Obstacle data processing method, controller, vehicle and medium

Abstract

The present invention relates to the field of data processing technology, and in particular to an obstacle data processing method, a controller, a vehicle and a medium. The method comprises: determining a type detection result according to an obstacle image of an obstacle; when the type detection result is determined to be a first detection result, determining the obstacle position of the obstacle according to wheel position information corresponding to a target wheel; the first detection result indicates that the obstacle image contains a target vehicle having a target wheel; the target wheel includes two wheels located on the same side of the target vehicle. The present invention improves the positioning accuracy of the obstacle position.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an obstacle data processing method, a controller, a vehicle and a medium. Background Art

[0002] At present, in scenarios such as automatic parking where obstacles (such as vehicles) need to be located and tracked, the position of the obstacle is usually determined based on the image of the obstacle captured by the camera. The inventors have found that the images taken by the camera have different degrees of distortion. For example, the images taken by the fisheye camera have small distortion in the center and large distortion on both sides. In this way, due to the influence of image distortion, the final obstacle position may have a large error, and the positioning accuracy of the obstacle is poor. Summary of the invention

[0003] The present invention aims at solving the technical problem in the prior art that the obstacle position is poorly positioned when the obstacle image is used to determine the obstacle position, and provides an obstacle data processing method, a controller, a vehicle and a medium.

[0004] The present invention provides an obstacle data processing method, comprising:

[0005] Determine a type detection result based on an obstacle image of the obstacle;

[0006] When it is determined that the type detection result is a first detection result, the obstacle position of the obstacle is determined according to the wheel position information corresponding to the target wheel; the first detection result indicates that the obstacle image contains a target vehicle with a target wheel; the target wheel includes two wheels located on the same side of the target vehicle.

[0007] The present invention also provides a controller, comprising a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, wherein the processor implements the above-mentioned obstacle data processing method when executing the computer-readable instructions.

[0008] The present invention also provides a vehicle, comprising the above controller.

[0009] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the above obstacle data processing method is implemented.

[0010] In the obstacle data processing method, controller, vehicle and medium provided by the present invention, in the present invention, the type detection result is first determined according to the obstacle image of the obstacle, and then, when the type detection result is determined to be a first detection result for characterizing that the obstacle is a target vehicle including a target wheel, the obstacle position is determined using the wheel position information corresponding to the two wheels located on the same side of the target vehicle. In this way, the obstacle position corresponding to the target vehicle (i.e., the obstacle) is located by the wheel position information, which will reduce the error caused by image distortion, thereby improving the positioning accuracy of the obstacle position. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0012] Figure 1 is a flow chart of an obstacle data processing method in one embodiment of the present invention;

[0013] Figure 2 is a flow chart of step S10 of the obstacle data processing method in one embodiment of the present invention;

[0014] Figure 3 is a flow chart of an obstacle data processing method in another embodiment of the present invention;

[0015] Figure 4 is a flow chart of an obstacle data processing method in another embodiment of the present invention;

[0016] Figure 5 is a flow chart of step S50 of the obstacle data processing method in one embodiment of the present invention;

[0017] Figure 6 is a schematic diagram of an obstacle image corresponding to a first detection result in an embodiment of the present invention;

[0018] Figure 7 is a schematic diagram of a top view of a driving scene corresponding to a first detection result in an embodiment of the present invention;

[0019] Figure 8 is a schematic diagram of an obstacle image corresponding to the third detection result in one embodiment of the present invention;

[0020] Fig. 9 is a schematic diagram of a top view of a driving scene corresponding to a third detection result in an embodiment of the present invention;

[0021] Fig.10Schematic diagram of a controller in one embodiment of the present invention. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] The present invention provides an obstacle data processing method. In one embodiment, Figure 1 As shown, the obstacle data processing method includes the following steps S10-S20:

[0024] S10, determining a type detection result according to an obstacle image of the obstacle; wherein the type detection result is used to characterize whether the obstacle image contains a target vehicle having target wheels. The target wheels refer to the two wheels located on the same side of the target vehicle in the obstacle image. Therefore, the type detection result is actually used to characterize whether the obstacle image includes the target vehicle and the two wheels located on the same side of the target vehicle. wherein the type detection result may include but is not limited to the first detection result, the second detection result and the third detection result mentioned later. Specifically, the first detection result characterizes that the obstacle image contains a target vehicle having target wheels; that is, the first detection result is used to characterize that the obstacle image includes the target vehicle and the two wheels located on the same side of the target vehicle. The second detection result characterizes that the obstacle image does not contain the target vehicle; that is, the second detection result is used to characterize that the obstacle image does not include the target vehicle, and at this time, the obstacle image will certainly not include the two wheels located on the same side of the target vehicle. The third detection result characterizes that the obstacle image does not contain the target vehicle having target wheels. That is, the third detection result is used to represent that the obstacle image includes the target vehicle but does not include two wheels located on the same side of the target vehicle. At this time, the target vehicle in the obstacle image may include only one wheel or no wheel.

