Vehicle irregular contour construction method and system based on direct regression

Abstract

The application discloses a vehicle irregular contour construction method and system based on direct regression, and belongs to the technical field of automatic driving. The method comprises the following steps: acquiring vehicle point cloud data of surrounding vehicles of an automatic driving vehicle by using a vehicle-mounted radar; processing the vehicle point cloud data by using a pre-trained contour key point prediction model, and determining contour key points corresponding to the surrounding vehicles in the vehicle point cloud data; connecting the contour key points in a preset order to obtain vehicle contours corresponding to the surrounding vehicles. The application processes the point cloud data of the surrounding vehicles acquired by the vehicle-mounted radar, determines a plurality of contour key points in the point cloud data, and obtains the contours of the surrounding vehicles according to the contour key points. The method can adapt to the acquisition of different vehicle postures and vehicle irregular contours, the acquired vehicle contours are more accurate, the method can be better applied in the automatic driving process of the vehicle, the precision of the process control in the automatic driving process is improved, and the surrounding vehicles can be better avoided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic driving, and in particular to a vehicle irregular contour construction method and system based on direct regression. BACKGROUND

[0002] In the process of automatic driving, in order to better improve the safety in the process of automatic driving and better avoid surrounding driving vehicles, the surrounding vehicles of the automatic driving vehicle are often detected, and a more optimal automatic driving route is planned. In the existing common method, the point cloud data of the surrounding vehicles is obtained by using the vehicle-mounted radar on the automatic driving vehicle, the point cloud data of the surrounding vehicles is processed by using the existing point cloud processing algorithm, and the corresponding rectangular vehicle frame of the surrounding vehicles is obtained. The position of the vehicle is represented by using the rectangular vehicle frame. However, there are problems in the prior art because the vehicle posture formats are various and there are various irregular contours. For example, the vehicle door is opened, and the vehicle turns. Especially for some large trucks or trailers, when they turn, the simple rectangular vehicle frame cannot completely represent the specific position of the vehicle, and the turning vehicle head is outside the rectangular vehicle frame. The rectangular vehicle frame is large, although it can ensure that the vehicle is completely covered, but the range represented by it has far exceeded the actual range of the vehicle, which is not conducive to the accuracy control of the automatic driving process. SUMMARY

[0003] In view of the problem in the prior art that the rectangular vehicle frame cannot accurately represent the position and shape of the surrounding vehicle when representing the position of the surrounding vehicle, especially for some irregular contours, a large deviation occurs, the present application provides a vehicle irregular contour construction method and system based on direct regression.

[0004] In one technical solution of the present application, a vehicle irregular contour construction method based on direct regression is provided, which comprises: obtaining vehicle point cloud data of a surrounding vehicle of an automatic driving vehicle by using a vehicle-mounted radar; processing the vehicle point cloud data by using a pre-trained contour key point prediction model, determining contour key points corresponding to the surrounding vehicle in the vehicle point cloud data; and connecting the contour key points in a predetermined order to obtain a vehicle contour corresponding to the surrounding vehicle.

[0005] Optionally, the vehicle irregular contour construction method based on direct regression of the present application further comprises: preprocessing the point cloud data obtained by the vehicle-mounted radar, and extracting vehicle point cloud data from the point cloud data according to target characteristics of the vehicle, the target characteristics including vehicle position characteristics.

[0006] Optionally, the point cloud data is processed by using a pre-trained contour key point prediction model to determine the contour key points corresponding to the surrounding vehicle in the point cloud data, including: determining a vehicle rectangular frame corresponding to the vehicle point cloud data according to the vehicle point cloud data, wherein the vehicle rectangular frame contains the vehicle point cloud data of the corresponding vehicle; determining foreground points in the vehicle point cloud data according to the vehicle rectangular frame; and dividing the foreground points into regions and determining the contour key points corresponding to each region.

