," "game consoles," "electronic control units (ECUs)," "virtual reality systems," and / or other device or system types, as all are considered Fig.13 within the range of computing devices.

[0197] The interconnection system 1302 may represent one or more links or buses, such as an address bus, a data bus, a control bus, or a combination thereof. The interconnection system 1302 may include one or more bus or link types, such as an industry standard architecture (ISA) bus, an extended industry standard architecture (EISA) bus, a video electronics standard association (VESA) bus, a peripheral component interconnect (PCI) bus, a peripheral component interconnect express (PCIe) bus, and / or another type of bus or link. In some embodiments, there is a direct connection between components. As an example, the CPU 1306 may be directly connected to the memory 1304. Further, the CPU 1306 may be directly connected to the GPU 1308. In the case where there is a direct or point-to-point connection between components, the interconnection system 1302 may include a PCIe link to perform the connection. In these examples, the PCI bus does not need to be included in the computing device 1300.

[0198] Memory 1304 may include any of a variety of computer-readable media. Computer-readable media may be any available media that can be accessed by computing device 1300. Computer-readable media may include volatile and non-volatile media, as well as removable and non-removable media. By way of example and not limitation, computer-readable media may include computer storage media and communication media.

[0199] Computer storage media may include volatile and nonvolatile media and / or removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules, and / or other data types. For example, memory 1304 may store computer readable instructions (e.g., representing one or more programs and / or one or more program elements, such as an operating system). Computer storage media may include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage devices or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by computing device 1300. As used herein, computer storage media does not include the signals themselves.

[0200] Computer storage media may embody computer readable instructions, data structures, program modules, and / or other data types in a modulated data signal such as a carrier wave or other transport mechanism, and include any information transfer media. The term "modulated data signal" may refer to a signal that has one or more of its characteristics set or changed in a manner that encodes the information in the signal. By way of example and not limitation, computer storage media may include wired media (such as a wired network or a direct wired connection) and wireless media (such as acoustic, RF, infrared, and other wireless media). Combinations of any of the above should also be included within the scope of computer readable media.

[0201] The CPU 1306 may be configured to execute at least some of the computer-readable instructions to control one or more components of the computing device 1300 to perform one or more of the methods and / or processes described herein. The CPUs 1306 may each include one or more cores (e.g., one, two, four, eight, twenty-eight, seventy-two, etc.) capable of handling numerous software threads simultaneously. The CPU 1306 may include any type of processor, and may include different types of processors (e.g., a processor with fewer cores for a mobile device and a processor with more cores for a server) depending on the type of computing device 1300 implemented. For example, depending on the type of computing device 1300, the processor may be an Advanced RISC Machine (ARM) processor implemented using Reduced Instruction Set Computing (RISC) or an x86 processor implemented using Complex Instruction Set Computing (CISC). The computing device 1300 may also include one or more CPUs 1306 in addition to one or more microprocessors or supplementary coprocessors (such as a math coprocessor).

[0202] In addition to or in place of one or more CPUs 1306, one or more GPUs 1308 may be configured to execute at least some of the computer-readable instructions to control one or more components of the computing device 1300 to perform one or more of the methods and / or processes described herein. One or more of the GPUs 1308 may be an integrated GPU (e.g., with one or more of the CPUs 1306) and / or one or more of the GPUs 1308 may be a discrete GPU. In an embodiment, one or more of the GPUs 1308 may be a coprocessor for one or more of the CPUs 1306. The GPUs 1308 may be used by the computing device 1300 to render graphics (e.g., 3D graphics) or perform general-purpose computing. For example, the GPUs 1308 may be used for general-purpose computing on GPUs (GPGPU). The GPUs 1308 may include hundreds or thousands of cores capable of handling hundreds or thousands of software threads simultaneously. The GPUs 1308 may generate pixel data for an output image in response to a rendering command (e.g., a rendering command received from the CPUs 1306 via a host interface). GPU 1308 may include graphics memory (e.g., display memory) for storing pixel data or any other suitable data (e.g., GPGPU data). Display memory may be included as part of memory 1304. GPU 1308 may include two or more GPUs operating in parallel (e.g., via a link). The link may connect the GPUs directly (e.g., using NVLINK) or may connect the GPUs through a switch (e.g., using NVSwitch). When combined together, each GPU 1308 may generate pixel data or GPGPU data for different portions or for different outputs (e.g., a first GPU for a first image and a second GPU for a second image). Each GPU may include its own memory, or may share memory with other GPUs.