[0025] In one embodiment, if Figure 2 As shown, the step S10, i.e., determining the type detection result according to the obstacle image of the obstacle, includes:

[0026] S101, collecting obstacle images in real time through at least one camera installed on the driving vehicle, and determining current obstacle information based on all the obstacle images collected at the current moment; that is, in this embodiment, one or more cameras may be installed on the driving vehicle, for example, multiple cameras may be installed on the driving vehicle to achieve surround view, specifically, multiple (for example, four) fisheye cameras may be arranged at intervals around the vehicle, so that the 360° scene around the driving vehicle can be observed based on the obstacle images captured by multiple fisheye cameras. Among them, the obstacle image only refers to the image captured by the camera in real time, which may include obstacles of different obstacle types or may not include obstacles. The current obstacle information may refer to the obstacle type obtained by performing image recognition on the obstacle image or by identifying it through a preset obstacle recognition model, and the obstacle detection data corresponding to each different obstacle type, etc.

[0027] It is understandable that the obstacle type may include but is not limited to wall columns, pedestrians, motor vehicles, non-motor vehicles, wheel types, ground locks, traffic cones, traffic columns, limit blocks, no-parking signs, etc. In one embodiment, the obstacle data processing method is applied in an automatic parking scenario. Since the obstacle that has the greatest impact on the automatic parking process is a motor vehicle, it is necessary to particularly enhance the recognition accuracy of the motor vehicle. Therefore, in the automatic parking scenario, the target vehicle type (that is, the obstacle type of the target vehicle) corresponding to the above-mentioned target vehicle is a motor vehicle. In this embodiment, the wheel type will be identified as an independent obstacle type, that is, whether the target vehicle is identified in the obstacle image or not, the presence of the wheel type in the obstacle image will be identified. It is understandable that when certain obstacle types are identified in the obstacle image, the positions of obstacles corresponding to different obstacle types in the obstacle image will be marked in the obstacle image through different identifiers, and the above-mentioned position marks for different obstacle types in the obstacle image can be regarded as part of the obstacle detection data. For example, the obstacle detection frame can be used to select the positions of different obstacles in the obstacle image. The obstacle detection frame can be a square, and the square obstacle detection frame can be a frame that selects the entire obstacle detected in the obstacle image and has the smallest area. It is understandable that the obstacle detection frames corresponding to different obstacle types may have certain differences, such as different parameters such as color or frame thickness, brightness, etc. Figure 6 As shown, it is used to select a motor vehicle (such as Figure 6 The obstacle detection frame of the obstacle position corresponding to the obstacle type (target vehicle type) of the target vehicle T1 shown in the obstacle image can be called vehicle position information (for example Figure 6The second detection frame 12 is shown in FIG. 12 ); that is, the vehicle position information is the obstacle detection data marked in the obstacle image in the above-mentioned current obstacle information and associated with the target vehicle model. Figure 6 As shown in , it is used to select the wheel type in the obstacle image (the obstacle detection frame of the obstacle position corresponding to this obstacle type can be called wheel position information (for example Figure 6 The first detection frame 11 shown in FIG. 11). That is, in Figure 6 In the figure, the first detection frame 11 and the second detection frame 12 are both obstacle detection frames.

[0028] Further, in step S101, determining the current obstacle information according to all the obstacle images collected at the current moment includes: inputting all the obstacle images collected at the current moment into a preset obstacle recognition model, and acquiring the current obstacle information output by the preset obstacle recognition model.

