Object detection using planar homography and self-supervised scene structure understanding

Abstract

The present invention discloses object detection using planar homography and self-supervised scene structure understanding. In various examples, a single camera is used to capture two images of a scene from different locations. A trained neural network, which takes the two images as input, outputs a scene structure graph indicating the ratio of the height and depth values ​​of the pixel positions associated with the images. The ratio may indicate the presence of an object above a surface (e.g., a road surface) within the scene. Object detection may then be performed on non-zero values ​​or regions in the scene structure graph.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of U.S. Provisional Application No. 63 / 020,527, filed May 5, 2020, the entire contents of which are incorporated herein by reference. Background Art

[0003] Embodiments of the present disclosure solve various problems involving computer vision and are applicable to a wide variety of situations, such as on-road obstacle detection for autonomous vehicles. For example, it is critical for a human driver to identify potential road hazards in the ego lane and then take timely actions, such as stopping or changing lanes, to avoid accidents. Similarly, self-driving cars are expected to have the ability to detect any potential hazardous conditions on the road.

[0004] The embodiments described herein represent improvements over previous attempts to solve the problem of detecting road hazards or other objects on the road surface. Summary of the invention

[0005] Embodiments of the present disclosure relate to object detection using planar homography and self-supervised understanding of scene structure. Systems and methods are disclosed that can capture images using a single camera mounted on the front of a mobile autonomous machine, and the images are used to train a deep neural network (DNN) to predict a scene structure graph. During a training phase, a planar homography between two image frames obtained from a camera at two different points in time is calculated, and a first warped frame is generated based at least in part on the planar homography. The remaining difference between the first warped image and the second frame is modeled as a residual optical flow, which is calculated at least in part based on the scene structure obtained as a DNN output using the two image frames as input. Ultimately, the residual optical flow is used to generate a second warped image based at least in part on the first warped image, and the difference between the second image and the second warped image (e.g., photometric loss) can be used to update the parameters of the DNN.

[0006] In the implementation phase, two image frames, which may be captured at two different time points, are provided as input to the DNN, and a scene structure is determined. The scene structure includes a set of numerical values, where a specific numerical value indicates the height and depth ratio of a specific pixel included in the two image frames. In various embodiments, a non-zero value in the scene structure graph indicates the presence of an obstacle above the road surface, which may result in the activation of one or more obstacle detection algorithms.

[0007] Compared to conventional systems discussed herein, embodiments of the present disclosure can utilize minimal sensors to detect road hazards for simplicity and cost savings. For example, the disclosed systems and methods can operate effectively with a single image camera (e.g., a monocular camera) mounted on a vehicle facing forward (e.g., a dashboard camera or other front-facing camera described in more detail below).

[0008] For example, some previous techniques use two images from a sequence as input and iteratively process them based on feature matching and consensus-based algorithms to compute a ground homography. Thereafter, dangerous objects are detected from the difference between the outputs of the processed images. This simple approach may not be applicable to complex scenes where many feature points are extracted from off-ground objects, which adversely affects the plane homography estimation. Furthermore, extracting image features and performing iterative matching for each frame (partly due to the timing requirements of the iterations) may not be feasible for real-time deployment.

[0009] Other techniques utilize low-level image features (e.g., image gradients) and depth images from stereo cameras to segment obstacles by optimizing a Markov Random Field (MRF) model. This approach is based on the assumption that the region around obstacles exhibits high depth curvature, high depth variance, and high image gradients. The implementation of this technique relies on stereo camera input and external depth estimation methods. In addition, this technique requires a deep neural network trained for road segmentation. For many applications, the complexity and computational requirements of this pipeline may be prohibitive and / or impractical. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The present system and method for object detection using planar homography and self-supervised scene structure understanding is described in detail below with reference to the accompanying drawings, wherein:

[0011] Figure 1 is an illustration of an exemplary single camera system for detecting objects on a road surface according to some embodiments of the present disclosure;

[0012] Figure 2 is an example process for generating a warped image for use in a training process according to some embodiments of the present disclosure;

[0013] Figure 3 According to some embodiments of the present disclosure, the training Figure 1 Example process of using a model for an example single camera system;

[0014] Figure 4 is a method for generating a distorted image for use in a training process and a training image according to some embodiments of the present disclosure. Figure 1 Example process of using a model for an example single camera system;

[0015] Figure 5 According to some embodiments of the present disclosure, Figure 1 An example single-camera system uses a trained model to detect an object.

[0016] Figure 6 is an example scene structure diagram generated by a trained model according to some embodiments of the present disclosure;

[0017] Fig. 7A is an example of an input image for a trained DNN according to some embodiments of the present disclosure;

[0018] Figure 7B is an example scene structure diagram generated by a trained model according to some embodiments of the present disclosure;

[0019] Fig. 8A is an example of an input image for a trained DNN according to some embodiments of the present disclosure;

[0020] Figure 8B is an example scene structure diagram generated by a trained model according to some embodiments of the present disclosure;

[0021] Fig. 9A is an example of an input image for a trained DNN according to some embodiments of the present disclosure;

[0022] Fig. 9B is an example scene structure diagram generated by a trained model according to some embodiments of the present disclosure;

[0023] Fig.10 is an illustration of an example autonomous vehicle according to some embodiments of the present disclosure;

[0024] Fig.11 According to some embodiments of the present disclosure Fig.10 Examples of autonomous vehicle camera positions and fields of view;

[0025] Fig.12 According to some embodiments of the present disclosure Fig.10 a block diagram of an example system architecture for an example autonomous vehicle;

[0026] Fig.13 According to some embodiments of the present disclosure, a method for Fig.10 A system diagram of an exemplary autonomous vehicle communicating with each other; and

[0027] Fig.14 is a block diagram of an example computing device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0028] Systems and methods related to object detection using planar homographies and self-supervised scene structure understanding are disclosed.

[0029] Object detection, including detection of potential obstacles, is an important feature of autonomous machines with assisted driving functions or fully autonomous driving capabilities. Embodiments of the present disclosure provide camera-based methods for object detection using computer vision and deep learning techniques. In addition, due to its simplicity and cost-effectiveness, this solution can be implemented with minimal sensor setup and computational requirements. Specifically, the proposed method and system can detect objects above the road surface using only a single image camera (e.g., a monocular camera) mounted on a moving vehicle (such as a dash cam or similar front-facing camera).

[0030] In various approaches, autonomous machines may utilize inputs from multimodal sensors, such as cameras, radars, and / or lidars, when performing various operations, such as detecting obstacles on a road service. However, the use of additional inputs comes at an additional cost, both in terms of additional hardware and additional computational requirements to process these additional inputs. In various embodiments described in the present disclosure, detection of obstacles on the road is performed using a minimal setup of sensor systems and computational intensity. For example, the systems and methods described herein may be used in deployments with only a single forward-facing camera mounted on a moving vehicle or robot. Although the systems and methods are not so limited, the inclusion of advanced sensors, such as lidars, may not be cost-effective and greatly increases system complexity, making large-scale deployment impossible in the near future.

[0031] Furthermore, supervised learning methods will be difficult to scale and generalize due to the lack of labeled data for road obstacles and the impossibility of enumerating all possible kinds of obstacles on the road. Furthermore, manually labeling large datasets to train neural networks will be very expensive.

