Image processing method, device and mobile tool

Abstract

The embodiment of the present application relates to an image processing method, comprising: obtaining collected original image information; the original image information comprises the size of a target; according to the size of the target, the original image information is cropped and / or spliced and / or processed to generate an image region of interest, so as to obtain image information of a preset pixel size to be detected; and the image information to be detected is input into a preset target detection model to generate target information. Through the image processing method, the original image can be enhanced, the calculation amount is effectively controlled, and the real-time performance of target detection is finally ensured.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to an image processing method, apparatus, and mobile tool. Background Technology

[0002] Roads typically contain various types of targets, including motor vehicles, non-motor vehicles, pedestrians, and special obstacles. Autonomous vehicles need to accurately and promptly identify these traffic obstacles to provide accurate and reliable information for downstream decision-making and control, ensuring the vehicle's safe, stable, and compliant operation on the road. To this end, autonomous vehicles are equipped with multiple sensors, including LiDAR, ultrasonic radar, and cameras, designed to provide road perception capabilities. Among these, cameras provide visual perception capabilities to the autonomous driving system; by analyzing the image data collected by the cameras, reliable status information is provided to downstream decision-making and control.

[0003] With the development of deep learning artificial intelligence technology, object detection technology, as one of the earliest and most mature technologies in deep learning, has been widely applied. In the field of autonomous driving, deep learning algorithms are now almost entirely used for obstacle and object detection. Current autonomous driving visual object detection solutions mainly rely on deep learning algorithms to design an end-to-end object detection network model and train it using pre-labeled data. Models trained on large amounts of data can achieve end-to-end object detection capabilities. The object detection model consists of multiple convolutional layers stacked in a specific structure, with each layer containing many convolutional kernels. Through training with large datasets, the parameters of each convolutional kernel in the network are continuously adjusted, enabling the network to automatically learn to extract image features and fit the correct results of the training data, which represent the coordinates and types of objects in the image. The trained model receives input image data, uses the network to extract features from the image, and calculates the coordinates and types of objects present in the output image, achieving end-to-end detection. Currently, there are many end-to-end object detection models in the field of autonomous driving, with representative examples including YOLO, SSD, and Faster R-CNN. However, deep learning algorithms have high requirements for computing resources and high computational complexity. How to detect obstacles quickly and accurately with limited computing resources has always been a problem that the industry is constantly exploring and urgently needs to solve.

[0004] In the field of autonomous driving, a prominent problem facing visual detection is the wide size distribution of obstacles that need to be detected. Taking an input image resolution of 720P (1280*720) as an example, small targets at a distance (200 meters) have a pixel size of less than 10 pixels; large targets nearby (within 20 meters), such as large freight vehicles, fill the entire image. If the visual detection algorithm needs to detect targets ranging from 10 to 1280 pixels, the most effective method for detecting small targets is to ensure a larger image resolution input, retain more details of small targets, and perform small target detection on high-resolution feature maps in the shallow layers of the network model. Detecting large targets requires the network model to have a larger receptive field and better feature representation; this requires a deeper network (more convolutional layers) and more channels in each convolutional layer.

[0005] Based on the above analysis, if good detection performance for both "small targets" and "large targets" is required, deep learning models need very high-resolution image input, and the network must be sufficiently deep with many convolutional layers and a large number of channels in each layer. However, deep learning models that meet these requirements have an extremely high computational cost and long computation latency. Coupled with very high image resolution input, the computational cost increases proportionally, making it impossible to meet the requirements of real-time target detection for autonomous driving. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing an image processing method that can guarantee the effectiveness and real-time performance of target detection.

[0007] To achieve the above objectives, in a first aspect, the present invention provides an image processing method, comprising:

[0008] Acquire the raw image information; the raw image information includes the size of the target;

[0009] Based on the size of the target, the original image information is cropped and / or stitched and / or a region of interest is generated to obtain image information to be detected with a preset pixel size;

[0010] The image information to be detected is input into a preset target detection model to generate target information.

[0011] A second aspect of the present invention provides an image processing apparatus, comprising:

[0012] The raw image information acquisition module is used to acquire the acquired raw image information, including the size of the target.