[0203] In addition to or in lieu of the CPU 1306 and / or GPU 1308, the logic unit 1320 may be configured to execute at least some of the computer-readable instructions to control one or more components of the computing device 1300 to perform one or more of the methods and / or processes described herein. In embodiments, the one or more CPUs 1306, the one or more GPUs 1308, and / or the one or more logic units 1320 may discretely or jointly perform any combination of methods, processes, and / or portions thereof. One or more of the logic units 1320 may be a part of and / or integrated into one or more of the CPU 1306 and / or GPU 1308 and / or one or more of the logic units 1320 may be a discrete component or otherwise external to the CPU 1306 and / or GPU 1308. In embodiments, one or more of logic units 1320 may be a co-processor to one or more of CPUs 1306 and / or one or more of GPUs 1308 .

[0204] Examples of logic unit 1320 include one or more processing cores and / or components thereof, such as a data processing unit (DPU), a tensor core (TC), a tensor processing unit (TPU), a pixel vision core (PVC), a vision processing unit (VPU), a graphics processing cluster (GPC), a texture processing cluster (TPC), a streaming multiprocessor (SM), a tree transverse unit (TTU), an artificial intelligence accelerator (AIA), a deep learning accelerator (DLA), an arithmetic logic unit (ALU), an application specific integrated circuit (ASIC), a floating point unit (FPU), an input / output (I / O) element, a peripheral component interconnect (PCI) or a peripheral component interconnect express (PCIe) element, etc.

[0205] The communication interface 1310 may include one or more receivers, transmitters, and / or transceivers that enable the computing device 1300 to communicate with other computing devices via an electronic communication network (including wired and / or wireless communications). The communication interface 1310 may include components and functions that implement communication through any of a plurality of different networks, such as a wireless network (e.g., Wi-Fi, Z-Wave, Bluetooth, Bluetooth LE, ZigBee, etc.), a wired network (e.g., via Ethernet or wireless band communication), a low power wide area network (e.g., LoRaWAN, SigFox, etc.), and / or the Internet. In one or more embodiments, the logic unit 1320 and / or the communication interface 1310 may include one or more data processing units (DPUs) for transmitting data received through the network and / or through the interconnect system 1302 directly to (e.g., memory of) one or more GPUs 1308.

[0206] I / O ports 1312 may enable computing device 1300 to be logically coupled to other devices including I / O components 1314, one or more presentation components 1318, and / or other components, some of which may be built into (e.g., integrated into) computing device 1300. Illustrative I / O components 1314 include microphones, mice, keyboards, joysticks, game pads, game controllers, satellite dishes, scanners, printers, wireless devices, and the like. I / O components 1314 may provide a natural user interface (NUI) that processes air gestures, voice, or other physiological inputs generated by a user. In some cases, the input may be transmitted to an appropriate network element for further processing. NUI may implement any combination of voice recognition, stylus recognition, facial recognition, biometric recognition, gesture recognition on and near the screen, air gestures, head and eye tracking, and touch recognition (as described in more detail below) associated with the display of computing device 1300. Computing device 1300 may include a depth camera for gesture detection and recognition, such as a stereo camera system, an infrared camera system, an RGB camera system, touch screen technology, and combinations of these. Additionally, computing device 1300 may include an accelerometer or gyroscope that enables detection of motion (e.g., as part of an inertial measurement unit (IMU)). In some examples, computing device 1300 may use the output of the accelerometer or gyroscope to render an immersive augmented reality or virtual reality.

[0207] The power supply 1316 may include a hardwired power supply, a battery power supply, or a combination thereof. The power supply 1316 may provide power to the computing device 1300 to enable the components of the computing device 1300 to operate.

[0208] The presentation component 1318 may include a display (e.g., a monitor, a touch screen, a television screen, a head-up display (HUD), other display types, or a combination thereof), a speaker, and / or other presentation components. The presentation component 1318 may receive data from other components (e.g., GPU 1308, CPU 1306, DPU, etc.) and output the data (e.g., as an image, video, sound, etc.).

[0209] Sample Data Center

[0210] Fig.14 An example data center 1400 is shown that can be used in at least one embodiment of the present disclosure. The data center 1400 can include a data center infrastructure layer 1410, a frame layer 1420, a software layer 1430, and / or an application layer 1440.