[0032] Embodiments of the systems and methods described herein use a single camera mounted on a moving vehicle or robot to capture images of the vehicle's surroundings. In an example training phase, two image frames from the camera are used as inputs to train a deep neural network (DNN). In one example, one or more computer systems within the vehicle or robot then use the DNN to predict a scene structure graph, the size of which is the same as the size of the input image. In one embodiment, the structure graph is a numerical matrix that is determined at least in part based on the ratio of height to depth predicted for a particular pixel position. Using a road segmentation or lane detection algorithm (which may be based on current technology), points on the road with non-zero values ​​in the structure graph may be obstacles. In various embodiments, a bounding box surrounding the obstacle can then be determined. In various embodiments, conventional techniques such as thresholding and connected component analysis are applied to generate bounding boxes. In the above examples, the DNN can perform unsupervised or self-supervised learning. However, in various other embodiments, supervised learning can be used to improve accuracy (e.g., to predict the height and depth of objects on the road surface). For example, additional sensor data (e.g., lidar data) can be used to perform supervised learning of the DNN.

[0033] The proposed system and method can be included as a necessary component in existing self-driving vehicle software to ensure safe operation at all levels of the autonomous driving system. In addition, the application of these techniques in a single camera system saves costs (e.g., sensor costs and computational costs) and reduces the requirements for implementing a self-driving vehicle. Another particular advantage of the proposed method is that due to the nature of self-supervised learning, a lot of effort can be saved from data labeling, which will greatly speed up the development process and reduce costs. In addition, other application areas such as robotics can also take advantage of this method, for example, to enable a robot to stay away from obstacles or navigate around obstacles.

[0034] Reference Figure 1 , Figure 11 is an example of a vehicle 102 with a single front camera 108 according to some embodiments of the present disclosure, which uses a trained DNN to detect an object 106 on a road surface 110. In addition, as described in more detail below, the camera 108 is used to capture an image of a scene, which can be used to train the DNN and / or provided as an input to the DNN to detect an object on the road surface. It should be understood that this arrangement and other arrangements described herein are only set forth as examples. In addition to or in place of those shown, other arrangements and elements (e.g., machines, interfaces, functions, commands, groupings of functions, etc.) can be used, and some elements can be omitted entirely. In addition, many of the elements described herein are functional entities that can be implemented as discrete or distributed components or combined with other components and implemented in any suitable combination and position. The various functions described herein as being performed by an entity can be performed by hardware, firmware, and / or software. For example, various functions can be performed by a processor executing instructions stored in a memory.

[0035] Figure 1 Two scenes T0 and T1 are shown, where T0 shows a first time point at which a first image 104A of the scene is captured, and T1 shows a second time point at which a second image 104B is captured. According to one or more embodiments, an implementation may use two (or more) consecutive frames, or two (or more) frames separated by a time gap (e.g., 0.2 seconds). For example, in a video captured by camera 108, first image 104A is adjacent to second image 104B. In another example, first image 104A and second image 104B are captured by camera 108, but separated by a time interval, e.g., 0.2 seconds. In various embodiments, the time interval between first image 104A and second image 104B may be modified based at least in part on data obtained from an inertial measurement unit (IMU) sensor of vehicle 102, which will be described in more detail below. For example, the interval between capturing first image 104A and second image 104B may be increased or decreased based at least in part on the speed of the vehicle. In yet another example, if the steering angle of the vehicle exceeds a threshold (eg, 30%), the capture of images may be temporarily paused, increased, or decreased.

[0036] In various embodiments, the first image 104A and the second image 104B (or additional images as volumetric image data, as described in more detail below) contain information of nearly the same scene but from slightly different views, so the three-dimensional structure of the scene, such as depth and height at each pixel location, can be inferred from the frame differences. Described below is a two-stage process, including a training stage and a deployment stage, for detecting objects on a road surface using the first image 104A and the second image 104B using a DNN.

[0037] In the exemplary training phase (combined with Figure 2 and Figure 3 During a training phase (described in more detail below), the first image 104A and the second image 104B may be used as input to train a DNN or similar algorithm to predict objects on the road. In other embodiments, multiple images and / or data (e.g., sensor data) may be provided as input to the DNN. During the training phase, a corpus of training images may be used to train the DNN. Additionally, the camera 108 may capture multiple images and / or videos, which may be used as training data in addition to or in lieu of the corpus of training images. Taking the first image 104A and / or the second image 104B as an example, during the training phase, a road segmentation and / or lane detection algorithm may be applied to the image to identify an approximate area of ​​the road in the first image 104A and / or the second image 104B.

[0038] Additionally, in various embodiments, these algorithms and other suitable algorithms are used to perform feature extraction on images (e.g., first image 104A and second image 104B) used to train the DNN. Extracting features from road areas in the training images enables feature matching-based homography, where, for example, the first image 104A is warped to the second image 104B to generate a warped image. As shown below in conjunction with Figure 3 As described in more detail, the warped image generated as a result of the homography is further warped using a residual optical flow map (e.g., vector) computed at least in part based on a scene structure graph output by the DNN.

[0039] In various embodiments, a homography transformation (e.g., a 3×3 matrix) is estimated using a random sample consensus (RANSAC) or similar algorithm so that correspondences can be established between key points detected on the road surface using feature extractors such as Speeded Up Robust Features (SURF), Robust Independent Elementary Features (BRIEF), Oriented FAST and Rotated BRIEF (ORB), and Scale Invariant Feature Transform (SIFT). This is called a planar homography. In other embodiments where additional information about the camera 108 or multiple cameras within the vehicle 110 is known (e.g., camera pose or other camera parameters), estimation of the homography transformation is not required because the correspondence between image coordinates and real-world coordinates can be determined by the pose or other camera parameters of the camera.

[0040] return Figure 1In the example of FIG. 1 , a planar homography is used to warp the first image 104A to the second image 104B. In the warped image generated as a result of the planar homography, any object (e.g., object 106) that is not on the plane (e.g., the road surface identified using the above algorithm) will not be aligned with the corresponding point in the second image 104B. Based at least in part on the warped image, the remaining difference between the warped image and the second image 104B can be modeled by a residual optical flow, which is determined at least in part based on the height of a specific point (e.g., a pixel representing object 106 in the image) located above the plane. In addition, while a plane is used as an illustrative example, the technology can be applied to other surfaces. For example, a curved surface (e.g., a hill, a river bank, a pothole, a curb, a ramp, etc.) that is convex or concave, or otherwise curved in the direction of vehicle travel (as well as in other identical directions). In various embodiments, during the training phase, the residual optical flow is calculated based at least in part on the scene structure output by the DNN. Returning to the above example, a scene structure graph is generated by a DNN with the first image 104A and the second image 104B as input. As described in more detail below, the scene structure graph is a matrix representing the ratio of height to depth at pixel locations within an image.

[0041] In various embodiments, during training of the DNN, the input to the DNN is the first image 104A and the second image 104B, and the output of the DNN is a scene structure map that is converted into a residual optical flow map (e.g., using known techniques). The residual optical flow can then be used to further warp the warped image (e.g., an image generated by warping the first image 104A into the second image 104B as described above) toward the second image 104B, thereby obtaining a second warped image. In various embodiments, the purpose of the residual optical flow is to align points above a plane (e.g., road surface 110). As described in Figure 1 Just as the midplane homography aligns points on the surface of road 110 , the residual optical flow may be used to align points on object 106 .

[0042] Typically, the closer the second warped image (e.g., the result of the warped image being warped toward the second image 104B) is to the second image 104B, the more accurate the residual optical flow, and therefore, the more accurate the scene structure. As a result, the photometric loss between the second warped image (e.g., a multi-warped image produced by at least warping the first warped image) and the second image 104B (e.g., the difference between the images) can be used to update the parameters of the DNN during training. During the deployment phase, the DNN generates a scene structure graph, for example, based at least in part on the first image 104A and the second image 104B, planar homographies and residual optical flows may not be calculated, and the parameters of the DNN may not be adjusted.