[0013] The image information generation module is used to crop and / or stitch and / or generate a region of interest in the original image information according to the size of the target, so as to obtain the image information to be detected with a preset pixel size;

[0014] The target information generation module is used to input the image information to be detected into a preset target detection model to generate target information.

[0015] In a third aspect, the present invention provides a computer server, comprising: a memory, a processor, and a transceiver;

[0016] The processor is configured to be coupled to the memory, read and execute instructions in the memory to implement the image processing method described in any of the first aspects above;

[0017] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0018] In a fourth aspect, the present invention provides a chip system including a processor coupled to a memory storing program instructions, wherein when the program instructions stored in the memory are executed by the processor, the image processing method described in any of the first aspects is implemented.

[0019] In a fifth aspect, the present invention provides a computer system including a memory and one or more processors communicatively connected to the memory;

[0020] The memory stores instructions that can be executed by the one or more processors to cause the one or more processors to implement the image processing method as described in any of the first aspects above.

[0021] In a sixth aspect, the present invention provides a mobile tool including the computer server described in the third aspect above.

[0022] The image processing method provided in this invention can perform multiple scaling operations on the original image information according to the preset pixel size and the size of the original image. At the image input end, it can make small targets larger and large targets smaller, thereby enhancing the image. Ultimately, it can detect targets of different sizes through the preset target detection model, effectively controlling the amount of computation and ensuring real-time target detection. Attached Figure Description

[0023] Figure 1 This is one of the flowcharts of the image processing method provided in the embodiments of the present invention;

[0024] Figure 2A schematic diagram illustrating the camera mounting position of an autonomous vehicle and image cropping for detecting targets of standard size, provided in an embodiment of the present invention.

[0025] Figure 3 This is a schematic diagram of image cropping processing for small-sized target detection provided in an embodiment of the present invention;

[0026] Figure 4 This is a schematic diagram of image stitching processing for large-size target detection provided in an embodiment of the present invention;

[0027] Figure 5 This is the second flowchart of the image processing method provided in the embodiments of the present invention;

[0028] Figure 6 This is a schematic diagram illustrating the positional relationship between vehicles and key points on a map, provided in an embodiment of the present invention.

[0029] Figure 7 This is a schematic diagram of image detection region of interest generation provided in an embodiment of the present invention;

[0030] Figure 8 This is a schematic diagram of the image processing method provided in an embodiment of the present invention;

[0031] Figure 9 This is a structural diagram of an image processing device module provided in an embodiment of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0033] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0034] The image processing method provided in this invention, when applied to the target detection process of autonomous vehicles, can improve the target detection model's ability to detect targets.

[0035] Example 1

[0036] Figure 1 This is a flowchart of an image processing method provided in an embodiment of the present invention, which is described below in conjunction with... Figure 1 The technical solution of the present invention will be described with reference to specific embodiments.

[0037] An image processing method provided by an embodiment of the present invention mainly includes the following steps:

[0038] Step 110: Obtain the acquired raw image information.

[0039] In a specific example, such as Figure 2 As shown, an autonomous vehicle (hereinafter referred to as the vehicle) has multiple cameras installed around its body, for example, seven cameras: two front-view cameras, one left front-view camera, one right front-view camera, one left rear-view camera, one right rear-view camera, and one rear-view camera, forming a 360-degree surround view to collect raw image information of the vehicle's driving environment from different angles. The raw image information includes the size of the target. Specific size types can include: standard size (e.g., the target's long side occupies less than 1 / 2 and more than 1 / 40 of the original image's long side), large size (the target's long side occupies more than 1 / 2 of the original image's long side), and small size (the target's long side occupies less than 1 / 40 of the original image's long side).

[0040] More specifically, for targets of standard size, such as Figure 2 As shown in the diagram, the image range of the raw image information captured by each camera is represented by a rectangle. The upper part of the rectangle, i.e., the area indicated by the diagonal lines in the diagram, occupies approximately 20% to 40% of the entire image. In practical applications, this 20% to 40% portion is almost entirely sky. Ordinary family cars, pedestrians, and non-motorized vehicles located 20 to 100 meters away from autonomous vehicles will almost never appear within this 20% to 40% area, meaning that this 20% to 40% portion is practically useless for object detection.