[0211] like Fig.14 As shown, data center infrastructure layer 1410 may include resource coordinator 1412, grouped computing resources 1414, and node computing resources ("node CRs") 1416(1)-1416(N), where "N" represents any complete positive integer. In at least one embodiment, node CRs 1416(1)-1416(N) may include, but is not limited to, any number of central processing units (CPUs) or other processors (including DPUs, accelerators, field programmable gate arrays (FPGAs), graphics processors or graphics processing units (GPUs), etc.), memory devices (e.g., dynamic read-only memories), storage devices (e.g., solid-state or disk drives), network input / output (NW I / O) devices, network switches, virtual machines (VMs), power modules and / or cooling modules, etc. In some embodiments, one or more node CRs from node CRs 1416(1)-1416(N) may correspond to a server having one or more of the above-mentioned computing resources. Furthermore, in some embodiments, node CRs 1416 ( 1 )- 1416 (N) may include one or more virtual components, such as vGPUs, vCPUs, etc., and / or one or more of node CRs 1416 ( 1 )- 1416 (N) may correspond to a virtual machine (VM).

[0212] In at least one embodiment, the grouped computing resources 1414 may include separate groups of node CRs 1416 housed in one or more racks (not shown), or many racks housed in data centers at different geographical locations (also not shown). The separate groups of node CRs 1416 within the grouped computing resources 1414 may include grouped computing, network, memory, or storage resources that can be configured or allocated to support one or more workloads. In at least one embodiment, several node CRs 1416 including CPUs, GPUs, DPUs, and / or other processors may be grouped in one or more racks to provide computing resources to support one or more workloads. One or more racks may also include any number of power modules, cooling modules, and / or network switches in any combination.

[0213] Resource coordinator 1412 may configure or otherwise control one or more node CRs 1416(1)-1416(N) and / or grouped computing resources 1414. In at least one embodiment, resource coordinator 1412 may include a software design infrastructure (SDI) management entity for data center 1400. Resource coordinator 1412 may include hardware, software, or some combination thereof.

[0214] In at least one embodiment, Fig.14 As shown, the frame layer 1420 may include a job scheduler 1433, a configuration manager 1434, a resource manager 1436, and / or a distributed file system 1438. The frame layer 1420 may include frames of software 1432 supporting the software layer 1430 and / or one or more applications 1442 of the application layer 1440. The software 1432 or the application 1442 may include network-based service software or applications, such as those provided by Amazon Web Services, Google Cloud, and Microsoft Azure. The frame layer 1420 may be, but is not limited to, a free and open source software network application framework (such as Apache Spark) that can utilize the distributed file system 1438 for large-scale data processing (e.g., "big data"). TM(hereinafter referred to as "Spark")). In at least one embodiment, the job scheduler 1433 may include a Spark driver to facilitate scheduling workloads supported by different layers of the data center 1400. The configuration manager 1434 may be able to configure different layers, such as the software layer 1430 and the frame layer 1420 (which includes Spark and a distributed file system 1438 for supporting large-scale data processing). The resource manager 1436 may be able to manage clustered or grouped computing resources mapped to the distributed file system 1438 and the job scheduler 1433 or allocated to support the distributed file system 1438 and the job scheduler 1433. In at least one embodiment, the clustered or grouped computing resources may include grouped computing resources 1414 at the data center infrastructure layer 1410. The resource manager 1436 may coordinate with the resource coordinator 1412 to manage these mapped or allocated computing resources.

[0215] In at least one embodiment, the software 1432 included in the software layer 1430 may include software used by at least a portion of the node CRs 1416(1)-1416(N), the grouped computing resources 1414, and / or the distributed file system 1438 of the frame layer 1420. The one or more types of software may include, but are not limited to, Internet web page search software, email virus scanning software, database software, and streaming video content software.

[0216] In at least one embodiment, the applications 1442 included in the application layer 1440 may include one or more types of applications used by at least a portion of the node CRs 1416(1)-1416(N), the grouped computing resources 1414, and / or the distributed file system 1438 of the frame layer 1420. The one or more types of applications may include, but are not limited to, any number of genomic applications, cognitive computing, and machine learning applications, including training or inference software, machine learning framework software (e.g., PyTorch, TensorFlow, Caffe, etc.), and / or other machine learning applications used in conjunction with one or more embodiments.

[0217] In at least one embodiment, any of configuration manager 1434, resource manager 1436, and resource coordinator 1412 can implement any number and type of self-modification actions based on any amount and type of data acquired in any technically feasible manner. The self-modification actions can save a data center operator of data center 1400 from making potentially poor configuration decisions and potentially avoiding underutilized and / or poorly performing portions of the data center.