[0007] Optionally, the foreground points in the entire point cloud data of the corresponding vehicle are determined according to the surrounding vehicle rectangular frame, including: calculating a first distance from each point cloud point in the vehicle point cloud data to the center of the vehicle rectangular frame and a second distance from the corresponding bounding box point of each point cloud point to the center of the vehicle rectangular frame in the vehicle rectangular frame; calculating the ratio of the first distance to the second distance; and if the ratio is greater than a preset threshold, determining the point cloud point corresponding to the ratio as a foreground point.

[0008] Optionally, the foreground points are divided into regions and the contour key points corresponding to each region are determined, including: constructing a plurality of sector regions according to the center of the vehicle rectangular frame and a preset angle threshold to divide the foreground points into regions; and selecting the point cloud point farthest from the center of the surrounding vehicle rectangular frame in the foreground points in each sector region as a contour key point.

[0009] Optionally, a plurality of sector regions are constructed according to the center of the vehicle rectangular frame and a preset angle threshold, including: constructing a plurality of sector regions with the same angle according to the center of the vehicle rectangular frame and a preset angle threshold to uniformly divide the foreground points into regions.

[0010] In one technical solution of the present application, a vehicle irregular contour construction system based on direct regression is provided, including: a vehicle-mounted radar that obtains vehicle point cloud data of a surrounding vehicle of an autonomous vehicle; and a processor that processes the vehicle point cloud data by using a contour key point prediction model to determine contour key points corresponding to the surrounding vehicle in the vehicle point cloud data, and connects the contour key points in a preset order to obtain a vehicle contour corresponding to the surrounding vehicle.

[0011] In one technical solution of the present application, a vehicle automatic driving control method is provided, including: obtaining vehicle point cloud data of a surrounding vehicle of an autonomous vehicle by using a vehicle-mounted radar; processing the vehicle point cloud data by using a pre-trained contour key point prediction model to determine contour key points corresponding to the surrounding vehicle in the vehicle point cloud data; connecting the contour key points in a preset order to obtain a vehicle contour corresponding to the surrounding vehicle; and performing driving control on the autonomous vehicle according to the vehicle contour to avoid the surrounding vehicle.

[0012] In one technical solution of the present application, a computer readable storage medium is provided, the storage medium stores computer instructions, and the computer instructions are operated to execute the method in solution one or solution three.

[0013] In one technical solution of the present application, a computer device is provided, which comprises a processor and a memory, and the memory stores computer instructions, wherein the processor operates the computer instructions to execute the method in solution one or solution three.

[0014] The present application has the beneficial effects that: the present application processes the point cloud data of surrounding vehicles obtained by the vehicle-mounted radar, determines a plurality of contour key points in the point cloud data, and obtains the contour of the surrounding vehicle according to the contour key points. It can adapt to the acquisition of different vehicle postures and irregular vehicle contours, the acquired vehicle contour is more accurate, can be better applied in the automatic driving process of the vehicle, improve the precision of the control in the automatic driving process, and better avoid the surrounding vehicles. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0016] Figure 1 is a flowchart of an embodiment of the vehicle irregular contour construction method based on direct regression of the present application.

[0017] Figure 2 is a flowchart of an embodiment of the vehicle irregular contour construction method based on direct regression of the present application;

[0018] Figure 3 is a flowchart of an embodiment of the vehicle irregular contour construction method based on direct regression of the present application;

[0019] Figure 4 is an example of the prospect point determination process in the present application;

[0020] Figure 5 is a schematic diagram of an example of the fan segmentation in the present application;

[0021] Figure 6 is a result example of the vehicle irregular contour construction method based on direct regression of the present application;

[0022] Figure 7 is a structural schematic diagram of an embodiment of the vehicle irregular contour construction system based on direct regression of the present application;

[0023] Figure 8 is a flowchart of an embodiment of the vehicle automatic driving control method of the present application.

[0024] The above-described drawings show specific embodiments of the present application, which will be described in more detail hereinafter. These drawings and the written description are not intended to restrict the scope of the present application in any way, but are merely meant to illustrate specific embodiments of the present application. DETAILED DESCRIPTION

[0025] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0026] The terms "first", "second", "third", "fourth" and the like (if any) in the description and claims of the present application and the above drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "comprise" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a product or device comprising a series of steps or units does not necessarily have to include only those units clearly listed, but can include other units not clearly listed or inherent to the product or device.