[0041] Large targets are typically captured when they are relatively close to the vehicle, meaning the camera's capture distance is no greater than a first threshold. This is an example, not a limitation; the first threshold could specifically be 20 meters. Large targets often have very large pixel areas, and a single target may even occupy the entire original image, such as... Figure 3 The shaded area is shown in the middle. In real-world applications, this could be a large truck in front of a vehicle.

[0042] For small targets, that is, targets whose acquisition distance by the camera is not less than the second threshold (not a limitation), this is an example; the second threshold could specifically be 200 meters. Figure 3 In this context, the portion represented by an ellipse typically represents a target located in the central region of the original image information. Due to the small size of the target pixels, the detection rate is extremely low.

[0043] Step 120: Based on the size of the target, crop and / or stitch and / or generate a region of interest in the original image information to obtain the image information to be detected with a preset pixel size.

[0044] Specifically, currently, the camera resolutions used in autonomous vehicles range from 1 million (1280*720) pixels to 8 million (3840*2160) pixels. Even with a 1-megapixel camera, directly inputting the raw image information into a pre-defined target detection model cannot meet the real-time requirements of the vehicle's processor. Therefore, the raw image information needs to be processed according to a preset pixel size and the size of the target. The preset pixel size can be determined based on the receptive field of the preset target detection model.

[0045] The following example uses an original image with an image size of 1280*720 pixels and a preset pixel size of 640*360 to illustrate the execution process of this step.

[0046] When the target size is standard, the original image information is cropped and reduced according to a preset first ratio to obtain the first image information to be detected with a preset pixel size. For example... Figure 2 As shown, parts of the image that are not useful for object detection can be cropped, such as the sky. In a specific example, the first ratio could be 20%. The remaining 80% of the image is scaled up for detection, making the computational cost only 80% of that before cropping. That is, a 640*288 pixel image is used as the image information to be detected.

[0047] When the target size is large, the original image information is scaled down according to a preset second ratio and then stitched together to obtain a second image information to be detected with a preset pixel size. For example... Figure 4 As shown, even the largest target in the processed original image information occupies less than 1 / 9 of the pixel area of ​​the stitched image, thus making the large target appear smaller.

[0048] When the target size is small, multiple regions of interest (ROIs) are generated based on vehicle location information and a map of the vehicle's driving area, resulting in a third image of the target size with a preset pixel dimensions. Detecting small targets becomes extremely difficult if the image is scaled down further. Therefore, it's necessary to generate ROIs centered on the small target, essentially enlarging it. These ROIs have preset dimensions and shapes. In this example, the preset dimensions are the preset pixel sizes required by the target detection model, and the preset shape is rectangular, as shown below. Figure 3The dashed rectangle shown in the image illustrates this. For straight roads, when the vehicle's heading is parallel to the road, the region of interest (ROI) for image detection is the center region of the image. However, when vehicles change lanes or the road curves, the ROI for image detection becomes uncertain. Furthermore, when there is an intersection ahead of the vehicle, and it is necessary to focus on targets on both sides of the intersection, it is difficult to determine the ROI based solely on the image. Therefore, the generation of the ROI for image detection is crucial.

[0049] The process of generating regions of interest in image detection is as follows: Figure 5 As shown:

[0050] Step 1201: Obtain vehicle location information and vehicle driving area map.

[0051] Specifically, vehicle location information can be obtained through a positioning module.

[0052] Step 1202: Based on the vehicle positioning information and the vehicle driving area map, determine the vehicle's coordinates and pose information in the vehicle driving area map.

[0053] Step 1203: Based on the coordinate information and pose information, determine the key points within the preset range of the vehicle.

