Vehicle control method and apparatus
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NAVINFO
- Filing Date
- 2021-05-18
- Publication Date
- 2026-06-26
AI Technical Summary
Existing deep learning-based object detection algorithms often fail to accurately determine signage areas, resulting in low accuracy of pose and position information for autonomous vehicles and impacting driving safety.
In the object detection model, an image segmentation branch and a key point detection branch are built to output the segmentation mask and key points of the sign. Based on this information, the corner coordinates of the sign are determined, thereby accurately locating the sign area.
This improved the accuracy of the signage area, which in turn improved the accuracy of pose and position information, thus enhancing the driving safety of autonomous vehicles.
Smart Images
Figure CN115439822B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of autonomous driving, and more particularly to a vehicle control method and apparatus. Background Technology
[0002] During operation, autonomous vehicles need to acquire their own pose and position information in real time, and combine this information with high-precision maps to control the vehicle's movement. A key step in acquiring their pose and position information is locating the sign area in the image captured by the camera.
[0003] In existing technologies, most object detection algorithms based on deep learning are used to determine the sign area. However, most deep learning object detection algorithms use classification plus regression. When regressing the sign area, the center point coordinates and width and height of the sign are regressed, and the rectangular area determined by the center point coordinates and width and height is taken as the sign area.
[0004] However, the rectangular area determined by the above method is not a sign area in the strict sense. The rectangular area includes both signs and non-sign areas. The pose and position information determined based on this rectangular area is not very accurate. Since pose and position information are important references during the driving process of autonomous vehicles, inaccurate pose and position information will undoubtedly affect the driving safety of autonomous vehicles. Summary of the Invention
[0005] This application provides a vehicle control method and apparatus. A target detection model is constructed with two branches: one branch outputs a segmentation mask of the target object, and the other branch outputs key points of the target object. Based on the segmentation mask and key points of the target object, the corner coordinates of a sign are determined, and then the sign area is determined based on these corner coordinates. Compared to rectangular areas in existing technologies, the sign area determined by this application is more accurate, and the pose and position information determined based on the sign area are also more accurate. Since pose and position information are important references during the operation of autonomous vehicles, this application's solution can improve vehicle driving safety.
[0006] In a first aspect, this application provides a vehicle control method, comprising: acquiring an image to be processed; acquiring a segmentation mask of a target object in the image to be processed and key points of the target object based on the image to be processed and a pre-trained target detection model; acquiring pose information and position information of a vehicle based on the segmentation mask of the target object and the key points of the target object; and controlling the vehicle to drive based on the pose information and position information.
[0007] Optionally, the vehicle's pose and position information are obtained based on the segmentation mask of the target object and the key points of the target object, including: determining the corner coordinates of the target object based on the segmentation mask of the target object and the key points of the target object; and obtaining the pose and position information based on the corner coordinates of the target object.
[0008] Optionally, determining the corner coordinates of the target object based on the segmentation mask of the target object and the key points of the target object includes: obtaining the segmentation contour of the target object based on the segmentation mask of the target object; performing polygon approximation on the segmentation contour of the target object to obtain an approximate polygon corresponding to the segmentation contour; and determining the corner coordinates of the target object based on the vertex coordinates of the approximate polygon and the key point coordinates of the target object.
[0009] Optionally, determining the corner coordinates of the target object based on the vertex coordinates of the approximate polygon corresponding to the segmented contour and the key point coordinates of the target object includes: for each vertex on the approximate polygon corresponding to the segmented contour, obtaining the key point corresponding to the vertex from the key points of the target object; and averaging the coordinates of the vertex and the coordinates of the key point corresponding to the vertex to obtain the corner coordinates of the target object.
[0010] Optionally, obtaining the pose information and the position information based on the corner coordinates of the target object includes: determining the region of the target object on the image to be processed based on the corner coordinates of the target object; extracting feature information of the region; determining a first object matching the target object in the high-precision map based on the feature information of the region and the feature information of all objects in the high-precision map; and obtaining the pose information and the position information based on the corner coordinates of the target object and the three-dimensional coordinates of each corner of the first object.