[0043] In addition, a DNN or other suitable network can be trained to take any number of images and / or sensor data as input. Non-limiting examples include two (or more) images taken one second apart from a single camera, non-contiguous images from a video feed of a single camera, or even four images captured by two cameras at two points in time. The systems and methods described in the present disclosure are flexible, and in various embodiments, object detection can be performed using only two images from a single camera. As described in the present disclosure, the DNN then predicts a scene structure graph from the two images. The DNN can be trained to use any image-based input format. In addition, the images can include grayscale images or color images in all kinds of color spaces such as RGB and YUV.

[0044] In addition to being able to use multiple images to generate a scene structure graph, multiple images (e.g., 30 frames captured in one second) can be used as input to the DNN during the example training phase. For example, as described above, given a series of images (e.g., image 1, image 2, ..., image 30), multiple combinations of images (e.g., image 1 and image 2, image 1 and image 30, image 2 and image 30, ...) can be used to calculate homography and photometric loss. In another example, the entire sequence of images (e.g., 30 images) can be stacked as a volume (e.g., volume data) including image 1, image 2, ..., image 30 and used as input to the DNN. In such an example, the additional images can provide more fine-grained information and scene details to help the DNN improve performance.

[0045] In various embodiments, during the deployment phase, the first image 104A and the second image 104B are used as inputs to the DNN, which detects the object 106 above the road surface 110 using information in the scene structure graph (e.g., the ratio of height to depth). Figure 4 As described in more detail, when performing hazard detection, non-zero locations within the road region of the scene structure graph may be identified. These non-zero locations indicate that the height above the road 110 is non-zero, and therefore indicate that the object 106 (e.g., debris) is on the road 110. The system executing the DNN or other system may generate a bounding box surrounding these non-zero regions using standard image processing techniques (e.g., thresholding, dilation / erosion, connected component analysis, or any other suitable image processing algorithm).

[0046] Using a trained DNN and using the techniques described in this disclosure, an accurate scene structure map can be generated from only two images obtained from a camera 108 on a vehicle 102 in motion. In various embodiments, any area corresponding to a non-zero value in the scene structure map is considered a potential obstacle. In addition, the DNN can operate in a self-supervised learning or unsupervised learning manner without any manual labeling of the specific shape, type, or location of obstacles.

[0047] Finally, despite Figure 1 A vehicle 102 with a single camera 108 is shown in , but it should be noted that the proposed method can be easily extended to use multiple image frames to better capture the scene structure. In addition, images from multiple cameras (e.g., stereo cameras) can provide additional visual cues of the same scene, resulting in better accuracy. In addition, although the described systems and methods use DNNs, other network models can be used. For example, the above-mentioned DNN takes two or more images as input, and the output is a scene structure graph (which can be generated as an image). As a result, any fully convolutional network or convolutional network that produces an image-shaped output can be utilized in conjunction with any of the embodiments described in the present disclosure. Non-limiting examples of suitable networks include: U-Net, DeeplabV3+, DeeplabV3, SegNet, and FCN. Note that, in general, any network suitable for semantic segmentation can be used.

[0048] The systems and methods described in the present disclosure may also be used in other applications. As examples, non-limiting applications of the present disclosure are directed to depth estimation, height estimation, off-ground object detection, off-ground object segmentation, robotics-related applications, navigation, or other image processing and computer vision applications. In general, the described neural networks predict a scene structure that contains more basic geometric information about the real-world scene. Therefore, the neural networks are suitable for applications that use estimation and / or prediction of real-world geometric information of the scene.

[0049] Embodiments of the present disclosure provide a novel method for using self-supervised learning for obstacle detection using a single camera; compared to traditional computer vision and image processing-based methods, the use of DNNs can model complex patterns in real data; according to the self-learning mechanism of the embodiments of the present disclosure, expensive human data labeling is also avoided compared to supervised learning. Instead, a large amount of easily accessible unlabeled data can be used to improve the accuracy of the calculated predictions. In addition, the systems and methods described herein do not require any prior knowledge about depth or disparity. In various embodiments, only color images are required as input during the deployment phase.

[0050] Reference now Figure 2 , Figure 3 , Figure 4 and Figure 5 , each block of methods 200, 300, 400, and 500 described herein includes a computing process that can be performed using any combination of hardware, firmware, and / or software. For example, various functions can be performed by a processor executing instructions stored in a memory. Each method can also be embodied as computer usable instructions stored on a computer storage medium. These methods 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, to name a few. In addition, by way of example, with respect to Figure 1 The method 200 is described with reference to an object detection system of FIG. However, these methods may additionally or alternatively be performed by any one system or any combination of systems, including but not limited to those described herein.

[0051] Figure 2 is a flow chart illustrating a method 200 for generating a warped image for use during a training phase on a DNN according to some embodiments of the present disclosure. At box 202, method 200 includes acquiring a first image and a second image. As described above, the images can be acquired from a single camera (e.g., a dash cam) mounted on a vehicle. In one embodiment, the images are adjacent images in a video stream. In other embodiments, there is a time shift (e.g., 0.2 seconds) between the images. The time shift can be modified based at least in part on information about the movement of the camera or the vehicle to which the camera is attached (e.g., speed, global positioning system (GPS) coordinates, steering angle, etc.). In addition, as described above, method 200 can be modified to include volume data, such as additional images and / or sensor data.

[0052] At box 204, the system performing method 200 identifies a road region within the first image and the second image. As described above, road segmentation, free space estimation, and / or lane detection algorithms can be used to identify the road region. The road region is identified so as to detect objects above the road surface. At box 206, the system performing method 200 detects key points and extracts key point features from the first image and the second image. For example, ORB or a similar algorithm can be used to detect key points and extract features from the images. As described below, the features extracted from the images and the corresponding key points allow for the estimation of a plane homography and the determination of a plane associated with the road.

[0053] At block 208, the system performing method 200 estimates a homography transformation to warp the first image into the second image. In various embodiments, RANSAC is used to estimate the homography transformation based at least in part on information obtained from the images. For example, the planar homography is based at least in part on feature matching and consensus from pre-segmented road surfaces in the first image and the second image. The system performing method 200 does not require precise isolation of the road area from the scene. In addition, in combination with Figure 2 The operations described can be performed during the training phase, do not need to be performed during the deployment phase, and do not increase the computational burden during the deployment phase.

[0054] As a result of performing direct homography estimation via keypoint matching and / or feature extraction, no information about the camera's calibration (e.g., camera pose) is required, and coordinate transformations such as from the image plane to world coordinates are bypassed. This greatly reduces system complexity. In embodiments where camera calibration is readily available and reliable, method 200 can be modified to compute homography using camera poses, rather than performing the processes described above at blocks 206 and 208.

[0055] At block 210, the system performing method 200 performs a planar homography using the estimated homography transformation to warp the first image toward the second image and generate a first warped image. At block 212, the processing of the image and the generation of the warped image are completed, and Figure 3 Continue training the DNN at block 302 of FIG. 1 . In various embodiments, Figure 2 The operations described in the above may be omitted, may be performed in a different order, may be performed in parallel, or may be performed in a combination of serial and parallel. Figure 2 ), optionally choosing to perform transformations such as mean subtraction, standard deviation normalization, and preprocessing the input to obtain better results (such as image denoising).

[0056] Figure 3 is a flow chart illustrating a method 300 for training a DNN to detect objects on a road surface according to some embodiments of the present disclosure. At block 302, the method 300 includes providing a first image and a second image as inputs to the DNN. In various embodiments, the method 300 may be part of a larger training phase to prepare the DNN for deployment. Figure 2 The described computer vision and image processing operations generate data (eg, a warped image) that can be used in conjunction with method 300. As described above, the DNN outputs a scene structure graph based at least in part on the first image and the second image.