[0029] The preset obstacle recognition model can be obtained based on neural network training. Specifically, the neural network model can be trained through sample images in the obstacle sample set, and each sample image is associated with a set of sample information; the sample information (corresponding to the above current obstacle information) includes the sample type in the sample image (corresponding to the above obstacle type) and the sample obstacle information corresponding to the sample type one by one (corresponding to the above obstacle detection data). The sample image is input into the neural network model. If the output detection result matches the sample information corresponding to the sample image, the detection result (corresponding to the above-mentioned current obstacle information) includes the recognition type (corresponding to the above-mentioned obstacle type) obtained by the neural network model for image recognition of the sample image and the recognition obstacle information corresponding to the recognition type (corresponding to the above-mentioned obstacle detection data). Then, the recognition type is matched with the sample type, and the sample obstacle information is matched with the recognized obstacle. When any match fails, the initial parameters of the neural network model are modified, and the sample image in the obstacle sample set is recognized again by the neural network model after the initial parameters are modified. In this way, iterative training is performed according to the above process until the recognition type in the detection result obtained by the neural network model successfully matches the sample type, and the sample obstacle information successfully matches the recognized obstacle, confirming that the neural network model training is completed; at this time, the current neural network model is the preset obstacle recognition model. After recognizing the obstacle image, the trained preset obstacle recognition model can output the current obstacle information, that is, output the obstacle type and the obstacle detection data corresponding to each different obstacle type. It is understandable that the current obstacle information output by the preset obstacle recognition model may include an obstacle image marked with obstacle detection data and / or obstacle type, that is, the obstacle detection data and / or obstacle type may be output together with the obstacle image as the current obstacle information, and the obstacle detection data may include but is not limited to an obstacle detection frame for selecting an obstacle in the obstacle image. Figure 6 and Figure 8 In the figure, the first detection frame 11 and the second detection frame 12 are both obstacle detection frames.

[0030] S102, determining the type detection result according to the current obstacle information. The type detection result can be obtained according to the current obstacle information. For example, when the obstacle type of the current obstacle information includes the obstacle type corresponding to the target vehicle (such as a motor vehicle) and two wheel types, and when it is determined according to the obstacle detection data that both of the two wheel types belong to the target vehicle, it means that the obstacle image includes the target vehicle and the two wheels located on the same side of the target vehicle. At this time, the type detection result is the first detection result.

[0031] For another example, when the obstacle type of the current obstacle information does not include the obstacle type corresponding to the target vehicle (such as a motor vehicle), regardless of whether the obstacle type includes a wheel type, it means that the obstacle image does not include the target vehicle. At this time, the type detection result is the second detection result.

[0032] For another example, the obstacle types in the current obstacle information include the obstacle type corresponding to the target vehicle (such as a motor vehicle) and two wheel types, but when it is determined based on the obstacle detection data that these two wheel types do not belong to the target vehicle, it means that the obstacle image includes the target vehicle and two wheels that are not located on the same side of the target vehicle. At this time, the type detection result is the third detection result.

[0033] For another example, the obstacle type in the current obstacle information includes the obstacle type corresponding to the target vehicle (such as a motor vehicle) and a wheel type. At this time, regardless of whether the wheel type belongs to the target vehicle when it is determined based on the obstacle detection data, it means that the obstacle image does not include a target vehicle having two wheels located on the same side of the target vehicle. At this time, the type detection result is also the third detection result.

[0034] S20, when it is determined that the type detection result is the first detection result, the obstacle position of the obstacle is determined according to the wheel position information corresponding to the target wheel; the first detection result indicates that the obstacle image contains a target vehicle with a target wheel; the target wheel includes two wheels located on the same side of the target vehicle. That is, the first detection result is used to indicate that the obstacle image includes the target vehicle and the two wheels located on the same side of the target vehicle. When it is determined that the type detection result is the first detection result, the obstacle is the target vehicle with the target wheel. At this time, the wheel position information corresponding to the target wheel can be directly obtained from the current obstacle information, and then the obstacle position corresponding to the target vehicle is determined according to the above wheel position information.

[0035] In the above embodiment of the present invention, the type detection result is first determined according to the obstacle image of the obstacle, and then, when the type detection result is determined to be a first detection result for characterizing that the obstacle is a target vehicle including a target wheel, the obstacle position is determined using the wheel position information corresponding to the two wheels located on the same side of the target vehicle. In this way, the obstacle position corresponding to the target vehicle (i.e., the obstacle) is located by the wheel position information, which will reduce the error caused by image distortion, thereby improving the positioning accuracy of the obstacle position. In the above embodiment, since the obstacle position of the target vehicle is determined according to the wheel position information corresponding to the target wheel (i.e., the two wheels located on the same side of the target vehicle), the distortion of the obstacle image has a small effect on the error of determining the obstacle position, so that the positioning of the obstacle type of the target vehicle is more accurate. In this way, in the automatic parking scene, the accurate information of the obstacle position occupied by the target vehicle can be output, and then the accurate angle information of the driving vehicle that needs to be parked can be output according to the above obstacle position, and the angle reference can also be provided for the release of the radar oblique parking space. In addition, the system delay of the obstacle data processing method of the present invention is small and does not increase the calculation cost.