[0011] Optionally, obtaining the pose information and the position information based on the corner coordinates of the target object and the three-dimensional coordinates of each corner of the first object includes: obtaining the pose information and the position information through a pose estimation algorithm (PnP) based on the corner coordinates of the target object and the three-dimensional coordinates of each corner of the first object.
[0012] Secondly, this application provides a vehicle control device, comprising: an acquisition module, configured to acquire an image to be processed, and further configured to acquire a segmentation mask of a target object in the image to be processed and key points of the target object based on the image to be processed and a pre-trained target detection model, and further configured to acquire pose information and position information of a vehicle based on the segmentation mask of the target object and the key points of the target object; and a control module, configured to control the vehicle to drive based on the pose information and the position information.
[0013] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method provided in the first aspect.
[0014] Fourthly, this application provides a navigation device, including: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to implement the method provided in the first aspect by executing the executable instructions.
[0015] The vehicle control method and apparatus provided in this application establish two branches in the target detection model: an image segmentation branch and a key point detection branch. After the image acquisition device acquires an image, it sends the acquired image to the target detection model. The image segmentation branch outputs the segmentation mask of the target object in the image to be processed, and the key point detection branch outputs the key points of the target object in the image to be processed. Based on the segmentation mask and key points of the sign, the corner coordinates of the sign can be determined, and then the sign area can be determined based on the corner coordinates. Compared with the rectangular area in the prior art, the sign area determined by the scheme of this application is more accurate. Therefore, the pose and position information determined based on the sign area are also more accurate, resulting in higher vehicle driving safety. Attached Figure Description
[0016] Figure 1 Schematic diagram of the signage area provided in this application Figure 1 ;
[0017] Figure 2 The schematic diagram of the target detection model provided in this application;
[0018] Figure 3 A schematic flowchart of an embodiment of the vehicle control method provided in this application;
[0019] Figure 4 The schematic diagram for obtaining approximate polygons provided in this application;
[0020] Figure 5 A schematic diagram of key points provided for this application;
[0021] Figure 6Schematic flowchart of the second embodiment of the vehicle control method provided by this application;
[0022] Figure 7 Schematic diagram of the signboard area provided by this application Figure 2 ;
[0023] Figure 8 Schematic structural diagram of the vehicle control device provided by this application;
[0024] Figure 9 Schematic hardware structure diagram of the navigation device provided by this application. Detailed implementation manners
[0025] To make the objectives, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below with reference to the accompanying drawings in this application. Apparently, the described embodiments are some but not all of the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art without making creative efforts based on the embodiments in this application belong to the scope protected by this application.
[0026] In this application, it should be explained that the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance. In addition, "at least one" means one or more, and "multiple" means two or more. "And / or" describes the association relationship of associated objects and indicates that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone, where A and B may be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship. "At least one (item)" or similar expressions thereof refer to any combination of these items, including any combination of single item (item) or plural items (items). For example, at least one (item) of a, b, or c may represent: a alone, b alone, c alone, the combination of a and b, the combination of a and c, the combination of b and c, or the combination of a, b, and c, where a, b, and c may be single or multiple.
[0027] Since an autonomous vehicle has no driver, during its driving process, it needs to obtain its own pose information and position information in real time, and control the vehicle to drive in combination with the pose information, position information and high-precision map. In the prior art, the pose information and position information are obtained through the following methods: The autonomous vehicle periodically collects images through an image acquisition device, finds the signboard area on the collected images, and determines the pose information and position information of the autonomous vehicle based on the signboard area.
[0028] In existing technologies, most object detection algorithms are based on deep learning to determine the sign area. However, most deep learning object detection algorithms use a classification plus regression approach. When regressing the sign area, they regress the coordinates of the sign's center point and its width and height, and then use the rectangular area determined by the center point coordinates and the width and height as the sign area. Figure 1 The rectangular area is indicated by the dashed box. However, the rectangular area determined by the above method is not a sign area in the strict sense. This rectangular area includes both signs and non-sign areas, and the pose and position information determined based on this rectangular area is not very accurate.