[0218] According to one or more embodiments described herein, data center 1400 may include tools, services, software, or other resources to train one or more machine learning models or use one or more machine learning models to predict or infer information. For example, one or more machine learning models may be trained by calculating weight parameters according to a neural network architecture using software and / or computing resources described above with respect to data center 1400. In at least one embodiment, a trained or deployed machine learning model corresponding to one or more neural networks may be used to infer or predict information using the resources described above with respect to data center 1400 using weight parameters calculated using one or more training techniques, such as, but not limited to, those described herein.

[0219] In at least one embodiment, the data center 1400 may use a CPU, an application-specific integrated circuit (ASIC), a GPU, an FPGA, and / or other hardware (or virtual computing resources corresponding thereto) to perform training and / or inference using the above resources. In addition, one or more software and / or hardware resources described above may be configured to allow a user to train or perform services that infer information, such as image recognition, speech recognition, or other artificial intelligence services.

[0220] Example network environment

[0221] A network environment suitable for implementing embodiments of the present disclosure may include one or more client devices, servers, network attached storage (NAS), other backend devices, and / or other device types. The client devices, servers, and / or other device types (e.g., each device) may be configured to communicate with the client device, server, or other device types. Fig.13 The backend device 1400 may be implemented on one or more instances of one or more computing devices 1300 of the present invention - for example, each device may include similar components, features and / or functions of one or more computing devices 1300. In addition, in the case of implementing a backend device (e.g., a server, NAS, etc.), the backend device may be included as part of the data center 1400, and examples of the data center 1400 are described herein with reference to the present invention. Fig.14 Describe in more detail.

[0222] The components of the network environment can communicate with each other via a network, which can be wired, wireless, or both. The network can include multiple networks or one of multiple networks. For example, the network can include one or more wide area networks (WANs), one or more local area networks (LANs), one or more public networks (such as the Internet and / or a public switched telephone network (PSTN)), and / or one or more private networks. In the case where the network includes a wireless telecommunications network, components such as base stations, communication towers, or even access points (and other components) can provide wireless connections.

[0223] Compatible network environments may include one or more peer-to-peer network environments (in which case the server may not be included in the network environment) and one or more client-server network environments (in which case one or more servers may be included in the network environment). In a peer-to-peer network environment, the functionality described herein for the server may be implemented on any number of client devices.

[0224] In at least one embodiment, the network environment may include one or more cloud-based network environments, distributed computing environments, combinations thereof, and the like. A cloud-based network environment may include a frame layer, a job scheduler, a resource manager, and a distributed file system implemented on one or more servers, which may include one or more core network servers and / or edge servers. The frame layer may include frames of software supporting the software layer and / or one or more applications of the application layer. The software or application may include network-based service software or applications, respectively. In an embodiment, one or more client devices may use network-based service software or applications (e.g., by accessing the service software and / or application via one or more application programming interfaces (APIs)). The frame layer may be, but is not limited to, a free and open source software network application frame that can use a distributed file system for large-scale data processing (e.g., "big data").

[0225] A cloud-based network environment can provide cloud computing and / or cloud storage that performs any combination of the computing and / or data storage functions (or one or more portions thereof) described herein. Any of these different functions can be distributed across multiple locations from a central or core server (e.g., one or more data centers that can be distributed across a state, region, country, global, etc.). If the connection to the user (e.g., client device) is relatively close to an edge server, the core server can assign at least a portion of the functionality to the edge server. A cloud-based network environment can be private (e.g., limited to a single organization), public (e.g., available to many organizations), and / or a combination thereof (e.g., a hybrid cloud environment).

[0226] One or more client devices may include the Fig.13At least some of the components, features, and functionality of one or more of the described example computing devices 1300. By way of example and not limitation, a client device may be implemented as a personal computer (PC), a laptop computer, a mobile device, a smart phone, a tablet computer, a smart watch, a wearable computer, a personal digital assistant (PDA), an MP3 player, a virtual reality headset, a global positioning system (GPS) or device, a video player, a camera, a surveillance device or system, a vehicle, a boat, a spacecraft, a virtual machine, a drone, a robot, a handheld communication device, a hospital device, a gaming device or system, an entertainment system, a vehicle computer system, an embedded system controller, a remote control, an appliance, a consumer electronic device, a workstation, an edge device, any combination of the depicted devices, or any other suitable device.