[0036] In one embodiment, the current obstacle information includes at least one obstacle type and obstacle detection data associated with each obstacle type; further, the step S102, i.e., determining the type detection result according to the current obstacle information, includes:

[0037] When it is determined that the obstacle type includes the target vehicle type and the wheel type, and it is determined according to the obstacle detection data that the wheel type and the target vehicle type meet the preset association condition, the type detection result is determined to be the first detection result; the target vehicle type refers to the vehicle type corresponding to the target vehicle. In this embodiment, the target vehicle type refers to the obstacle type of motor vehicle. After the obstacle type of the current obstacle information includes the obstacle type corresponding to the target vehicle (that is, the motor vehicle) and the wheel type, it can be determined according to the obstacle detection data whether the wheel type and the target vehicle type meet the preset association condition. When it is determined that the wheel type and the target vehicle type meet the preset association condition, it is determined that the obstacle is the target vehicle, and the target vehicle includes two wheels (that is, the target wheels) located on the same side in the obstacle image, and the type detection result is the first detection result.

[0038] In a specific embodiment, determining, according to the obstacle detection data, that the wheel type and the target vehicle type meet a preset association condition includes:

[0039] The vehicle position information and wheel position information are obtained from the obstacle detection data, wherein the vehicle position information is associated with the target vehicle model, and the wheel position information is associated with the wheel type. That is, the vehicle position information refers to the obstacle detection data associated with the target vehicle model in the current obstacle information; the wheel position information refers to the obstacle detection data associated with the wheel type in the current obstacle information; understandably, when it is recognized that certain obstacle types exist in the obstacle image, the obstacle detection data may include an obstacle detection box for selecting the positions of obstacles of different obstacle types in the obstacle image. Among them, if Figure 6 As shown, it is used to select a motor vehicle (such as Figure 6 The obstacle detection frame of the obstacle position corresponding to the obstacle type (target vehicle type) of the target vehicle T1 shown in the obstacle image can be called vehicle position information (for example Figure 6 , as shown in the second detection frame 12); and Figure 6 As shown in , it is used to select the wheel type in the obstacle image (the obstacle detection frame of the obstacle position corresponding to this obstacle type can be called wheel position information (for example Figure 6 The first detection frame 11 shown in FIG. 11). That is, in Figure 6 In the figure, the first detection frame 11 and the second detection frame 12 are both obstacle detection frames.

[0040] When it is determined that the vehicle position information includes the wheel position information corresponding to the target wheel, it is determined that the wheel type and the target vehicle type meet the preset association condition. That is, the vehicle position information includes the wheel position information corresponding to the target wheel, indicating that in the obstacle image, the wheel position information of two wheels corresponding to two wheel types is within the position range of the wheel position information corresponding to the target vehicle. Specifically, Figure 6 As shown, the second detection frame 12 corresponding to the target wheel contains two first detection frames 11. At this time, it can be determined that the wheels corresponding to the two first detection frames 11 are the target wheels of the target vehicle T1, and in the same obstacle image, the two wheels belonging to the same target vehicle must be located on the same side of the target vehicle. Therefore, it can be determined that the obstacle image includes the target vehicle and the two wheels located on the same side of the target vehicle, and it can be determined that the wheel type and the target vehicle type meet the preset association condition, and then the type detection result is determined to be the first detection result.

[0041] In this embodiment, based on the vehicle position information of the target vehicle and the wheel position information corresponding to the wheel type in the obstacle detection data, it is determined that the wheel type and the target vehicle type meet the preset association condition, and then the type detection result is determined. This can reduce the error effect of the obstacle image distortion on the determination of the type detection result, thereby further improving the positioning accuracy of the obstacle position determined according to the type detection result.

[0042] In one embodiment, if Figure 6 As shown, the wheel position information includes a first detection frame 11 for selecting a wheel corresponding to the wheel type; the vehicle position information includes a second detection frame 12 for selecting the target vehicle; further, determining that the vehicle position information includes the wheel position information corresponding to the target wheel includes:

[0043] When the second detection frame 12 includes two of the first detection frames 11, it is determined that the vehicle position information includes the wheel position information corresponding to the target wheel. That is, in this implementation, since each first detection frame 11 is used to select a wheel corresponding to a wheel type, and the second detection frame 12 is used to select the target vehicle T1 corresponding to the target vehicle type (motor vehicle), therefore, when the second detection frame 12 includes two of the first detection frames 11, at this time, it can be determined that the vehicle position information includes the wheel position information corresponding to the target wheel, thereby determining that the wheel type and the target vehicle type meet the preset association condition, and also indicating that the two wheels selected by the two first detection frames 11 are two wheels belonging to the same side of the target vehicle T1, and then determining that the type detection result is the first detection result.