[0057] At block 304, the system executing method 300 converts the scene structure graph output by the DNN into a residual optical flow graph. In various embodiments, the residual optical flow can be modeled as a vector, which can be used to further convert the scene structure graph output by the DNN into a residual optical flow graph. Figure 2The warped image generated by the planar homography described in is warped toward the second image. The determination of the residual optical flow can be performed using known computer vision and pattern recognition techniques. At box 306, based at least in part on the residual optical flow, the first warped image (e.g., the warped image generated at box 210 of method 200) is further warped toward the second image to generate a second warped image. In various embodiments, the planar homography described above is aligned between the first image and the second image to the road surface, and the warped image at box 306 of method 300 is further aligned to an object above the road surface.

[0058] At block 308, the system performing method 300 determines a photometric difference between the second warped image and the second image. The photometric difference and / or photometric loss measures the difference between the input image (e.g., the second image) and the warped image (e.g., the second warped image generated at block 306) based on the optical flow predicted by the DNN. At block 310, the system performing method 300 may update parameters of the DNN based at least in part on the photometric difference determined at block 308.

[0059] Furthermore, in various embodiments, each of the methods 200 and 300 may be generalized to different settings, e.g., multiple input frames, multiple camera inputs, multiple additional sensors, etc. Furthermore, the design of the DNN may be aligned with the latest techniques to further address accuracy or latency requirements.

[0060] Figure 4 4 is a flowchart illustrating a method 400 according to some embodiments of the present disclosure, the method 400 being used to generate a distorted image used during a training phase on a DNN and training a DNN to detect an object on a road surface. At boxes 402A and 402B, the method 400 includes acquiring a first image frame and a second image frame. As described above, images can be acquired from a single camera (e.g., a dash cam) mounted on a vehicle. In one embodiment, the images are adjacent images in a video stream. In other embodiments, there may be a time shift (e.g., 0.2 seconds) between images. The time shift can be modified based at least in part on information about the movement of the camera or the vehicle to which the camera is attached (e.g., speed, global positioning system (GPS) coordinates, steering angle, etc.). In addition, as described above, the method 400 can be modified to include volume data, such as additional images and / or sensor data.

[0061] At block 404, method 400 includes providing first image frame 402A and second image frame 402B as inputs to the DNN. Figure 2The computer vision and image processing operations described herein generate data, such as a warped image 410 as described in more detail below. As described above, the DNN outputs a scene structure map 406 based at least in part on the first image frame 402A and the second image frame 402B. In various embodiments, the scene structure map 406 includes images and data described in the present disclosure, such as the following in combination Figure 5 Those described.

[0062] At block 408, the system performing method 400 converts the scene structure map 406 into a residual optical flow map given the calibrated camera height and the estimated homography. In various embodiments, the residual optical flow 408 can be modeled as a vector that can be used to convert the scene structure map 406 into a residual optical flow map. Figure 2 The warped image 1410 generated by the planar homography described in is further warped towards the second image. In various embodiments, the determination of the residual optical flow is performed using known computer vision and pattern recognition techniques.

[0063] At block 412, the system performing process 400 further warps the warped image 1410 toward the second image frame 2402B using the residual optical flow 408. In various embodiments, warping the warped image 1410 is performed using the residual optical flow 408. The residual optical flow 408 provides information associated with objects above the road surface that is used to warp the objects above the road surface in the warped image 1410 toward the objects above the road surface in the second image frame 402B. The image warping that occurs in block 412 generates another warped image 1414. As described above, the further warped image 1414 is an image generated as a result of warping the warped image 1410 toward the second image frame 402B.

[0064] At block 416, the system performing method 400 determines a photometric difference (e.g., loss calculation) between the further warped image 1414 and the second image frame 402B. The photometric difference and / or photometric loss measures the difference between the input image (e.g., the second image frame 402B) and the warped image (e.g., the further warped image 1414) based on the optical flow predicted by the DNN.

[0065] Figure 5 5 is a flow chart illustrating a method 500 for a deployment phase of a trained DNN according to some embodiments of the present disclosure. At block 502, the method 500 includes providing a first image and a second image to the trained DNN as input. At block 504, the system performing the method 500 obtains a scene structure graph generated as an output of the DNN. As described in the present disclosure, the DNN takes an image as input and outputs a scene structure graph indicating predicted heights above a road surface of objects within the scene.

[0066] At box 506, the system performing method 500 performs analysis of the scene structure graph. In various embodiments, the scene structure graph includes a height and depth ratio at each pixel in the scene predicted by the DNN. In such an embodiment, any non-zero value of a pixel can represent an object or a portion thereof above the road surface. A pixel with a value of zero can indicate that there is nothing on the road surface. Therefore, in one embodiment, the analysis of the scene structure graph includes masking pixels that do not represent objects on the road surface. Various techniques for focusing on non-zero values ​​included in the scene structure graph can be used in conjunction with the embodiments described in the present disclosure.

[0067] At box 508, the system performing method 500 can detect obstacles on the road surface. For example, by performing spot detection and / or connected component analysis on at least the non-zero pixels indicated in the scene structure map. According to one or more additional embodiments, the implementation of two-image input can be extended to multi-image input so that the time information embedded in the volume video data can be used to improve the results. Instead of using a single camera, multiple cameras such as stereo cameras can be utilized. The parallax of overlapping scenes between different cameras can supplement the information of scenes captured at different times carried by a single camera. Technical details such as the photometric loss function used in training neural networks (e.g., DNNs) can be extended to different volume data inputs (e.g., stereo cameras). For example, to calculate the similarity between the distorted image and the reference image, we can use L1 loss, L2 loss, structural similarity (SSIM) loss, or other suitable techniques. When multimodal data is available, the neural network can be modified to obtain more inputs. For example, when radar / lidar data is available, this information can be used in conjunction with camera images to predict the scene structure map.

[0068] Figure 6An example output or scene structure map 602 generated based at least in part on two images provided as input to the DNN is shown. In various embodiments, the output of the DNN is a single scene structure map 602. In addition, the scene structure map 602 can have the same size as the input image, or a reduced size, depending on accuracy and efficiency. For example, at a current timestamp using the DNN to predict the scene structure, the DNN takes the current frame as input and one or more previous frames as input (e.g., a 640×480 RGB color image captured from a single camera), and performs reasoning. As a result, a scene structure map 602 of size 640×480 is output, so that at each pixel location, the value of the scene structure map 602 indicates the ratio of height to depth. As described above, the scene structure map can also have a reduced output size, such as 320×240; in this case, the resolution will be coarser, meaning that now in the map, one pixel will tell the height / depth information corresponding to four pixels in the original image. In various embodiments, the reduced size of the scene structure map can improve efficiency.

[0069] Fig. 7A An example input image 700A depicts a box in a road scene. In various embodiments, as described above, the input image 700A is captured by a front-facing camera of a moving vehicle. Figure 7B An example output scene structure graph depicting the box depicted in image 700A is depicted. In various embodiments, the example output scene structure graph is generated by a trained DNN as described above, to which is provided as input image 700A and at least one other image.

[0070] Fig. 8A An example input image 800A depicts a vehicle in a road scene. In various embodiments, as described above, the input image 800A is captured by a front-facing camera of a moving vehicle. Figure 8B An example output scene structure graph is depicted for the vehicle depicted in image 800A. In various embodiments, the example output scene structure graph is generated by a trained DNN as described above, to which image 800A and at least one other image are provided as input.