[0227] The present disclosure may be described in the general context of machine-usable instructions or computer codes executed by a computer or other machine such as a personal digital assistant or other handheld device, including computer-executable instructions such as program modules. Typically, program modules including routines, programs, objects, components, data structures, etc. refer to codes that perform specific tasks or implement specific abstract data types. The present disclosure may be practiced in a variety of system configurations, including handheld devices, consumer electronics, general-purpose computers, more specialized computing devices, etc. The present disclosure may also be practiced in a distributed computing environment in which tasks are performed by remote processing devices linked through a communication network.

[0228] As used herein, the statement of "and / or" with respect to two or more elements should be interpreted as referring to only one element or a combination of elements. For example, "element A, element B, and / or element C" may include only element A, only element B, only element C, element A and element B, element A and element C, element B and element C, or elements A, B, and C. In addition, "at least one of element A or element B" may include at least one of element A, at least one of element B, or at least one of element A and at least one of element B. Further, "at least one of element A and element B" may include at least one of element A, at least one of element B, or at least one of element A and at least one of element B.

[0229] The subject matter of the present disclosure is described in detail herein to meet statutory requirements. However, the description itself is not intended to limit the scope of the present disclosure. On the contrary, the present inventors have contemplated that the claimed subject matter may also be embodied in other ways to include steps different from the steps described herein in conjunction with other current or future technologies or combinations of similar steps. Moreover, although the terms "step" and / or "block" may be used herein to imply different elements of the method employed, these terms should not be interpreted as implying any particular order among or between the various steps disclosed herein, unless the order of the steps is explicitly described.

Claims

1. A method comprising: determining a reference frame and a target frame using frames of image data representing overlapping views of an environment surrounding the ego object; generating a ground projection of the reference frame; calculating reference color statistics using the ground projection of the reference frame; and migrating the reference color statistics from the ground projection of the reference frame to the target frame.

2. The method of claim 1, wherein determining the reference frame comprises: One of the frames of image data is selected based at least on one or more of: a direction of ego-motion of the ego-object, an active viewport viewing the environment, or a gaze of an operator of the ego-object.

3. The method of claim 1 , wherein generating the ground projection of the reference frame comprises: projecting the reference frame onto a three-dimensional (3D) bowl to generate a projected 3D bowl representation; and using a portion of the projected 3D bowl representation corresponding to a ground plane of the environment as the ground projection of the reference frame.

4. The method of claim 1 , wherein the reference frame and the target frame represent different views of the environment in a same time slice, and migrating the reference color statistics from the ground projection of the reference frame to the target frame is based at least on determining that an overlap area between the ground projection of the reference frame and a second ground projection of the target frame has less than or equal to a threshold number of pixels belonging to a detected object.

5. The method of claim 1 , wherein the reference frame represents at least a portion of the environment from a previous time slice and the target frame represents at least a portion of the environment from a subsequent time slice, and migrating the reference color statistics comprises: The reference color statistics are migrated from the ground projection of the reference frame based at least on a determination that an overlap area between a second ground projection of the candidate reference frame from the subsequent time slice and a third ground projection of the target frame includes more than a threshold number or percentage of pixels corresponding to a detected object.

6. The method of claim 1 , wherein computing the reference color statistics from the ground projection of the reference frame comprises: The reference color statistics are calculated from a majority cluster of the ground projection of the reference frame, and migrating the reference color statistics includes migrating the reference color statistics from the majority cluster of the ground projection of the reference frame to the target frame based at least on a determination that an overlap area between the ground projection of the reference frame and a second ground projection of the target frame includes less than a threshold number or percentage of pixels belonging to a detected object.

7. The method of claim 1, wherein migrating the reference color statistics from the ground projection of the reference frame to the target frame generates a modified target frame, and the method further comprises: At least the reference frame and the modified target frame are stitched into a stitched image and a visualization is rendered based on at least the stitched image.

8. A method as claimed in claim 1, wherein the method is performed by at least one of the following: a control system for an autonomous or semi-autonomous machine; a perception system for an autonomous or semi-autonomous machine; a system for performing simulation operations; a system for performing real-time streaming; a system for presenting one or more of augmented reality content, virtual reality content, or mixed reality content; a system for performing digital twin operations; a system for performing deep learning operations; a system implemented using an edge device; a system implemented using a robot; a system including one or more virtual machines (VMs); a system implemented at least in part in a data center; a system for performing light transport simulations; a system for performing collaborative content creation of 3D assets; a system for generating synthetic data; or a system implemented at least in part using cloud computing resources.

9. A processor, comprising: One or more processing units, the one or more processing units being configured to: determine a reference frame and a target frame; And migrating reference color statistics from the first ground projection of the reference frame to the target frame using the first ground projection of the reference frame and the second ground projection of the target frame.