[0044] In one embodiment, the step S20, i.e., determining the obstacle position of the obstacle according to the wheel position information corresponding to the target wheel, includes:

[0045] The wheel position information corresponding to the target wheel is determined from the current obstacle information, and the obstacle position of the target vehicle is determined based on the wheel position information, wherein the wheel position information includes two first detection frames for selecting two wheels located on the same side of the target vehicle. Understandably, when the type detection result is the first detection result, it means that the two wheels selected by the two first detection frames are two wheels on the same side of the target vehicle. At this time, the obstacle position occupied by the target vehicle can be determined based on the two first detection frames. In this way, the error effect of the distortion of the obstacle image on the determination of the obstacle position is reduced, thereby improving the positioning accuracy.

[0046] In one embodiment, if Figure 6 and Figure 7As shown, in the step S201, determining the obstacle position of the target vehicle according to the wheel position information includes:

[0047] The first positioning point A2 and the second positioning point B2 are marked in the top view of the driving scene; the top view of the driving scene is generated according to the obstacle image; the first positioning point and the second positioning point refer to the midpoints of the bottom edges of the two first detection frames (i.e. Figure 6 A1 and B1 in the figure are converted to the corresponding coordinate points in the top view of the driving scene; wherein, after all obstacle images at the current moment (such as four obstacle images taken by four fisheye cameras) are distorted, a single corrected obstacle image can be homographically transformed, and then all obstacle images of the side view image perspective are converted into top view perspective images; then, all the converted top view perspective images are spliced ​​to obtain a top view of the driving scene corresponding to the surrounding environment of the driving vehicle. The driving vehicle will be set at the center of the top view of the driving scene. It can be understood that when the obstacle type is a target vehicle, the obstacle image includes a side view of the target vehicle. Therefore, the top view of the driving scene that can be converted from multiple obstacle images also has a top view corresponding to the target vehicle. In the obstacle image, the bottom midpoints A1 and B1 of the two first detection frames are both tire grounding points, and the line between the bottom midpoints A1 and B1 of the two first detection frames is the grounding line between the target vehicle and the ground. Since two first detection frames 11 are marked in the obstacle image, the bottom midpoints A1 and B1 of the first detection frames can be converted and marked in the top view of the driving scene during the conversion to the top view, and the first positioning point A2 and the second positioning point B2 are obtained. Understandably, in the top view of the driving scene, the line between the first positioning point A2 and the second positioning point B2 corresponding to the bottom midpoints A1 and B1 of the two first detection frames after conversion is still the grounding line between the target vehicle and the ground, and the position accuracy will not change due to the distortion of the obstacle image and the error in the conversion process.

[0048] The obstacle position of the obstacle in the driving scene top view is determined according to the first positioning point A2 and the second positioning point B2 in the driving scene top view. It can be understood that in the driving scene top view, the line between the first positioning point A2 and the second positioning point B2 is the grounding line between the target vehicle and the ground, and the positions of the first positioning point A2 and the second positioning point B2 will not change due to the distortion of the obstacle image and the error in the conversion process. In this way, the obstacle position of the obstacle in the driving scene top view is determined according to the first positioning point A2 and the second positioning point B2, which will improve the accuracy of the obstacle position finally obtained.

[0049] In one embodiment, if Figure 6 and Figure 7 As shown, determining the obstacle position of the obstacle in the driving scene top view according to the first positioning point and the second positioning point in the driving scene top view includes:

[0050] The target rectangular area enclosed by the two target vertical lines L2 and the length sideline L1 is determined as the obstacle position of the obstacle; the length sideline L1 refers to the line segment obtained by extending the first connecting line between the first positioning point A2 and the second positioning point B2 in the driving scene top view toward the two ends by a preset length; the two target vertical lines L2 refer to two vertical lines of a preset height extending in a direction away from the center point of the driving scene top view (that is, the center point of the driving vehicle 2) with the two end points of the length sideline L1 as the vertical foot. Among them, the length sideline L1 can be understood as the projection of the target vehicle corresponding to one side of the target tire in the driving scene top view. Since the driving vehicle is set at the center of the driving scene top view, the length sideline is the side edge line of the target vehicle close to the driving vehicle in the driving scene top view. That is, the side of the target vehicle closest to the driving vehicle corresponds to the length sideline L1. The preset length can be set according to actual needs, such as being set to 1m. The preset length represents the distance from the center of the target tire to the front bumper of the target vehicle, and the distance from the center of the target tire to the rear bumper. The preset height can be set according to actual needs, for example, it is set to 2m, and the preset height represents the width of the target vehicle. That is, the above two target vertical lines L2 can be regarded as the sides corresponding to the front and rear surrounding positions of the vehicle. The target rectangular area surrounded by the above length side line L1 and the two target vertical lines L2 is the area corresponding to the obstacle position occupied by the target vehicle in the top view of the driving scene. Therefore, after determining the above obstacle position of the target vehicle, the precise position occupied by the target vehicle is also determined. At this time, the relative position relationship between the driving vehicle and the target vehicle can be determined by the above length side line L1 and the two target vertical lines L2, wherein the angle corresponding to the length side line L1 is the current parking angle of the target vehicle; further, in the automatic parking scene, the angle required for the driving vehicle to be parked can be calculated based on the above target rectangular area. The present invention can solve the problem that radar parking spaces cannot release diagonal parking spaces. Specifically, when the radar parking space needs to be released, the above-mentioned target rectangular area can provide an angle reference, and the parking angle of the driving vehicle can be set to be consistent with the current parking angle corresponding to the length edge line of the target rectangular area of ​​the target vehicle, thereby realizing radar diagonal parking space.