[0071] Fig. 9A An example input image 900A depicts two-dimensional road markings in a road scene. In various embodiments, as described above, the input image 900A is captured by a front-facing camera of a moving vehicle. Fig. 9B An example output scene structure graph is depicted for the two-dimensional road markings depicted in image 900A. In various embodiments, the example output scene structure graph is generated by a trained DNN as described above, to which image 900A and at least one other image are provided as input.

[0072] Example Autonomous Vehicle

[0073] Fig.10 An example of an autonomous vehicle 1000 according to some embodiments of the present disclosure is shown. Autonomous vehicle 1000 (alternatively referred to herein as "vehicle 1000") 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 car, an ambulance, a boat, a construction vehicle, an underwater vehicle, an unmanned aircraft, 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 according to the 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 1000 may be capable of implementing functions consistent with one or more of levels 3-5 of autonomous driving levels. For example, depending on the embodiment, the vehicle 1000 may be capable of conditional automation (level 3), high automation (level 4), and / or full automation (level 5).

[0074] The vehicle 1000 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 1000 may include a propulsion system 1050, such as an internal combustion engine, a hybrid power plant, an all-electric engine, and / or another type of propulsion system. The propulsion system 1050 may be connected to a drive train of the vehicle 1000, which may include a transmission, to achieve propulsion of the vehicle 1000. The propulsion system 1050 may be controlled in response to receiving a signal from a throttle / accelerator 1052.

[0075] A steering system 1054, which may include a steering wheel, may be used to steer vehicle 1000 (e.g., along a desired path or route) when propulsion system 1050 is operating (e.g., when the vehicle is in motion). Steering system 1054 may receive signals from steering actuator 1056. For fully automated (Level 5) functionality, a steering wheel may be optional.

[0076] Brake sensor system 1046 may be used to operate vehicle brakes in response to receiving signals from brake actuator 1048 and / or brake sensors.

[0077] May include one or more system on chip (SoC) 1004 ( Fig.12 ) and / or one or more controllers 1036 of one or more GPUs can provide signals (e.g., representing commands) to one or more components and / or systems of the vehicle 1000. For example, the one or more controllers can send signals to operate vehicle brakes via one or more brake actuators 1048, to operate the steering system 1054 via one or more steering actuators 1056, and / or to operate the propulsion system 1050 via one or more throttles / accelerators 1052. The one or more controllers 1036 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 1000. The one or more controllers 1036 may include a first controller 1036 for autonomous driving functions, a second controller 1036 for functional safety functions, a third controller 1036 for artificial intelligence functions (e.g., computer vision), a fourth controller 1036 for infotainment functions, a fifth controller 1036 for redundancy in emergency situations, and / or other controllers. In some examples, a single controller 1036 may handle two or more of the above functions, two or more controllers 1036 may handle a single function, and / or any combination thereof.

[0078] The one or more controllers 1036 may provide signals for controlling one or more components and / or systems of the vehicle 1000 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 sensor 1058 (e.g., a global positioning system sensor), a RADAR sensor 1060, an ultrasonic sensor 1062, a LIDAR sensor 1064, an inertial measurement unit (IMU) sensor 1066 (e.g., an accelerometer, a gyroscope, a magnetic compass, a magnetometer, etc.), a microphone 1096, a stereo camera 1068, a wide angle camera 1070 (e.g., a fisheye camera), an infrared camera 1072, a surround camera 1074 (e.g., a 360 degree camera), a long-range and / or mid-range camera 1098, a speed sensor 1044 (e.g., for measuring the velocity of the vehicle 1000), a vibration sensor 1042, a steering sensor 1040, a brake sensor (e.g., as part of a brake sensor system 1046), and / or other sensor types.

[0079] One or more of the controllers 1036 may receive input (e.g., represented by input data) from the instrument panel 1032 of the vehicle 1000 and provide output (e.g., represented by output data, display data, etc.) via a human machine interface (HMI) display 1034, an audible annunciator, a speaker, and / or via other components of the vehicle 1000. These outputs may include information such as vehicle speed, velocity, time, map data (e.g., Fig.12 The HMI display 1034 may include information such as the HD map 1022 of the vehicle 1000, location data (e.g., the location of the vehicle 1000 on the map), directions, locations of other vehicles (e.g., occupancy grid), information about objects and states of objects as sensed by the controller 1036, etc. For example, the HMI display 1034 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.).

[0080] The vehicle 1000 further includes a network interface 1024 that can communicate via one or more networks using one or more wireless antennas 1026 and / or a modem. For example, the network interface 1024 may be capable of communicating via LTE, WCDMA, UMTS, GSM, CDMA2000, etc. The one or more wireless antennas 1026 may 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 LE, Z-wave, ZigBee, etc. and / or one or more low power wide area networks (LPWANs) such as LoRaWAN, SigFox, etc.

[0081] Fig.11 For use according to some embodiments of the present disclosure Fig.10 1000. 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 1100.

[0082] The camera type 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 1100. 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.

[0083] 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).

[0084] One or more of the cameras may be mounted in a mounting assembly such as a custom designed (3-D printed) assembly to cut off 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 3-D 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 side view cameras, one or more cameras may also be integrated into the four pillars at each corner of the cab.

[0085] 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 1100 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 1136 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.

[0086] A variety of cameras may be used in the front-facing configuration, including, for example, a monocular camera platform including a CMOS (complementary metal oxide semiconductor) color imager. Another example may be a wide-angle camera 1170, which may be used to sense objects (e.g., pedestrians, intersection traffic, or bicycles) entering the field of view from the periphery. Fig.11 Only one wide-angle camera is shown, but there may be any number of wide-angle cameras 1170 on the vehicle 1100. In addition, long-range cameras 1198 (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 1198 may also be used for object detection and classification and basic object tracking.

[0087] One or more stereo cameras 1168 may also be included in the front configuration. The stereo camera 1168 may include an integrated control unit including an expandable processing unit that may provide a multi-core microprocessor and programmable logic (e.g., FPGA) with an integrated CAN or Ethernet interface on a single chip. Such a unit may be used to generate a 3-D map of the vehicle environment, including distance estimates for all points in the image. An alternative stereo camera 1168 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 1168 may be used in addition to or alternatively to those described herein.

[0088] A camera having a field of view that includes portions of the environment to the sides of the vehicle 1100 (e.g., a side view camera) can be used for surround viewing, providing information used to create and update an occupancy grid and generate side impact collision warnings. Fig.11 The four surround cameras 1174 shown in FIG. 1174 may be placed around the vehicle 1100. The surround cameras 1174 may include a wide-angle camera 1170, a fisheye camera, a 360-degree camera, and / or the like. For example, 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 1174 (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.

[0089] A camera having a field of view that includes a portion of the environment behind the vehicle 1100 (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 1198, stereo cameras 1168, infrared cameras 1172, etc.).

[0090] Fig.12 For use according to some embodiments of the present disclosure Fig.10 Block diagram of an example system architecture of an example autonomous vehicle 1000. 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.

[0091] Fig.12 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.

[0092] 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.

[0093] The vehicle 1200 may include one or more controllers 1236, such as those described herein. Fig.10 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.

[0094] 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.13 one or more servers 1378) to obtain map refreshes and / or updates.

[0095] 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.

[0096] 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.

[0097] 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 computer-based 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).

[0098] 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.

[0099] 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.

[0100] 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.

[0101] 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.

[0102] 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.

[0103] The SoC 1204 may include an arithmetic logic unit (ALU) that may be utilized when performing processing for any of the various tasks or operations of 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.