[0027] In the process of automatic driving, in order to better improve the safety in the process of automatic driving, and better avoid the surrounding driving vehicles, the surrounding vehicles of the automatic driving vehicle are often detected, and a better automatic driving route is planned. In the existing common method, the point cloud data of the surrounding vehicles is obtained by using the vehicle-mounted radar on the automatic driving vehicle, the point cloud data of the surrounding vehicles is processed by using the existing point cloud processing algorithm, and the corresponding rectangular vehicle frame of the surrounding vehicles is obtained. The position of the vehicle is represented by using the rectangular vehicle frame. However, there are problems in the prior art. Because the vehicle posture formats are various, there are various irregular contours. For example, the vehicle door is opened, and the vehicle turns. Especially for some large trucks or trailers, when they turn, the simple rectangular vehicle frame cannot completely represent the specific position of the vehicle, and the turning vehicle head is outside the rectangular vehicle frame. The rectangular vehicle frame is large, although it can completely cover the vehicle, but the range represented by it is far beyond the actual range of the vehicle, which is not conducive to the accuracy control of the automatic driving process.

[0028] In order to better and more accurately describe and construct the irregular contour of the vehicle, the application proposes a vehicle irregular contour construction method and system based on direct regression, which comprises the following steps: obtaining the point cloud data of the surrounding vehicles of the automatic driving vehicle by using the vehicle-mounted radar; processing the point cloud data by using a pre-trained contour key point prediction model to determine the contour key points corresponding to the surrounding vehicles in the point cloud data; and connecting the contour key points in a predetermined order to obtain the vehicle contour corresponding to the surrounding vehicles.

[0029] The vehicle irregular contour construction method based on direct regression of the application directly uses the point cloud data of the surrounding vehicles of the automatic driving vehicle to directly obtain a plurality of contour points of the surrounding vehicles, and connects the plurality of contour points to determine the contour of the vehicle. Compared with the method of representing the position of the vehicle by using the rectangular vehicle frame, the method of the application can more accurately represent the irregular contour of the vehicle in different postures, and greatly improves the safety of automatic driving and the accuracy of automatic driving control. The accuracy of the process control of the automatic driving is improved, and the surrounding vehicles are better avoided.

[0030] The technical solutions of the application and how the technical solutions of the application solve the above technical problems will be described in detail in the specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes can not be described again in some embodiments. The embodiments of the application will be described below with reference to the drawings.

[0031] Figure 1 A flowchart of an embodiment of the vehicle irregular contour construction method based on direct regression of the application is shown.

[0032] In Figure 1 In the embodiment shown, the vehicle irregular contour construction method based on direct regression of the application comprises a process S101 of obtaining vehicle point cloud data of a surrounding vehicle of an autonomous vehicle by using a vehicle-mounted radar.

[0033] In this embodiment, during the driving of the autonomous vehicle, the vehicle point cloud data of the surrounding vehicle of the autonomous vehicle is obtained by using a vehicle-mounted radar, and the obtaining method of the point cloud includes but is not limited to a laser radar device, a millimeter wave radar device or an ultrasonic radar device. The specific device for obtaining the point cloud can be appropriately selected according to different use environments and accuracy requirements, and the application does not make specific limitations.

[0034] Optionally, the vehicle irregular contour construction method based on direct regression further comprises: preprocessing the point cloud data obtained by the vehicle-mounted radar, and extracting vehicle point cloud data from the point cloud data according to target characteristics of the vehicle, wherein the target characteristics include vehicle position characteristics.