[0054] Specifically, the preset range is determined based on the current vehicle speed, specifically the distance the vehicle can travel in 10 seconds at that speed. For example, if the current speed is 10 m / s, the preset range is 100 meters. Key points include key point coordinates. These coordinates include the coordinates of the road center point directly in front of the vehicle and the coordinates of the intersection's core point. Therefore, based on the vehicle's position and pose information, the coordinates of the road center point within the preset range directly in front of the vehicle can be determined, such as... Figure 6 P5 in the map shows that the core points of intersections are pre-marked on the vehicle's driving area map. Specifically, a core point can be the intersection of the intersection and the current straight-ahead intersection. Therefore, based on the vehicle's coordinate information, the coordinates of the core points of intersections within a preset range can be determined. For example... Figure 6 P1, P2, P3, and P4 are shown in the diagram.

[0055] Step 1204: Process the key point coordinates to generate the projected coordinates of the key points in the original image information.

[0056] Specifically, firstly, using the coordinates of the vehicle in the vehicle driving area map as the origin, the coordinates of the key points are transformed into the vehicle coordinate system to generate the coordinates of the key points in the vehicle coordinate system.

[0057] Secondly, taking the camera as the origin, the key point coordinates in the vehicle coordinate system are transformed into the camera coordinate system according to the pre-calibrated camera extrinsic parameters, thus generating the key point coordinates in the camera coordinate system.

[0058] Finally, based on the pre-calibrated camera intrinsic parameters, the keypoint coordinates in the camera coordinate system are projected to generate the projected coordinates of the keypoints in the original image information. These projected coordinates are the positions of the keypoints in the image.

[0059] Step 1205: Generate the region of interest for image detection based on the projection coordinates of the key points in the original image information.

[0060] Specifically, when there is no intersection within a preset range ahead of the vehicle, a region of interest (ROI) is generated centered on the projected coordinates of the road center point. When there is an intersection within the preset range ahead of the vehicle, an ROI is generated centered on the projected coordinates of the road center point; and, based on the projected coordinates of the intersection's core point, the coordinates of the lower corner point adjacent to the vehicle are calculated; and an ROI is generated based on the lower corner point coordinates.

[0061] The following is based on Figure 6 Taking the key points in the middle as an example, combined with Figure 7 This step will be explained.

[0062] For the intersection on the left, the projected coordinates of the core points P1 and P2 are P1' and P2'. Based on P1' and P2', calculate the coordinates of point A (x...). max ,y max Using the coordinates of point A as the coordinates of the lower right corner, the first image detection region of interest Z1 is generated.

[0063] For the intersection on the right, the projected coordinates of the core points P3 and P4 are P3' and P4'. Based on P3' and P4', calculate the coordinates of point B (x...). min ,y max Using the coordinates of point B as the coordinates of the lower left corner, the second image is used to detect the region of interest Z2.

[0064] The projected coordinates of the road center point P5 within a preset range ahead of the vehicle's current straight-ahead road are P5'. Using P5' as the center point, a third image is generated to detect the region of interest Z3.

[0065] In this example, the width of the regions of interest for the first, second, and third images can be 640, and the height can be 288.

[0066] Step 130: Input the image information to be detected into the preset target detection model to generate target information.

[0067] Specifically, the image information to be detected at the image input end has actually undergone image enhancement processing. Therefore, inputting it into the preset target detection model controls the computational load of the vehicle processor while achieving real-time detection of targets of different sizes. The target information here can specifically be obstacle information.

[0068] The image processing method provided by the embodiments of the present invention has been described above. To better understand the method of the present invention, the following will be combined with... Figure 8 Here is a brief introduction to the generation principle of the region of interest in image detection in this invention.

[0069] This invention combines vehicle positioning with high-precision maps. Key points in the high-precision map are transformed into the vehicle coordinate system based on the vehicle coordinates. Then, combined with camera extrinsic and intrinsic parameters, the key points in the high-precision map are projected onto the image, thereby generating a region of interest based on the projection of the key points.

[0070] The image processing method provided in this invention can perform multiple scaling operations on the original image information according to the preset pixel size and the size of the original image. At the image input end, it can make small targets larger and large targets smaller, thereby enhancing the image. Ultimately, it can detect targets of different sizes through the preset target detection model, effectively controlling the amount of computation and ensuring real-time target detection.

[0071] Example 2

[0072] Figure 9 This is a module structure diagram of an image processing device provided in Embodiment 2 of the present invention. The device includes:

[0073] The raw image information acquisition module 10 is used to acquire the acquired raw image information, including the size of the target.