[0104] 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 acceleration 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 acceleration cluster to accelerate neural networks and other calculations. The hardware acceleration 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 easily control 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).

[0105] Accelerator 1214 (e.g., a hardware acceleration 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.

[0106] 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.

[0107] 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.

[0108] The accelerator 1214 (e.g., a hardware acceleration cluster) may include a programmable vision accelerator (PVA), which may be alternatively referred to herein as a computer vision accelerator. 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.

[0109] 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.

[0110] 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.

[0111] 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.

[0112] 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 acceleration 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.

[0113] The accelerator 1214 (e.g., a hardware acceleration 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, which may be accessed by both the PVA and the DLA. Each pair of memory blocks may include an advanced peripheral bus (APB) interface, a configuration circuit system, 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).

[0114] 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.

[0115] 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 traversal units (TTUs) may be used to perform one or more ray tracing-related operations.

[0116] 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.

[0117] 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 motion, pedestrian recognition, lane detection, etc.). PVA can perform computer stereo vision functions on input from two monocular cameras.

[0118] In some examples, PVA can be used to perform dense optical flow. For example, PVA can be used to process raw RADAR data (e.g., using a 4D fast Fourier transform) to provide a processed RADAR signal before transmitting the next RADAR pulse. In other examples, PVA is used for time-of-flight depth processing, for example, by processing raw time-of-flight data to provide processed time-of-flight data.

[0119] 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 as 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.

[0120] 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 1216 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.

[0121] 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).

[0122] Processor 1210 may further include a set of embedded processors that may be used as an audio processing engine. The audio processing engine may be an audio subsystem that allows for 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 having a digital signal processor with dedicated RAM.

[0123] Processor 1210 may further include an always-on processor engine that may 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 (e.g., timers and interrupt controllers), various I / O controller peripherals, and routing logic.

[0124] The processor 1210 may further include a safety cluster engine, which 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.

[0125] Processor 1210 may further include a real-time camera engine, which may include a dedicated processor subsystem for handling real-time camera management.

[0126] Processor 1210 may further 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.

[0127] 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.

[0128] 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.

[0129] 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 actively performing 3D rendering, the video image compositor may be used to offload the GPU 1208 to improve performance and responsiveness.

[0130] SoC 1204 may further 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 further 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.

[0131] The SoC 1204 may further 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 further 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.

[0132] 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.

[0133] 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.

[0134] In contrast to conventional systems, by providing a CPU complex, a GPU complex, and a hardware acceleration cluster, 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 further include a neural network capable of recognizing, interpreting, and providing semantic understanding of the signs, and passing that semantic understanding to a path planning module running on the CPU complex.

[0135] 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.

[0136] 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.

[0137] 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.

[0138] 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.

[0139] 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 on inputs (e.g., sensor data) from sensors of the vehicle 1200.

[0140] The vehicle 1200 may further 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., to 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.

[0141] 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.

[0142] The vehicle 1200 may further include data storage 1228, which may include off-chip (e.g., outside the SoC 1204) storage. 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.

[0143] The vehicle 1200 may further include a GNSS sensor 1258 (e.g., GPS and / or assisted GPS sensor) for assisting with 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.

[0144] The vehicle 1200 may further 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 ASILB. 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.

[0145] 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.

[0146] 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 on both ends of a rear bumper. When mounted on both ends of a rear bumper, such a RADAR sensor system may create two beams that continuously monitor the blind spots behind and beside the vehicle.

[0147] Short-range RADAR systems can be used in ADAS systems for blind spot detection and / or lane change assistance.

[0148] The vehicle 1200 may further 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.

[0149] 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 ASIL B for functional safety level. In some examples, the vehicle 1200 may include multiple LIDAR sensors 1264 (e.g., two, four, six, etc.) that may use Ethernet (e.g., to provide data to a Gigabit Ethernet switch).

[0150] In some examples, the LIDAR sensor 1264 may be able to provide a list of objects and their distances for a 360-degree field of view. Commercially available LIDAR sensors 1264 may have, for example, an advertised range of approximately 1400m, an accuracy of 2cm-3cm, and support for 100Mbps 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.

[0151] 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.

[0152] The vehicle may further 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, for example, in a six-axis application, the IMU sensor 1266 may include an accelerometer and a gyroscope, and in a nine-axis application, the IMU sensor 1266 may include an accelerometer, a gyroscope, and a magnetometer.

[0153] 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.

[0154] 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.

[0155] The vehicle may further 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.10 and Fig.11 Described in more detail.

[0156] The vehicle 1200 may further 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 freely rotating shaft).

[0157] 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.

[0158] 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 LC and CWS.

[0159] 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.

[0160] 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.

[0161] 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.

[0162] 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.

[0163] 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.

[0164] 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 merging or changing lanes is unsafe. 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 that is electrically coupled to driver feedback such as a display, speaker, and / or vibration components.

[0165] 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.

[0166] 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.

[0167] 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.

[0168] 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 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 and / or be included as a component of the SoC 1204.

[0169] 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 used by the main computer does not cause a substantial error.

[0170] 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.

[0171] The vehicle 1200 may further 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.

[0172] 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.

[0173] The vehicle 1200 may further include a dashboard 1232 (e.g., a digital dashboard, an electronic instrument cluster, a digital instrument panel, etc.). The dashboard 1232 may include a controller and / or a supercomputer (e.g., a discrete controller or a supercomputer). The dashboard 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 dashboard 1232. In other words, the dashboard 1232 may be included as part of the infotainment SoC 1230, or vice versa.

[0174] Fig.13 For a cloud-based server and Fig.10 1376. System 1376 may include server 1378, network 1390, and vehicles including vehicle 1300. Server 1378 may include multiple GPUs 1384(A)-1384(H) (collectively referred to herein as GPU 1384), PCIe switches 1382(A)-1382(H) (collectively referred to herein as PCIe switches 1382), and / or CPUs 1380(A)-1380(B) (collectively referred to herein as CPU 1380). GPU 1384, CPU 1380, and PCIe switch may be interconnected with a high-speed interconnect and / or PCIe connection 1386 such as, for example and without limitation, NVLink interface 1388 developed by NVIDIA. In some examples, GPU 1384 is connected via NVLink and / or NVSwitch SoC, and GPU 1384 and PCIe switch 1382 are connected via PCIe interconnect. Although eight GPUs 1384, two CPUs 1380, and two PCIe switches are illustrated, this is not intended to be limiting. Depending on the embodiment, each of the servers 1378 may include any number of GPUs 1384, CPUs 1380, and / or PCIe switches. For example, each of the servers 1378 may include eight, sixteen, thirty-two, and / or more GPUs 1384.

[0175] Server 1378 may receive image data over network 1390 and from a vehicle, the image data representing images showing unexpected or changed road conditions, such as recently begun road work. Server 1378 may transmit neural network 1392, updated neural network 1392, and / or map information 1394, including information about traffic and road conditions, over network 1390 and to the vehicle. Updates to map information 1394 may include updates to HD map 1322, such as information about construction sites, potholes, curves, flooding, or other obstacles. In some examples, neural network 1392, updated neural network 1392, and / or map information 1394 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 1378 and / or other servers).

[0176] Server 1378 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 one or more classes of machine learning techniques, including but not limited to the following classes: supervised training, semi-supervised training, unsupervised training, self-learning, reinforcement learning, joint learning, transfer learning, feature learning (including principal components and cluster analysis), multilinear subspace learning, manifold learning, representation learning (including sparse 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 1390), and / or the machine learning model can be used by server 1378 to remotely monitor the vehicle.