10. The processor of claim 9, wherein the reference frame and the target frame are generated using a camera disposed on a self-object in an environment, and the one or more processing units are further configured to determine the reference frame based on at least one or more of: a direction of self-motion of the self-object, an active viewport viewing the environment, or a gaze of an operator of the self-object.

11. The processor of claim 9, wherein the one or more processing units are further configured to generate the first ground projection of the reference frame based at least on projecting the reference frame onto a three-dimensional (3D) bowl to generate a projected 3D bowl representation and using a portion of the projected 3D bowl representation corresponding to a ground plane as the first ground projection of the reference frame.

12. The processor of claim 9, wherein the reference frame and the target frame represent different views of an environment in a same time slice, and the one or more processing units are further configured to: migrate the reference color statistics from the first ground projection of the reference frame to the target frame based at least on a determination that an overlapping area between the first ground projection of the reference frame and the second ground projection of the target frame does not include any detected objects.

13. A processor as described in claim 9, wherein the reference frame represents a previous time slice and the target frame represents a subsequent time slice, and the one or more processing units are further used to: migrate the reference color statistics from the reference frame from the previous time slice based on determining that the overlapping area between the third ground projection of the candidate reference frame from the subsequent time slice and the second ground projection of the target frame includes more than a threshold number or percentage of pixels belonging to the detected object.

14. A processor as described in claim 9, wherein the one or more processing units are further used to: determine to migrate the reference color statistics from the majority cluster of the reference frame to the target frame based on determining that the overlapping area between the first ground projection of the reference frame and the second ground projection of the target frame includes less than a threshold number or percentage of pixels belonging to the detected object.

15. A processor as described in claim 9, wherein the processor is included in at least one of the following: a control system for an autonomous or semi-autonomous machine; a perception system for an autonomous or semi-autonomous machine; a system for performing simulation operations; a system for performing digital twin operations; a system for performing deep learning operations; a system for performing real-time streaming; a system for presenting one or more of augmented reality content, virtual reality content, or mixed reality content; a system implemented using an edge device; a system implemented using a robot; a system including one or more virtual machines (VMs); a system implemented at least in part in a data center; a system for performing light transport simulations; a system for performing collaborative content creation of 3D assets; a system for generating synthetic data; or a system implemented at least in part using cloud computing resources.

16. A system comprising: one or more processing units, the one or more processing units configured to: identify a reference image and a target image of an environment; generating a ground projection of the reference image; calculating reference color statistics from the ground projection of the reference image; and modifying one or more colors of the target image to correspond to the reference color statistics from the ground projection of the reference image.

17. A system as claimed in claim 16, wherein the reference image and the target image represent different views of the environment in the same time slice, and the one or more processing units are further used to: based on determining that the overlapping area between the ground projection of the reference image and the second ground projection of the target image has less than or equal to a threshold number of pixels belonging to the detected object, modify the color of the target image to correspond to the reference color statistics of the ground projection from the reference image.

18. A system as claimed in claim 16, wherein the reference image represents the environment from a previous time slice and the target image represents the environment from a subsequent time slice, and the one or more processing units are further used to: based on determining that the overlapping area between the second ground projection of the candidate reference image from the subsequent time slice and the third ground projection of the target image includes more than a threshold number or percentage of pixels belonging to the detected object, modify the color of the target image to correspond to the reference color statistics of the ground projection of the reference image from the previous time slice.

19. The system of claim 16, wherein computing the reference color statistics from the ground projection of the reference image comprises: The reference color statistics are calculated from a majority cluster of the ground projections of the reference image, and the one or more processing units are further used to: based on determining that an overlapping area between the ground projection of the reference image and a second ground projection of the target image includes less than a threshold number or percentage of pixels belonging to a detected object, modify the color of the target image to correspond to the reference color statistics from the majority cluster of the ground projections of the reference image.

20. A system as described in claim 16, wherein the system is included in at least one of the following: a control system for an autonomous or semi-autonomous machine; a perception system for an autonomous or semi-autonomous machine; a system for performing simulation operations; a system for performing digital twin operations; a system for performing deep learning operations; a system for performing real-time streaming; a system for presenting one or more of augmented reality content, virtual reality content, or mixed reality content; a system implemented using an edge device; a system implemented using a robot; a system including one or more virtual machines (VMs); a system implemented at least in part in a data center; a system for performing light transport simulations; a system for performing collaborative content creation of 3D assets; a system for generating synthetic data; or a system implemented at least in part using cloud computing resources.

Images