[0035] In this optional embodiment, in order to reduce the interference of non-vehicle points on the determination process of vehicle contour points and reduce the processing amount of point clouds, after obtaining the point cloud data of the vehicle-mounted radar of the autonomous vehicle, vehicle-related vehicle point cloud data can be extracted from the point cloud data according to different target characteristics of different point clouds, and noise points can be filtered out. On the one hand, the data amount of the model for processing the point cloud data can be reduced, the processing speed can be accelerated, noise points can be removed, and the accuracy and precision of the obtained vehicle contour can be improved. In the specific processing process, the vehicle point cloud data can be extracted according to the characteristics of the vehicle point cloud, such as being located at a certain distance above the road surface, and then the vehicle point cloud data related to the surrounding vehicle can be extracted from the numerous point cloud data, and then the subsequent determination process of the vehicle contour points can be performed, so as to reduce the data processing amount and improve the accuracy.

[0036] In Figure 1 In the embodiment shown, the vehicle irregular contour construction method based on direct regression of the application comprises a process S102 of utilizing a pre-trained contour key point prediction model to process the vehicle point cloud data, and determining the contour key points corresponding to the surrounding vehicle in the vehicle point cloud data.

[0037] In this embodiment, the contour key point prediction model is pre-trained by a convolutional neural network. In the actual autonomous driving process, the contour key point prediction model that has been trained is used to analyze and process the vehicle point cloud data of the surrounding vehicle obtained by the point cloud device, and a plurality of contour key points of the surrounding vehicle are determined in the point cloud data. The plurality of contour key points represent the contour of the surrounding vehicle. Compared with the vehicle rectangular frame in the prior art, the plurality of contour key points can better and more accurately represent the irregular contour of the surrounding vehicle. The accuracy of contour recognition is improved.

[0038] Optionally, Figure 2 A flowchart of one embodiment of the vehicle irregular contour construction method based on direct regression of the present application is shown.

[0039] In Figure 2 In the optional embodiment shown, the pre-trained contour key point prediction model is used to process the point cloud data, and the contour key points corresponding to the surrounding vehicle are determined in the point cloud data, including: process S201, determining the corresponding vehicle rectangular frame according to the vehicle point cloud data, wherein the surrounding vehicle rectangular frame contains the vehicle point cloud data of the corresponding vehicle.

[0040] In this optional embodiment, in the trained contour key point prediction model, the corresponding vehicle rectangular frame is extracted according to the vehicle point cloud data of the surrounding vehicle, wherein the vehicle rectangular frame can contain most of the point clouds, and the center of the vehicle rectangular frame can represent the center position of the surrounding vehicle. The identification process of the vehicle rectangular frame is a common practice in the prior art, and will not be described herein.

[0041] In Figure 2 In the optional embodiment shown, the pre-trained contour key point prediction model is used to process the point cloud data, and the contour key points corresponding to the surrounding vehicle are determined in the point cloud data, including: process S202, determining the foreground points in the vehicle point cloud data according to the vehicle rectangular frame.

[0042] In this optional embodiment, when the point cloud of the surrounding vehicle is obtained by the point cloud device on the autonomous vehicle, only the vehicle point cloud on one side relative to the autonomous vehicle can be obtained, and the vehicle point cloud facing away from the autonomous vehicle cannot be obtained. The foreground points are used to describe the vehicle point cloud that is visible and actually exists in the entire vehicle point cloud data relative to the autonomous vehicle, rather than the noise points or false detection point clouds on the back of the vehicle. By judging the foreground points, the wrong point clouds can be filtered out, and the accuracy of the finally obtained vehicle contour is further improved.

[0043] Optionally, Figure 3 A flowchart of one embodiment of the vehicle irregular contour construction method based on direct regression of the present application is shown.

[0044] In Figure 3 In the optional embodiment shown, determining the foreground points in the entire point cloud data of the corresponding vehicle according to the surrounding vehicle rectangular frame comprises: process S2021, in the vehicle rectangular frame, respectively calculating a first distance of each point cloud point in the vehicle point cloud data to the center of the vehicle rectangular frame and a second distance of the edge frame point corresponding to each point cloud point to the center of the vehicle rectangular frame.