[0029] To solve the above technical problems, see [link to relevant documentation]. Figure 2 As shown, this application proposes to build two branches in the target detection model: an image segmentation branch and a key point detection branch. After the image acquisition device acquires an image, it sends the acquired image to the target detection model. The target detection model can then output the segmentation mask and key points of the sign. Based on the segmentation mask and key points, the corner coordinates of the sign can be determined, and the sign region can be determined based on the corner coordinates. Compared with the rectangular region in the prior art, the sign region determined by the scheme of this application is more accurate, and therefore the pose and position information determined based on the sign region are also more accurate.
[0030] It should be noted that the solution in this application can be executed by any software and / or hardware with corresponding processing capabilities, such as a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), etc., and this application is not limited to any of them.
[0031] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.
[0032] Figure 3 A flowchart illustrating an embodiment of the vehicle control method provided in this application. The vehicle control method provided in this embodiment includes:
[0033] S301. Obtain the image to be processed.
[0034] In one possible implementation, an image acquisition device can be installed on the autonomous vehicle. This image acquisition device can periodically acquire images. In this embodiment, the images acquired by the image acquisition device are referred to as images to be processed.
[0035] Optionally, the aforementioned image acquisition device can be any device capable of implementing the solution of this application, such as a camera, mobile phone, etc., and this application is not limited to this.
[0036] The image to be processed is a two-dimensional plane. The origin of the coordinate system is the top left corner of the image. The X-axis extends to the right and the Y-axis extends downwards. This allows us to determine the two-dimensional coordinates of each pixel in the image to be processed.
[0037] In one possible implementation, when the solution of this embodiment is executed by the processor, the processor and the image acquisition device can be connected. After the image acquisition device acquires the image, it can transmit the image and the two-dimensional coordinates of each pixel in the image to the processor.
[0038] S302. Based on the image to be processed and the pre-trained target detection model, obtain the segmentation mask of the target object in the image to be processed and the key points of the target object.
[0039] The object detection model comprises two branches: an image segmentation branch and a keypoint detection branch. After the image to be processed is input into the model, each branch processes the image. The image segmentation branch outputs the segmentation mask of the target object in the image, while the keypoint detection branch outputs the keypoints of the target object in the image.
[0040] The object detection model described above can be a convolutional neural network model. A training set can be pre-constructed, which consists of many training samples. Each training sample includes an image, a segmentation mask of the target object in the image, and the key point coordinates of the target object. The segmentation mask and key point coordinates of each training sample can be manually labeled. After obtaining the training set, the convolutional neural network model is trained using the training set. Training can be stopped when the model converges.
[0041] S303. Based on the segmentation mask of the target object and the key points of the target object, obtain the pose information and position information of the vehicle.
[0042] In one possible implementation, the corner coordinates of the target object can be determined based on the segmentation mask and key points of the target object; and the pose and position information of the vehicle can be obtained based on the corner coordinates of the target object.
[0043] The following describes how to determine the corner coordinates of a target object:
[0044] After obtaining the segmentation mask of the target object, the segmentation contour of the target object can be obtained based on the segmentation mask; then, the segmentation contour of the target object is approximated by polygons to obtain the approximate polygons corresponding to the segmentation contours; finally, the corner coordinates of the target object are determined based on the vertex coordinates of the approximate polygons and the key point coordinates of the target object.
[0045] As described above, the two-dimensional coordinates of each pixel in the image to be processed are known. After obtaining the approximate polygon in the above way, the coordinates of the pixel corresponding to the vertex of the approximate polygon are the coordinates of the vertex. Similarly, after obtaining the keypoint in the above way, the coordinates of the pixel corresponding to the keypoint are the coordinates of the keypoint.
[0046] In this embodiment, the segmentation mask for the target object is a bitmap. Each pixel in the segmentation mask has two pixel values, 0 and 1. A bitwise AND operation is performed between each pixel in the image to be processed and the corresponding pixel in the segmentation mask to extract the target object from the image. The outline of the extracted region can be used as the segmentation outline of the target object. The above-mentioned bitwise AND operation process can be found in existing technologies and will not be repeated here.