[0074] The image information generation module 20 is used to crop and / or stitch and / or generate a region of interest in the original image information according to the size of the target, so as to obtain the image information to be detected with a preset pixel size.

[0075] The target information generation module 30 is used to input the image information to be detected into a preset target detection model to generate target information.

[0076] The image processing apparatus provided in Embodiment 2 of the present invention can execute the method steps in the above method embodiments. The original image information acquisition module 10 implements step 110, the image information generation module 20 to be detected implements step 120 and steps 1201 to 1205, and the target information generation module 30 implements step 130.

[0077] The specific implementation principles and technical effects are similar, and will not be elaborated here.

[0078] It should be noted that the division of the various modules in the above device is merely a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, these modules can be implemented entirely in software via processing element calls; they can be fully implemented in hardware; or some modules can be implemented by processing element calls to software, while others are implemented in hardware. For example, a determination module can be a separate processing element, or it can be integrated into a chip within the above device. Alternatively, it can be stored as program code in the memory of the above device, and its function can be called and executed by a processing element of the device. The implementation of other modules is similar. Moreover, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed through integrated logic circuits in the hardware of the processor element or through software instructions.

[0079] For example, these modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more Digital Signal Processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). As another example, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a System-on-a-Chip (SOC).

[0080] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. This computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The aforementioned computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The aforementioned computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the aforementioned computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, Bluetooth, microwave, etc.) means. The aforementioned computer-readable storage medium can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The aforementioned available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0081] Example 3

[0082] Embodiment 3 of the present invention provides a computer server, including: a memory, a processor, and a transceiver;

[0083] The processor is used to couple with the memory, read and execute instructions in the memory to implement the image processing method of any one of the above embodiments;

[0084] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0085] Example 4

[0086] Embodiment 4 of the present invention provides a chip system including a processor, the processor being coupled to a memory, the memory storing program instructions, and when the program instructions stored in the memory are executed by the processor, the image processing method of any one of the embodiments described above is implemented.

[0087] Example 5

[0088] Embodiment 5 of the present invention provides a computer system, including a memory and one or more processors communicatively connected to the memory;

[0089] The memory stores instructions that can be executed by one or more processors to cause one or more processors to implement the image processing method as described in any of the above embodiments.

[0090] Example 6

[0091] Embodiment 6 of the present invention provides a mobile tool, including the computer server described in Embodiment 3 above.

[0092] The mobile tool can be any movable tool, such as vehicles (e.g., floor scrubbers, vacuum cleaners, sweepers, logistics vehicles, passenger cars, buses, coaches, vans, trucks, heavy-duty trucks, trailers, drop trailers, cranes, excavators, bulldozers, road trains, sweepers, water trucks, garbage trucks, engineering vehicles, rescue vehicles, logistics carts, automated guided vehicles (AGVs), etc.), motorcycles, bicycles, tricycles, handcarts, robots, sweepers, balance scooters, etc. This application does not strictly limit the types of mobile tools, and will not list them all here.

[0093] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0094] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, software modules executed by a processor, or a combination of both. The software modules can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROM power system control methods, or any other form of storage medium known in the art.

[0095] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An image processing method, characterized in that, include: Acquire the raw image information; The original image information includes the size of the target; Based on the size of the target, the original image information is cropped and / or stitched and / or a region of interest is generated to obtain image information to be detected with a preset pixel size; The image information to be detected is input into a preset target detection model to generate target information; Specifically, the step of cropping and / or stitching and / or generating a region of interest in the original image information according to the size of the target to obtain image information to be detected of a preset pixel size includes: When the target size is a normal size, the original image information is cropped and reduced according to a preset first ratio to obtain a first image information to be detected with a preset pixel size; When the target is large, the original image information is reduced and stitched according to a preset second ratio to obtain the second image information to be detected with the preset pixel size. When the target is small, multiple regions of interest are generated based on vehicle positioning information and vehicle driving area map to obtain the third image information to be detected with the preset pixel size; Specifically, generating multiple regions of interest based on vehicle location information and a map of the vehicle's driving area includes: Obtain vehicle location information and a map of the vehicle's driving area; Based on the vehicle positioning information and the vehicle driving area map, determine the vehicle's coordinate information and pose information in the vehicle driving area map; Based on the coordinate information and pose information, key points within a preset range from the vehicle are determined, and the key points include key point coordinates. The key point coordinates are processed to generate the projected coordinates of the key points in the original image information; Based on the projection coordinates of the key points in the original image information, a region of interest for image detection is generated.