[0045] In this optional embodiment, in order to determine the foreground points and avoid the influence of non-vehicle contour points on the judgment of vehicle contours, when judging the foreground points, the distance of the point cloud point from the center of the vehicle rectangular frame is used for judgment, if the distance meets certain conditions, the point cloud point is a contour point, if the distance does not meet the corresponding conditions, the contour point is a non-contour point and needs to be filtered out. Wherein, respectively calculating a first distance of each point cloud point in the entire point cloud data of the corresponding vehicle to the center of the surrounding vehicle rectangular frame and a second distance of the edge frame point corresponding to each point cloud point to the center of the surrounding vehicle rectangular frame. Wherein, the second distance is to extend from the center of the vehicle rectangular frame through the point cloud point until it intersects with the vehicle rectangular frame, wherein the distance between the intersection point and the center of the vehicle rectangular frame is the second distance.

[0046] In Figure 3 In the optional embodiment shown, determining the foreground points in the entire point cloud data of the corresponding vehicle according to the surrounding vehicle rectangular frame comprises: process S2022, calculating the ratio of the first distance to the second distance.

[0047] In this optional embodiment, considering different vehicle models and different vehicle sizes, the application adopts the ratio method for judgment when judging the foreground points. Thus, it can adapt to the prediction and construction of the contours of various vehicles. Therefore, the application calculates the ratio between the first distance and the second distance. If the ratio is small, it means that the point cloud point is close to the vehicle interior and has a low probability of being a foreground point; if the ratio is large, it means that the point cloud point is far from the vehicle interior and has a large probability of being a foreground point.

[0048] In Figure 3 In the optional embodiment shown, determining the foreground points in the entire point cloud data of the corresponding vehicle according to the surrounding vehicle rectangular frame comprises: process S2023, if the ratio is greater than the preset threshold, the point cloud point corresponding to the ratio is determined as the foreground point.

[0049] In this optional embodiment, the relationship between the ratio and the preset threshold is judged, if the ratio is greater than the preset threshold, it means that the corresponding point cloud point meets the vehicle contour point, then the point cloud point is determined as the foreground point.

[0050] Specifically, the preset threshold value can be selected as 0.2. It should be noted that the preset threshold value can be reasonably selected according to different judgment requirements and judgment accuracy, and the 0.2 mentioned in the present application is only a relatively optimal selection example. In this regard, the specific selection of the preset threshold value is not specifically limited in the present application.

[0051] Specifically, Figure 4 An example of the foreground point determination process in the present application is shown.

[0052] In Figure 4 In the example shown, the rectangular frame represents the vehicle rectangular frame, and the black points in the frame represent the point cloud points of the surrounding vehicle, wherein, for example, the record d1 of the distance between the point cloud point and the center point of the vehicle rectangular frame represents the first distance, and the distance between the intersection point obtained by continuing to extend from the center point of the vehicle rectangular frame to the point cloud point to the vehicle rectangular frame and the center point of the vehicle rectangular frame represents the second distance, such as d2 in Figure 4 The ratio of the first distance d1 and the second distance d2 is calculated, and it is determined whether the point cloud point is a foreground point.

[0053] In Figure 2 In the optional embodiment shown, the pre-trained contour key point prediction model is used to process the point cloud data to determine the contour key points corresponding to the surrounding vehicle in the point cloud data, including: process S203, regionally dividing the foreground points to determine the contour key points corresponding to each region.

[0054] In this embodiment, in order to accurately and comprehensively determine the vehicle contour points, after obtaining the foreground points, the foreground points are divided into different regions, and the contour key points of the corresponding regions are determined in different regions, so that the finally determined contour key points can fully and accurately express the specific posture and characteristics of the vehicle, and the accuracy is improved.

[0055] Optionally, the regionally dividing the foreground points to determine the contour key points corresponding to each region includes: constructing a plurality of fan-shaped regions according to the center of the vehicle rectangular frame and a preset angle threshold value, and regionally dividing the foreground points; and selecting, in each fan-shaped region, the point cloud point farthest from the center of the surrounding vehicle rectangular frame in the foreground points as the contour key point.