[0051] In one embodiment, if Figure 3 As shown, after step S10, that is, after determining the type detection result according to the obstacle image of the obstacle, the following further includes:

[0052] S30, when it is determined that the type detection result is the second detection result or the third detection result, the obstacle position of the obstacle is determined according to the current obstacle information; the second detection result indicates that the obstacle image does not include the target vehicle; the third detection result indicates that the obstacle image does not include the target vehicle with the target wheels. That is, the second detection result is used to indicate that the obstacle image does not include the target vehicle. When it is determined that the type detection result is the second detection result, the obstacle is not the target vehicle. At this time, it is necessary to directly determine the obstacle position corresponding to the obstacle according to the current obstacle information. When it is determined that the type detection result is the third detection result, the obstacle is the target vehicle, but the target vehicle does not include two wheels (i.e., target wheels) located on the same side in the obstacle image. At this time, it is also necessary to directly determine the obstacle position corresponding to the obstacle according to the current obstacle information.

[0053] In one embodiment, the current obstacle information includes at least one obstacle type; further, the step S102, i.e., determining the type detection result according to the current obstacle information, includes:

[0054] When it is determined that the obstacle type does not include the target vehicle type, the type detection result is determined to be the second detection result; the target vehicle type refers to the vehicle type corresponding to the target vehicle; in this embodiment, the target vehicle type refers to the obstacle type of motor vehicle. After the obstacle type corresponding to the target vehicle (that is, motor vehicle) is not included in the obstacle type of the current obstacle information, it can be directly determined that the obstacle is not the target vehicle, and the type detection result is the second detection result. Specifically, it can be determined whether the obstacle image includes an obstacle of the obstacle type by detecting whether the obstacle image contains an obstacle detection frame corresponding to the obstacle type. For example, in Figure 6 and Figure 8 In the obstacle image shown, as long as the second detection frame 12 is not included, it can be determined that the obstacle type detection result in the obstacle image is the second detection result.

[0055] When it is determined that the obstacle type includes the target vehicle type but does not include the wheel type, the type detection result is determined to be the third detection result. In this embodiment, after the obstacle type of the current obstacle information includes the obstacle type corresponding to the target vehicle (i.e., the motor vehicle) but does not include the wheel type, the obstacle can be directly determined to be the target vehicle, but the target vehicle does not include wheels (i.e., the target wheels) in the obstacle image, and the type detection result is the third detection result. For example, if Figure 8 The second detection frame 12 in the obstacle image shown exists, but the first detection frame 11 therein does not exist, so it can be determined that the obstacle type detection result in the obstacle image is the third detection result.

[0056] In one embodiment, the current obstacle information includes at least one obstacle type and obstacle detection data associated with each obstacle type; further, the step S102, i.e., determining the type detection result according to the current obstacle information, includes: when it is determined that the obstacle type includes a wheel type and a target vehicle type, and when it is determined according to the obstacle detection data that the wheel type and the target vehicle type do not meet a preset association condition, determining that the type detection result is a third detection result; the target vehicle type refers to the vehicle type corresponding to the target vehicle. In this embodiment, after the obstacle type corresponding to the target vehicle (i.e., a motor vehicle) and a wheel type are included in the obstacle type of the current obstacle information, it can be determined according to the obstacle detection data whether the wheel type and the target vehicle type meet a preset association condition. When it is determined that the wheel type and the target vehicle type do not meet the preset association condition, it is determined that the obstacle is a target vehicle, and the target vehicle does not include two wheels (i.e., target wheels) located on the same side in the obstacle image, and at this time, the type detection result is the third detection result. For example, in Figure 8 In the example, since it includes a second detection frame 12 and a first detection frame 11, it can be determined that the obstacle type corresponding to the obstacle in the obstacle image includes the wheel type (one first detection frame 11 represents one wheel type) and the target vehicle type (one second detection frame 12 represents one target vehicle type). However, since the second detection frame 12 only includes one first detection frame 11, and there are no two first detection frames 11 included in the same second detection frame 12, the obstacle corresponding to the target vehicle type (motor vehicle) does not include two wheels (i.e., target wheels) located on the same side, and the type detection result is the third detection result.