[0047] In one possible implementation, since signs are mostly polygons, after obtaining the segmented outline of the target object, the segmented outline of the target object can be approximated by polygons. Specifically, the inflection points on the segmented outline can be found, and the shape obtained by connecting each inflection point in sequence in one direction can be used as an approximate polygon.
[0048] In one possible implementation, for each vertex on the approximate polygon, the keypoint corresponding to that vertex can be obtained from the keypoints of the target object. Specifically, the coordinates of the vertex can be compared with the coordinates of all keypoints, and the keypoint closest to the vertex can be taken as the keypoint corresponding to that vertex. Then, the coordinates of the corner point of the target object can be obtained by averaging the coordinates of the vertex and the coordinates of the keypoint corresponding to that vertex.
[0049] The following example illustrates the process of determining the corner coordinates of the target object:
[0050] See Figure 4 As shown, Figure 4 From left to right, the image shows the segmentation mask of the target object, the segmentation contour of the target object, and the approximate polygon. For ease of description, the three vertices of the approximate polygon are represented by vertex 0, vertex 1, and vertex 2, respectively. Figure 5For the keypoint detection branch output, for ease of description, the three keypoints are represented by keypoint 0', keypoint 1', and keypoint 2' respectively. The keypoint corresponding to vertex 0 is keypoint 0', the keypoint corresponding to vertex 1 is keypoint 1', and the keypoint corresponding to vertex 2 is keypoint 2'. Taking the average of the coordinates of vertex 0 and keypoint 0', we can obtain the coordinates of one corner point of the target object. Taking the average of the coordinates of vertex 1 and keypoint 1', we can obtain the coordinates of another corner point of the target object. Taking the average of the coordinates of vertex 3 and keypoint 3', we can obtain the coordinates of yet another corner point of the target object.
[0051] S304. Control the vehicle's movement based on pose and position information.
[0052] Based on the vehicle's location information and planned route, the next driving direction of the vehicle can be determined; based on the vehicle's posture information and the next driving direction, the steering wheel can be controlled to rotate.
[0053] The vehicle control method provided in this embodiment establishes two branches in the target detection model: an image segmentation branch and a key point detection branch. After the image acquisition device acquires an image, it sends the acquired image to the target detection model. The image segmentation branch outputs the segmentation mask of the target object in the image to be processed, and the key point detection branch outputs the key points of the target object in the image to be processed. Based on the segmentation mask and key points of the sign, the corner coordinates of the sign can be determined, and then the sign area can be determined based on the corner coordinates. Compared with the rectangular area in the prior art, the sign area determined by the solution of this application is more accurate, and therefore the pose and position information determined based on the sign area are also more accurate.
[0054] Figure 6 A flowchart illustrating a second embodiment of the vehicle control method provided in this application. The vehicle control method provided in this embodiment includes:
[0055] S601. Obtain the image to be processed.
[0056] S602. Based on the image to be processed and the pre-trained target detection model, obtain the segmentation mask of the target object in the image to be processed and the key points of the target object.
[0057] S603. Determine the corner coordinates of the target object based on the segmentation mask of the target object in the image to be processed and the key points of the target object.
[0058] The implementation process of S601-S603 is similar to that of the above embodiments, and will not be described again in this embodiment.
[0059] S604. Determine the area of the target object on the image to be processed based on the corner coordinates of the target object.
[0060] In one possible implementation, after obtaining the corner coordinates of the target object, the area enclosed by connecting each corner in sequence in one direction can be used as the area of the target object on the image to be processed.
[0061] The following example illustrates this:
[0062] See Figure 7 As shown, assuming that the corner points of each target object are obtained through S601-S603, such as... Figure 7 The fork-shaped identifier illustrates how, for each target object, the corner points of that target object are connected sequentially in one direction to obtain... Figure 7 The area indicated by the dashed box in the image. Figure 1 Compared to the rectangular area, Figure 7 The area indicated by the dashed box is closer to the sign area.
[0063] S605. Extract feature information of the target object in the region of the image to be processed.
[0064] For ease of explanation, in this embodiment, the region of the target object on the image to be processed is referred to as the first region.