[0177] In some examples, server 1378 can receive data from the vehicle and apply the data to the latest real-time neural network for real-time intelligent reasoning. Server 1378 may include a deep learning supercomputer and / or a dedicated AI computer powered by GPU 1384, such as DGX and DGX Station machines developed by NVIDIA. However, in some examples, server 1378 may include a deep learning infrastructure of a data center powered only by CPUs.

[0178] The deep learning infrastructure of server 1378 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 1300. For example, the deep learning infrastructure may receive periodic updates from vehicle 1300, such as a sequence of images and / or objects located in the sequence of images that vehicle 1300 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 1300, and if the results do not match and the infrastructure concludes that the AI ​​in vehicle 1300 has failed, then server 1378 may transmit a signal to vehicle 1300 instructing the fail-safe computer of vehicle 1300 to take control, notify passengers, and complete a safe parking maneuver.

[0179] For reasoning, server 1378 may include GPU 1384 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.

[0180] Example computing device

[0181] Fig.14 14 is a block diagram of an example computing device 1400 suitable for implementing some embodiments of the present disclosure. The computing device 1400 may include an interconnect system 1402 that directly or indirectly couples the following devices: memory 1404, one or more central processing units (CPUs) 1406, one or more graphics processing units (GPUs) 1408, communication interfaces 1410, I / O ports 1412, input / output components 1414, power supplies 1416, one or more presentation components 1418 (e.g., one or more displays), and one or more logic units 1420.

[0182] Although Fig.14 The various blocks of are shown as being connected via an interconnect system 1402 having lines, but this is not intended to be limiting and is for purposes of clarity only. For example, in some embodiments, a presentation component 1418 such as a display device may be considered an I / O component 1414 (e.g., if the display is a touch screen). As another example, CPU 1406 and / or GPU 1408 may include memory (e.g., memory 1404 may represent a storage device in addition to the memory of GPU 1408, CPU 1406, and / or other components). In other words, Fig.14The computing devices described herein are illustrative only. No distinction is made between categories such as “workstation,” “server,” “laptop,” “desktop,” “tablet,” “client device,” “mobile device,” “handheld device,” “gaming console,” “electronic control unit (ECU),” “virtual reality system,” “augmented reality system,” and / or other device or system types, all of which are described herein. Fig.14 The scope of computing devices is considered.

[0183] Interconnection system 1402 can represent one or more links or buses, such as address bus, data bus, control bus or its combination.Interconnection system 1402 can include one or more bus or link types, such as industry standard architecture (ISA) bus, extended industry standard architecture (EISA) bus, video electronics standard association (VESA) bus, peripheral component interconnect (PCI) bus, peripheral component interconnect fast (PCIe) bus and / or another type of bus or link. In some embodiments, there is a direct connection between components. As an example, CPU 1406 can be directly connected to memory 1404. In addition, CPU 1406 can be directly connected to GPU 1408. In the case of direct or point-to-point connection between components, interconnection system 1402 can include PCIe link to perform connection. In these examples, PCI bus does not need to be included in computing device 1400.

[0184] Memory 1404 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 1400. Computer-readable media may include volatile and nonvolatile media and removable and non-removable media. By way of example and not limitation, computer-readable media may include computer storage media and communication media.

[0185] Computer storage media may include volatile and nonvolatile media and / or removable and non-removable media implemented in any method or technology for storing information such as computer-readable instructions, data structures, program modules, and / or other data types. For example, memory 1404 may store computer-readable instructions (e.g., representing programs and / or program elements, such as an operating system). Computer storage media may include, but are not limited to, RAM, ROM, EEPROM, flash memory or other storage technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, cassettes, magnetic tape, magnetic disk storage 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 1400. As used herein, computer storage media does not include the signals themselves.

[0186] Computer storage media may contain computer readable instructions, data structures, program modules, and / or other data types in a modulated data signal such as a carrier wave or other transmission mechanism, and include any information delivery media. The term "modulated data signal" may refer to a signal that has one or more of its characteristics set or changed in such a way that information is encoded into 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 sound, RF, infrared, and other wireless media. Any combination of the above should also be included within the scope of computer readable media.

[0187] The CPU 1406 may be configured to execute at least some of the computer-readable instructions in order to control one or more components of the computing device 1400 to perform one or more of the methods and / or processes described herein. Each of the CPUs 1406 may include one or more cores (e.g., one, two, four, eight, twenty-eight, seventy-two, etc.) capable of simultaneously processing a large number of software threads. The CPU 1406 may include any type of processor, and may include different types of processors, depending on the type of computing device 1400 implemented (e.g., a processor with fewer cores for mobile devices and a processor with more cores for servers). For example, depending on the type of computing device 1400, the processor may be an ARM processor implemented using reduced instruction set computing (RISC) or an x86 processor implemented using complex instruction set computing (CISC). In addition to one or more microprocessors or supplementary coprocessors such as mathematical coprocessors, the computing device 1400 may also include one or more CPUs 1406.

[0188] In addition to or alternatively to one or more CPUs 1406, one or more GPUs 1408 may be configured to execute at least some computer-readable instructions to control one or more components of the computing device 1400 to perform one or more of the methods and / or processes described herein. One or more GPUs 1408 may be integrated GPUs (e.g., with one or more CPUs 1406) and / or one or more GPUs 1408 may be discrete GPUs. In an embodiment, one or more GPUs 1408 may be coprocessors of one or more CPUs 1406. The computing device 1400 may use one or more GPUs 1408 to render graphics (e.g., 3D graphics) or perform general-purpose computing. For example, GPUs 1408 may be used for general-purpose computing (GPGPU) on GPUs. One or more GPUs 1408 may include hundreds or thousands of cores capable of processing hundreds or thousands of software threads simultaneously. GPUs 1408 may generate pixel data for outputting an image in response to a rendering command (e.g., a rendering command received from CPUs 1406 via a host interface). One or more GPUs 1408 may include graphics memory (e.g., display memory) for storing pixel data or any other suitable data, such as GPGPU data. Display memory may be included as part of memory 1404. GPU 1408 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 1408 may generate pixel data or GPGPU data for different parts of the output 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.

[0189] In addition to or as an alternative to the CPU 1406 and / or the GPU 1408, the logic unit 1420 may be configured to execute at least some of the computer-readable instructions to control one or more components of the computing device 1400 to perform one or more of the methods and / or processes described herein. In an embodiment, the one or more CPUs 1406, the one or more GPUs 1408, and / or the one or more logic units 1420 may discretely or jointly perform any combination of methods, processes, and / or portions thereof. The one or more logic units 1420 may be a part of and / or integrated into the one or more CPUs 1406 and / or the one or more GPUs 1408 and / or the one or more logic units 1420 may be a discrete component or otherwise located outside of the CPU 1406 and / or the GPU 1408. In some embodiments, the one or more logic units 1420 may be a coprocessor of the one or more CPUs 1406 and / or the one or more GPUs 1408.

[0190] Examples of logic unit 1420 include one or more processing cores and / or components thereof, such as 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 traversal 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 I / O element, a peripheral component interconnect (PCI) or a peripheral component interconnect express (PCIe) element, etc.

[0191] The communication interface 1410 may include one or more receivers, transmitters, and / or transceivers that enable the computing device 1400 to communicate with other computing devices via an electronic communication network, including wired and / or wireless communications. The communication interface 1410 may include components and functionality that enable communication over any of a number of different networks, such as wireless networks (e.g., Wi-Fi, Z-wave, Bluetooth, Bluetooth LE, ZigBee, etc.), wired networks (e.g., via Ethernet or InfiniBand communications), low power wide area networks (e.g., LoRaWAN, SigFox, etc.), and / or the Internet.