[0057] Further, the determining, according to the obstacle detection data, that the wheel type and the target vehicle type do not meet a preset association condition includes:

[0058] The vehicle position information and wheel position information are obtained, wherein the vehicle position information refers to the obstacle detection data associated with the target vehicle type in the current obstacle information; the wheel position information refers to the obstacle detection data associated with the wheel type in the current obstacle information; it is understandable that when certain obstacle types are identified in the obstacle image, the obstacle detection data may include an obstacle detection frame for selecting the positions of obstacles of different obstacle types in the obstacle image. Among them, the obstacle detection frame used to select the obstacle position corresponding to the motor vehicle (target vehicle) in the obstacle image can be referred to as the vehicle position information (including the above-mentioned second detection frame); and the obstacle detection frame used to select the obstacle position corresponding to the obstacle type of the wheel type in the obstacle image can be referred to as the wheel position information (including the above-mentioned first detection frame).

[0059] When it is determined that the vehicle position information does not include the wheel position information corresponding to the target wheel, it is determined that the wheel type and the target vehicle model do not meet the preset association condition. That is, the vehicle position information does not include the wheel position information corresponding to the target wheel, indicating that in the obstacle image, there is no wheel type or only one wheel type corresponding to the wheel position information is subordinate to the position range of the wheel position information corresponding to the target vehicle. Specifically, it can be understood that the second detection frame corresponding to the target wheel contains a first detection frame or does not contain any first detection frame. At this time, it can be determined that the target vehicle in the obstacle image does not contain two wheels located on the same side. Therefore, it can be determined that the wheel type and the target vehicle model do not meet the preset association condition, and then the type detection result is determined to be the third detection result.

[0060] In one embodiment, the wheel position information includes a first detection frame for selecting a wheel corresponding to the wheel type; the vehicle position information includes a second detection frame for selecting the target vehicle; and determining that the vehicle position information does not include the wheel position information corresponding to the target wheel includes:

[0061] The second detection frame does not include the first detection frame, or includes only one first detection frame (such as Figure 8In the embodiment, when a second detection frame 12 and a first detection frame 11 are included, but only one first detection frame 11 is included in the second detection frame 12, it is determined that the vehicle position information does not include the wheel position information corresponding to the target wheel. That is, in this implementation, since each first detection frame is used to select a wheel corresponding to a wheel type, and the second detection frame is used to select a target vehicle corresponding to a target vehicle type (motor vehicle), when the second detection frame includes or does not include the first detection frame, it can be determined that the vehicle position information does not include the wheel position information corresponding to any wheel, thereby determining that the wheel type and the target vehicle type do not meet the preset association condition; when the second detection frame includes one first detection frame, it can be determined that the vehicle position information only includes the wheel position information corresponding to one wheel (rather than two wheels located on the same side of the target vehicle), thereby determining that the wheel type and the target vehicle type do not meet the preset association condition; and further, in both of the above cases, the type detection result can be determined to be the third detection result.

[0062] In one embodiment, the current obstacle information includes obstacle detection data; the obstacle detection data includes an obstacle detection frame for selecting obstacles corresponding to different obstacle types; Figure 6 and Figure 8 As shown, the first detection frame 11 and the second detection frame 12 are both obstacle detection frames. Further, the step S30, i.e., determining the obstacle position of the obstacle according to the current obstacle information, includes: determining the obstacle position of the obstacle according to the obstacle detection frame. The obstacle detection frame is used to select an obstacle in the obstacle image. At this time, the two end points of the bottom edge of the obstacle detection frame can be regarded as the grounding points between the obstacle and the ground. In this embodiment, the obstacle position of the obstacle can be determined using the two end points of the bottom edge of the obstacle detection frame.