[0056] In the optional embodiment, in order to determine the vehicle contour points in different orientations, the application takes the center of the vehicle rectangular frame as the origin, constructs a plurality of sector regions with a preset angle threshold, so that the point cloud points of the vehicle are distributed on each sector region. By determining the contour key points in each sector region, the contour key points of the entire vehicle are determined. The contour points are the boundary points. In order to more accurately reflect the contour features of the vehicle, the point farthest from the center of the surrounding vehicle rectangular frame is selected as the contour key point in each sector region. Finally, a plurality of contour key points are obtained.

[0057] Optionally, constructing a plurality of sector regions according to the center of the vehicle rectangular frame and a preset angle threshold includes: constructing a plurality of sector regions with the same angle according to the center of the vehicle rectangular frame and a preset angle threshold, and uniformly dividing the foreground points into regions.

[0058] In the optional embodiment, when the regions are divided, the regions can be uniformly divided according to a preset angle to obtain sector regions with the same angle.

[0059] Specifically, the center of the vehicle rectangular frame is taken as the origin, and 10 degrees is taken as the preset angle threshold to divide the vehicle point cloud points. A total of 36 sector regions are determined, and the contour key points are determined in the 36 sector regions.

[0060] Figure 5 A schematic diagram of an example of the sector division of the application is shown.

[0061] As shown in Figure 5 , by dividing the sector regions, the vehicle point cloud points are divided into blocks, and the corresponding contour key points are calculated in each region, which can more accurately reflect the real vehicle contour.

[0062] In Figure 1 the embodiment shown, the vehicle irregular contour construction method based on direct regression of the application includes a process S103 of connecting the contour key points in a preset order to obtain the vehicle contour corresponding to the surrounding vehicle.

[0063] In the embodiment, after the vehicle contour points in different sector regions are determined, the contour key points are sequentially connected in a counterclockwise or clockwise order to obtain the vehicle contour corresponding to the surrounding vehicle.

[0064] Figure 6 is a result example of the vehicle irregular contour construction method based on direct regression of the application.

[0065] As shown in Figure 6As shown, the point cloud points pointed by each arrow are the finally determined contour key points. By determining the contour key points in different fan-shaped regions, the contour key points are finally connected in a counterclockwise or clockwise order, and the vehicle contour line is finally obtained.

[0066] The vehicle irregular contour construction method based on direct regression of the present application can process the point cloud data of the surrounding vehicle obtained by the vehicle-mounted radar, determine a plurality of contour key points in the point cloud data, and obtain the contour of the surrounding vehicle according to the contour key points. The vehicle irregular contour construction method based on direct regression of the present application can adapt to different vehicle postures and obtain the vehicle irregular contour. The obtained vehicle contour is more accurate, can be better applied in the automatic driving process of the vehicle, improve the control precision of the automatic driving process, and better avoid the surrounding vehicle.

[0067] Figure 7 A structural schematic diagram of one embodiment of the vehicle automatic driving control method of the present application is shown.

[0068] In the embodiment shown in the figure, the vehicle irregular contour construction system based on direct regression of the present application comprises: a vehicle-mounted radar 701 that obtains vehicle point cloud data of a surrounding vehicle of an automatic driving vehicle; a processor 702 that processes the vehicle point cloud data by using a contour key point prediction model, determines contour key points corresponding to the surrounding vehicle in the vehicle point cloud data, and connects the contour key points in a preset order to obtain a vehicle contour corresponding to the surrounding vehicle.

[0069] The vehicle irregular contour construction system based on direct regression of the present application can process the point cloud data of the surrounding vehicle obtained by the vehicle-mounted radar, determine a plurality of contour key points in the point cloud data, and obtain the contour of the surrounding vehicle according to the contour key points. The vehicle irregular contour construction system based on direct regression of the present application can adapt to different vehicle postures and obtain the vehicle irregular contour. The obtained vehicle contour is more accurate, can be better applied in the automatic driving process of the vehicle, improve the control precision of the automatic driving process, and better avoid the surrounding vehicle.