[0192] I / O ports 1412 can enable computing device 1400 to be logically coupled to other devices including I / O components 1414, presentation components 1418, and / or other components, some of which can be built into (e.g., integrated into) computing device 1400. Illustrative I / O components 1414 include microphones, mice, keyboards, joysticks, game pads, game controllers, satellite dishes, scanners, printers, wireless devices, and the like. I / O components 1414 can provide a natural user interface (NUI) that processes air gestures, voice, or other physiological inputs generated by a user. In some instances, the input can be transmitted to an appropriate network element for further processing. The NUI can implement any combination of voice recognition, stylus recognition, facial recognition, biometric recognition, gesture recognition on and adjacent to the screen, air gestures, head and eye tracking, and touch recognition associated with the display of computing device 1400 (as described in more detail below). Computing device 1400 may include a depth camera such as a stereo camera system, an infrared camera system, an RGB camera system, touch screen technology, and combinations of these for gesture detection and recognition. Additionally, computing device 1400 may include an accelerometer or gyroscope to enable motion detection (e.g., as part of an inertial measurement unit (IMU)). In some examples, the output of the accelerometer or gyroscope may be used by computing device 1400 to render immersive augmented reality or virtual reality.

[0193] The power supply 1416 may include a hardwired power supply, a battery power supply, or a combination thereof. The power supply 1416 may provide power to the computing device 1400 to enable the components of the computing device 1400 to operate.

[0194] The presentation component 1418 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 1418 may receive data from other components (e.g., GPU 1408, CPU 1406, etc.) and output the data (e.g., as an image, video, sound, etc.).

[0195] 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.

[0196] As used herein, the description of "and / or" about 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.

[0197] 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 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 system for object detection, comprising: one or more processors; as well as a memory storing instructions that, as a result of execution of the instructions by the one or more processors, cause the system to: capturing a first image and a second image, the first image being captured at a first time from a first location of a scene, the second image being captured at a second time from a second location of the scene; generating a scene structure graph based on the first image and the second image using a neural network, the scene structure graph representing planar homography estimates corresponding to one or more surfaces in the scene; and Objects in the scene are detected based at least in part on generating an enclosing shape that includes a subset of the set of values ​​of the scene structure graph, wherein the subset of values ​​indicates an area of ​​non-zero values.

2. The system of claim 1 , wherein the memory further comprises instructions that, as a result of execution of the instructions by the one or more processors, cause the system to: determining the planar homography estimate to warp the first image toward the second image based at least in part on a set of keypoints extracted from the first image and the second image; and A first warped image is generated by at least warping the first image toward the second image based at least in part on the planar homography estimate.

3. The system of claim 2, wherein the memory further comprises instructions that, as a result of execution of the instructions by the one or more processors, cause the system to: determining a residual optical flow based at least in part on the first warped image and the scene structure graph, wherein the residual optical flow is used to transform the first warped image into the second image by at least modeling a transformation of a subset of values ​​of the scene structure graph to corresponding pixels in the second image; and A second warped image is generated based at least in part on the first warped image, the scene structure map, and the residual optical flow.

4. The system of claim 3, wherein the memory further comprises instructions that, as a result of execution of the instructions by the one or more processors, cause the system to perform: determining a photometric difference between the second warped image and the second image; and Parameters of the neural network are modified based at least in part on the photometric difference. The system of claim 1 , wherein the first image and the second image are captured from a single camera.

6. The system of claim 1, wherein the instructions causing the system to detect objects in the scene further comprise instructions that, as a result of execution of the instructions by the one or more processors, cause the system to: generate the bounding shape based at least in part on dilation or erosion.

7. The system of claim 1, wherein the instructions causing the system to detect objects in the scene further comprise instructions that, as a result of execution of the instructions by the one or more processors, cause the system to: A connected component analysis is performed on pixel positions of the first image and the second image, the pixel positions corresponding to pixel positions associated with the subset of values.

8. The system of claim 1, wherein at least one of the one or more surfaces is a planar surface.

9. A computer-implemented method comprising: To train a deep neural network (DNN), perform at least the following steps: acquiring a first image captured during a first time interval and a second image captured during a second time interval; generating a first warped image by at least warping the first image toward the second image based at least in part on a planar homography; determining a residual optical flow based at least in part on the first warped image and a scene structure map generated by at least providing the first image and the second image as inputs to the DNN, wherein the scene structure map comprises height and depth ratios for particular pixels in the first image and the second image; generating a second warped image based at least in part on the first warped image and the residual optical flow; calculating a photometric loss between the first image and the second warped image; as well as Modifying parameters of the DNN based at least in part on the photometric loss.

10. The computer-implemented method of claim 9, wherein the method further comprises: acquiring a third image captured during a third time interval and a fourth image captured during a fourth time interval; generating a second scene structure graph by at least providing the third image and the fourth image as input to the DNN; as well as Obstacles in the second scene structure graph are determined based at least in part on a set of non-zero values ​​associated with a set of pixels represented in the second scene structure graph.

11. The computer-implemented method of claim 10, wherein the method further comprises: The planar homography is determined at least by determining a set of key points in the first image and the second image based at least in part on a set of matching features obtained by performing feature matching between the first image and the second image. 12 . The computer-implemented method of claim 10 , wherein the second scene structure graph indicates a height above a surface plane corresponding to a portion of a road surface.

13. The computer-implemented method of claim 9, wherein the first image and the second image are captured consecutively.

14. The computer-implemented method of claim 9, wherein the first image and the second image are captured from a single monocular camera.

15. A method for detecting an obstacle, comprising detecting the obstacle using a neural network trained according to the method of claim 9.

16. An autonomous driving vehicle comprising a system for detecting obstacles using a neural network trained using the method of claim 9.

17. A system for object detection, comprising: one or more processors; as well as a memory storing instructions that, as a result of execution of the instructions by the one or more processors, cause the system to: acquiring a first image of a scene and a second image of the scene, the first image and the second image being captured from different positions; generating a transformed first image by at least transforming the first image to the second image based at least in part on a planar homography estimate determined based at least in part on a set of features obtained from the first image and the second image; generating, based on the first image and the second image, a map associating height and depth ratios with corresponding locations in the scene using a neural network; further transforming the transformed first image into the second image using residual optical flow to generate a further transformed image; and The neural network is updated based on a measure of the difference between the further transformed image and the second image.

18. The system of claim 17, wherein the first image and the second image are non-consecutive images captured by a camera within a time interval.

19. The system of claim 18, wherein the camera is mounted in the vehicle facing forward.

20. The system of claim 19, wherein the memory further comprises instructions that, as a result of execution of the instructions by the one or more processors, cause the system to: A sampling rate of a set of images of which the first image and the second image are members is modified based at least in part on a speed associated with the vehicle.

21. The system of claim 19, wherein the memory further comprises instructions that, as a result of execution of the instructions by the one or more processors, cause the system to: A sampling rate of a set of images, wherein the first image and the second image are members of the set of images, is modified based at least in part on a steering wheel angle associated with the vehicle.

22. The system of claim 17, wherein the instructions that cause the system to generate the transformed first image further comprise instructions that, as a result of execution of the instructions by the one or more processors, cause the system to: A homography transformation is estimated based at least in part on a first subset of features of the set of features acquired from a first region of the first image and a second subset of features of the set of features acquired from a second region of the second image.

23. The system of claim 17, wherein the memory further comprises instructions that cause the system to extract the set of features at least by providing the first image and the second image as input to a consensus-based algorithm.

24. The system of claim 17, the map further comprising a set of values, wherein a first value in the set of values ​​comprises a ratio of a height and a depth of a pixel location in the first image.

Images