2. The image processing method according to claim 1, characterized in that, The process of processing the key point coordinates to generate the projected coordinates of the key points in the original image information specifically includes: Using the coordinate information of the vehicle in the vehicle driving area map as the origin, the coordinates of the key points are transformed into the vehicle coordinate system to generate the key point coordinates in the vehicle coordinate system. Using the camera as the origin, and based on the pre-calibrated camera extrinsic parameters, the key point coordinates in the vehicle coordinate system are transformed to the camera coordinate system to generate the key point coordinates in the camera coordinate system. Based on the pre-calibrated camera intrinsic parameters, the coordinates of key points in the camera coordinate system are projected to generate the projected coordinates of the key points in the original image information.

3. The image processing method according to claim 1, characterized in that, The step of determining the coordinates of key points within a preset range from the vehicle based on the coordinate information and pose information specifically includes: Based on the coordinate and pose information, determine the coordinates of the road center point within a preset range directly in front of the vehicle and the coordinates of the core point of the intersection within a preset range in front of the vehicle.

4. The image processing method according to claim 3, characterized in that, The step of generating a region of interest based on the projection coordinates of the key points in the original image information specifically includes: When there are no intersections within a preset range in front of the vehicle, the region of interest for image detection is generated with the projected coordinates of the road center point as the center. When there is an intersection within a preset range in front of the vehicle, the region of interest for image detection is generated with the projected coordinates of the road center point as the center; and the coordinates of the lower corner point adjacent to the vehicle are calculated based on the projected coordinates of the core point coordinates of the intersection; and the region of interest for image is generated based on the coordinates of the lower corner point.

5. An image processing apparatus, characterized in that, include: The raw image information acquisition module is used to acquire the raw image information collected. The original image information includes the size of the target; The image information generation module is used to crop and / or stitch and / or generate a region of interest in the original image information according to the size of the target, so as to obtain the image information to be detected with a preset pixel size; The image information generation module is specifically used for: When the target size is a normal size, the original image information is cropped and reduced according to a preset first ratio to obtain a first image information to be detected with a preset pixel size; When the target is large, the original image information is reduced and stitched according to a preset second ratio to obtain the second image information to be detected with the preset pixel size. When the target is small, multiple regions of interest are generated based on vehicle positioning information and vehicle driving area map to obtain the third image information to be detected with the preset pixel size; Specifically, generating multiple regions of interest based on vehicle location information and a map of the vehicle's driving area includes: Obtain vehicle location information and a map of the vehicle's driving area; Based on the vehicle positioning information and the vehicle driving area map, determine the vehicle's coordinate information and pose information in the vehicle driving area map; Based on the coordinate information and pose information, key points within a preset range from the vehicle are determined, and the key points include key point coordinates. The key point coordinates are processed to generate the projected coordinates of the key points in the original image information; Based on the projection coordinates of the key points in the original image information, a region of interest for image detection is generated. The target information generation module is used to input the image information to be detected into a preset target detection model to generate target information.

6. A computer server, characterized in that, include: Memory, processor, and transceiver; The processor is configured to be coupled to the memory, read and execute instructions in the memory to implement the image processing method according to any one of claims 1-4; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

7. A chip system, characterized in that, The device includes a processor coupled to a memory storing program instructions, which, when executed by the processor, implement the image processing method according to any one of claims 1-4.

8. A computer system, characterized in that, Includes a memory, and one or more processors communicatively connected to the memory; The memory stores instructions that can be executed by the one or more processors to cause the one or more processors to implement the image processing method as described in any one of claims 1-4.

9. A mobile tool, characterized in that, Includes the computer server described in claim 6 above.

Images