[0070] Figure 8 A structural schematic diagram of one embodiment of the vehicle automatic driving control method of the present application is shown.

[0071] In Figure 8In the illustrated embodiment, the vehicle automatic driving control method of the present application includes a process S801 of obtaining vehicle point cloud data of a surrounding vehicle of an autonomous vehicle using a vehicle-mounted radar; a process S802 of processing the vehicle point cloud data using a pre-trained contour key point prediction model to determine contour key points corresponding to the surrounding vehicle in the vehicle point cloud data; a process S803 of connecting the contour key points in a preset order to obtain a vehicle contour corresponding to the surrounding vehicle; and a process S804 of performing driving control on the autonomous vehicle according to the vehicle contour to avoid the surrounding vehicle.

[0072] In one specific embodiment of the present application, a computer-readable storage medium stores computer instructions, wherein the computer instructions are operated to perform the direct regression-based vehicle irregular contour construction method or the vehicle automatic driving control method described in any embodiment. Wherein the storage medium can be directly in hardware, in a software module executed by a processor, or a combination of both.

[0073] The software module can reside in the RAM memory, the flash memory, the ROM memory, the EPROM memory, the EEPROM memory, the register, the hard disk, the removable disk, the CD-ROM, or any other form of storage medium known in the art. The exemplary storage medium is coupled to the processor, such that the processor can read information from the storage medium and write information to the storage medium.

[0074] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof, etc. The general-purpose processor can be a microprocessor, but in the alternative, the processor can be any conventional processor, controller, microcontroller, or state machine. The processor can also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration. In the alternative, the storage medium can be integrated into the processor. The processor and the storage medium can reside in an ASIC. The ASIC can reside in the user terminal. In the alternative, the processor and the storage medium can reside as discrete components in the user terminal.

[0075] In an embodiment of the present application, a computer device includes a processor and a memory, and the memory stores computer instructions, wherein the processor operates the computer instructions to perform the direct regression based vehicle irregular contour construction method or the vehicle automatic driving control method described in any embodiment.

[0076] In the embodiments provided by the present application, it should be understood that the disclosed apparatus can be implemented in other manners. For example, the embodiments of the apparatus described above are merely schematic; for example, the division of the units is only a logical function division; there can be another division manner in actual implementation; for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between the units can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0077] The units described as separated components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. According to actual needs, some or all of the units can be selected to implement the purposes of the embodiments of the present application.

[0078] The above are merely embodiments of the present application, and do not limit the patent scope of the present application, and any equivalent structure transformation made by using the contents of the present application specification and drawings, or directly or indirectly applied to other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for constructing irregular vehicle contours based on direct regression, characterized in that, include: Vehicle point cloud data of surrounding vehicles are obtained using vehicle-mounted radar. The vehicle point cloud data is processed using a pre-trained contour key point prediction model to determine the contour key points corresponding to the surrounding vehicles in the vehicle point cloud data. Connect the key points of the outline in a preset order to obtain the vehicle outline corresponding to the surrounding vehicles. The process of using a pre-trained contour keypoint prediction model to process the vehicle point cloud data and determine the contour keypoints corresponding to the surrounding vehicles in the vehicle point cloud data includes: Based on the vehicle point cloud data, a corresponding vehicle bounding box is determined, wherein the vehicle bounding box contains the vehicle point cloud data of the corresponding vehicle. Within the vehicle rectangle, calculate the first distance from each point in the vehicle point cloud data to the center of the vehicle rectangle, and the second distance from the border point corresponding to each point in the vehicle point cloud data to the center of the vehicle rectangle. Calculate the ratio of the first distance to the second distance; If the ratio is greater than a preset threshold, then the point cloud point corresponding to the ratio is determined as the foreground point. Based on the center of the vehicle rectangle, and with a preset angle threshold, multiple fan-shaped regions are constructed to divide the foreground point into regions. Within each of the aforementioned sector-shaped regions, the point cloud point in the foreground point that is furthest from the center of the vehicle rectangle is selected as the contour key point.