[0065] Optionally, features such as the Histogram of Oriented Gradient (HOG), Speeded Up Robust Features (SURF), or Local Binary Pattern (LBP) of the first region can be extracted; however, this embodiment does not limit this.
[0066] S606. Based on the feature information of the first region and the feature information of all objects in the high-precision map, determine the first object in the high-precision map that matches the target object.
[0067] Taking HOG features as an example, the HOG features of the first region are compared with the HOG features of all objects in the high-precision map. If the similarity between the HOG features of the first region and the HOG features of a certain object reaches a preset value, then that object is taken as the first object to match the target object.
[0068] S607. Obtain pose information and position information based on the corner coordinates of the target object and the three-dimensional coordinates of each corner of the first object.
[0069] In one possible implementation, after obtaining the corner coordinates of the target object and the three-dimensional coordinates of each corner of the first object, the pose and position information of the vehicle can be determined by a pose estimation algorithm (Perspective-N-Point, PnP).
[0070] S608: Control the vehicle's movement based on pose and position information.
[0071] The implementation process of S608 is similar to that of the above embodiments, and will not be described again in this embodiment.
[0072] The vehicle control method provided in this embodiment offers an implementation method for determining pose information and position information. Since the pose information and position information in this embodiment are calculated based on a first region, which is closer to the actual sign area than the rectangular region in the prior art, the pose information and position information determined by the method in this embodiment are also more accurate.
[0073] Figure 8 A schematic diagram of the vehicle control device provided in this application. Figure 8 As shown, the vehicle control device provided in this application includes:
[0074] The acquisition module 801 is used to acquire the image to be processed, and is also used to acquire the segmentation mask of the target object in the image to be processed and the key points of the target object based on the image to be processed and the pre-trained target detection model, and is also used to acquire the pose information and position information of the vehicle based on the segmentation mask of the target object and the key points of the target object.
[0075] The control module 802 is used to control the vehicle's movement based on the pose information and position information.
[0076] Optionally, module 801 is specifically used for:
[0077] Based on the segmentation mask of the target object and the key points of the target object, determine the corner coordinates of the target object;
[0078] Based on the corner coordinates of the target object, obtain the vehicle's pose and position information.
[0079] Optionally, module 801 is specifically used for:
[0080] Based on the segmentation mask of the target object, obtain the segmentation contour of the target object;
[0081] The segmentation contour of the target object is approximated by a polygon to obtain the approximate polygon corresponding to the segmentation contour;
[0082] The corner coordinates of the target object are determined based on the vertex coordinates of the approximate polygon and the key point coordinates of the target object.
[0083] Optionally, module 801 is specifically used for:
[0084] For each vertex on the approximate polygon corresponding to the segmented contour, the key point corresponding to the vertex is obtained from the key points of the target object; the coordinates of the vertex and the coordinates of the key point corresponding to the vertex are averaged to obtain the corner coordinates of the target object.
[0085] Optionally, module 801 is specifically used for:
[0086] Based on the corner coordinates of the target object, determine the region of the target object on the image to be processed;
[0087] Extract the feature information of the region;
[0088] Based on the feature information of the region and the feature information of all objects in the high-precision map, determine the first object in the high-precision map that matches the target object;
[0089] The pose information and position information are obtained based on the corner coordinates of the target object and the three-dimensional coordinates of each corner of the first object.
[0090] Optionally, module 801 is specifically used for:
[0091] Based on the corner coordinates of the target object and the three-dimensional coordinates of each corner of the first object, the pose information and position information are obtained through the pose estimation algorithm PnP.
[0092] The vehicle control device provided in this embodiment can be used to execute the steps in any of the above method embodiments. Its implementation principle and technical effect are similar, and will not be described again here.
[0093] Figure 9 A schematic diagram of the hardware structure of the navigation device provided in this application. Figure 9 As shown, the navigation device in this embodiment may include:
[0094] Memory 901 is used to store program instructions.
[0095] The processor 902 is used to implement the vehicle control method described in any of the above embodiments when the program instructions are executed. The specific implementation principle can be found in the above embodiments, and will not be repeated here.