2. The method for constructing irregular vehicle contours based on direct regression according to claim 1, characterized in that, The process involves constructing multiple sector regions based on the center of the vehicle's rectangular frame and a preset angle threshold, including: Based on the center of the vehicle rectangle, a preset angle threshold is used to construct multiple fan-shaped regions with the same angle, and the foreground point is uniformly divided into regions.

3. The method for constructing irregular vehicle contours based on direct regression according to claim 1, characterized in that, Also includes: The point cloud data acquired by the vehicle-mounted radar is preprocessed, and the vehicle point cloud data is extracted from the point cloud data according to the target characteristics of the vehicle, including vehicle position characteristics.

4. A system for constructing irregular vehicle contours based on direct regression, comprising: Vehicle-mounted radar acquires point cloud data of vehicles surrounding the autonomous vehicle. The processor uses a pre-trained contour key point prediction model to process the vehicle point cloud data, determines the contour key points corresponding to the surrounding vehicles in the vehicle point cloud data, and connects the contour key points in a preset order to obtain the vehicle contours corresponding to the surrounding vehicles. The process of using a pre-trained contour keypoint prediction model to process the vehicle point cloud data and determine the contour keypoints corresponding to the surrounding vehicles in the vehicle point cloud data includes: Based on the vehicle point cloud data, a corresponding vehicle bounding box is determined, wherein the vehicle bounding box contains the vehicle point cloud data of the corresponding vehicle. Within the vehicle rectangle, calculate the first distance from each point in the vehicle point cloud data to the center of the vehicle rectangle, and the second distance from the border point corresponding to each point in the vehicle point cloud data to the center of the vehicle rectangle. Calculate the ratio of the first distance to the second distance; If the ratio is greater than a preset threshold, then the point cloud point corresponding to the ratio is determined as the foreground point. Based on the center of the vehicle rectangle, and with a preset angle threshold, multiple fan-shaped regions are constructed to divide the foreground point into regions. Within each of the aforementioned sector-shaped regions, the point cloud point in the foreground point that is furthest from the center of the vehicle rectangle is selected as the contour key point.

5. A method for controlling autonomous driving of a vehicle, comprising: Vehicle point cloud data of surrounding vehicles are obtained using vehicle-mounted radar. The vehicle point cloud data is processed using a pre-trained contour key point prediction model to determine the contour key points corresponding to the surrounding vehicles in the vehicle point cloud data. Connect the key points of the outline in a preset order to obtain the vehicle outline corresponding to the surrounding vehicles. The autonomous vehicle is controlled based on its outline to avoid surrounding vehicles. The process of using a pre-trained contour keypoint prediction model to process the vehicle point cloud data and determine the contour keypoints corresponding to the surrounding vehicles in the vehicle point cloud data includes: Based on the vehicle point cloud data, a corresponding vehicle bounding box is determined, wherein the vehicle bounding box contains the vehicle point cloud data of the corresponding vehicle. Within the vehicle rectangle, calculate the first distance from each point in the vehicle point cloud data to the center of the vehicle rectangle, and the second distance from the border point corresponding to each point in the vehicle point cloud data to the center of the vehicle rectangle. Calculate the ratio of the first distance to the second distance; If the ratio is greater than a preset threshold, then the point cloud point corresponding to the ratio is determined as the foreground point. Based on the center of the vehicle rectangle, and with a preset angle threshold, multiple fan-shaped regions are constructed to divide the foreground point into regions. Within each of the aforementioned sector-shaped regions, the point cloud point in the foreground point that is furthest from the center of the vehicle rectangle is selected as the contour key point.

6. A computer-readable storage medium, characterized in that, The storage medium stores computer instructions that are operated to perform the method described in any one of claims 1-3 or 5.

7. A computer device comprising a processor and a memory, the memory storing computer instructions, wherein: The processor operates computer instructions to perform the method described in any one of claims 1-3 or 5.

Images