[0096] This application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the vehicle control method described in any of the above embodiments.
[0097] This application also provides a program product comprising a computer program stored in a readable storage medium, wherein at least one processor can read the computer program from the readable storage medium, and the at least one processor executes the computer program to cause a navigation device to implement the vehicle control method described in any of the above embodiments.
[0098] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.
[0099] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0100] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in a combination of hardware and software functional units.
[0101] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0102] It should be understood that the processor described in this application can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in this application can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.
[0103] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.
Claims
1. A vehicle control method, characterized in that, include: The image to be processed is acquired using the image acquisition device on the vehicle; Based on the image to be processed and the pre-trained target detection model, the segmentation mask of the target object in the image to be processed and the key points of the target object are obtained. The target detection model includes two branches, namely an image segmentation branch and a key point detection branch. Based on the segmentation mask of the target object and the key points of the target object, obtain the vehicle's pose information and position information; The vehicle is controlled to move based on the pose information and the position information; Based on the segmentation mask of the target object and the key points of the target object, the pose information and position information of the vehicle are obtained, including: Based on the segmentation mask of the target object and the key points of the target object, determine the corner coordinates of the target object; Based on the corner coordinates of the target object, determine the region of the target object on the image to be processed; Extract the feature information of the region; Based on the feature information of the region and the feature information of all objects in the high-precision map, determine the first object in the high-precision map that matches the target object; The pose information and the position information are obtained based on the corner coordinates of the target object and the three-dimensional coordinates of each corner of the first object.
2. The method according to claim 1, characterized in that, Based on the segmentation mask of the target object and the key points of the target object, the corner coordinates of the target object are determined, including: Based on the segmentation mask of the target object, obtain the segmentation contour of the target object; The segmentation contour of the target object is approximated by a polygon to obtain the approximate polygon corresponding to the segmentation contour; The corner coordinates of the target object are determined based on the vertex coordinates of the approximate polygon corresponding to the segmented contour and the key point coordinates of the target object.
3. The method according to claim 2, characterized in that, Based on the vertex coordinates of the approximate polygon corresponding to the segmented contour and the key point coordinates of the target object, the corner coordinates of the target object are determined, including: For each vertex on the approximate polygon corresponding to the segmented contour, the key point corresponding to the vertex is obtained from the key points of the target object; the coordinates of the vertex and the coordinates of the key point corresponding to the vertex are averaged to obtain the corner coordinates of the target object.
4. The method according to any one of claims 1-3, characterized in that, Based on the corner coordinates of the target object and the three-dimensional coordinates of each corner of the first object, the pose information and the position information are obtained, including: Based on the corner coordinates of the target object and the three-dimensional coordinates of each corner of the first object, the pose information and the position information are obtained through the pose estimation algorithm PnP.
5. A vehicle control device, characterized in that, include: The acquisition module is used to acquire an image to be processed through an image acquisition device on the vehicle. It is also used to acquire a segmentation mask of a target object in the image to be processed and key points of the target object based on the image to be processed and a pre-trained target detection model. It is also used to acquire the pose information and position information of the vehicle based on the segmentation mask of the target object and the key points of the target object. The target detection model includes two branches, namely an image segmentation branch and a key point detection branch. The control module is used to control the vehicle's movement based on the pose information and the position information; The acquisition module is specifically used to determine the corner coordinates of the target object based on the segmentation mask and key points of the target object; determine the region of the target object on the image to be processed based on the corner coordinates of the target object; extract feature information of the region; determine a first object matching the target object in the high-precision map based on the feature information of the region and the feature information of all objects in the high-precision map; and acquire the pose information and the position information based on the corner coordinates of the target object and the three-dimensional coordinates of each corner of the first object.
6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1-4.
7. A computer program product, characterized in that, When the instructions contained in the computer program product are run on a computer, the computer causes the computer to perform the method according to any one of claims 1-4.
8. A navigation device, characterized in that, include: processor; as well as Memory for storing the executable instructions of the processor; The processor is configured to implement the method of any one of claims 1-4 by executing the executable instructions.