[0063] In one embodiment, if Figure 8 and Fig. 9 As shown, determining the obstacle position of the obstacle according to the obstacle detection frame includes:

[0064] The third positioning point C2 and the fourth positioning point D2 are marked in the top view of the driving scene; the top view of the driving scene is generated according to the obstacle image; the third positioning point and the fourth positioning point refer to the two end points of the bottom edge of the obstacle detection frame (i.e. Figure 8C1 and D1 in the figure) are converted to the corresponding coordinate points in the overhead view of the driving scene; wherein, after all the obstacle images at the current moment (such as the four obstacle images taken by four fisheye cameras) are subjected to distortion correction, a single corrected obstacle image can be subjected to homography transformation, thereby converting all obstacle images of the side view image perspective into overhead view images; thereafter, all the converted overhead view images are spliced ​​to obtain a driving scene overhead view corresponding to the environment surrounding the driving vehicle. The driving vehicle will be set at the center of the overhead view of the driving scene. Understandably, since the obstacle detection frame corresponding to the obstacle T2 is marked in the obstacle image, the two endpoints of the bottom edge of the obstacle detection frame (i.e., C1 and D1) can be followed and converted during the conversion process and marked in the overhead view of the driving scene to obtain the following. Fig. 9 The third positioning point C2 and the fourth positioning point D2 shown in .

[0065] The second connecting line L3 between the third positioning point C2 and the fourth positioning point D2 in the top view of the driving scene is determined as the obstacle position of the obstacle. Fig. 9 In the top view of the driving scene shown in , the second line L3 between the third positioning point C2 and the fourth positioning point D2 is the obstacle position, and the side of the second line L3 away from the center point of the top view of the driving scene (that is, the center point of the driving vehicle 2) can be regarded as the area occupied by the obstacle T2; that is, when the type detection result is not the first detection result, since there is no wheel position information corresponding to the target wheel, it is necessary to determine the type of the obstacle T2 based on the third positioning point C2 and the fourth positioning point D2. Figure 8 The obstacle T2 shown in FIG. Fig. 9 After the top view of the driving scene shown in , the obstacle position corresponding to the obstacle T2, specifically, since the second line L3 can be used to define the area occupied by the obstacle T2, the second line L3 can be determined as the obstacle position. Therefore, after determining the obstacle position (that is, the second line L3), the relative position relationship between the driving vehicle 2 and the obstacle T2 can be further determined through the second line L3 (obstacle position).

[0066] In one embodiment, if Figure 4 As shown, after step S20 or step S30, that is, after determining the obstacle position of the obstacle, the following steps are included:

[0067] S40, matching the obstacle position with the historical obstacle position in the historical position set; the historical obstacle position is obtained by converting the historical position of the historical obstacle according to the current vehicle body posture.

[0068] It is understandable that at a certain historical moment in a preset historical period before the current moment, one or more historical obstacles have also been determined by the obstacle data processing method of the present invention, and each historical obstacle corresponds to an obstacle position also determined by the obstacle data processing method of the present invention, and the obstacle position can be called the initial obstacle position. Among them, when the driving vehicle is started, a world coordinate system is established with the rear axle center of the driving vehicle as the origin of the world coordinate system, and the current vehicle body posture refers to the x-coordinate, y-coordinate and heading angle of the rear axle center in the world coordinate system after the driving vehicle moves.

[0069] Since the initial obstacle position is the coordinate in the top view of the driving scene at the historical moment, in this embodiment, the initial obstacle position can be converted according to the historical body posture corresponding to the driving vehicle at the historical moment when the initial obstacle position is determined (that is, the current body position corresponding to the historical moment) to obtain the historical position of the initial obstacle position in the world coordinate system. As the driving vehicle moves, the origin of the world coordinate system changes. At this time, it is necessary to convert the historical position according to the current body posture in the changed current world coordinate system to obtain the historical obstacle position corresponding to the historical position in the top view of the driving scene at the current moment, so as to facilitate the comparison of the historical obstacle position with the obstacle position determined in step S20 or 30 of the above embodiment (also in the top view of the driving scene at the current moment).

[0070] Understandably, during the driving process, if the distance between the historical position and the center point of the driving vehicle in the current world coordinate system exceeds the preset distance range (for example, with the center point of the driving vehicle as the origin, within the distance range of 5m in front, back, left, and right), it can be considered that the historical obstacle corresponding to the historical position has little effect on the driving vehicle. Therefore, in this embodiment, the historical obstacle position in the historical position set should be the historical obstacle position corresponding to the historical obstacle within the preset distance range from the center point of the driving vehicle, and the historical obstacle position exceeding the preset distance range will be deleted from the historical position set. Understandably, the historical position set stores the obstacle types corresponding to different historical obstacles and their associated historical obstacle positions.

[0071] S50, when the obstacle position matches the historical obstacle position successfully, the historical obstacle position that matches successfully is updated according to the obstacle position; understandably, when the obstacle position matches the historical obstacle position successfully, it means that the obstacle corresponding to the obstacle position and the historical obstacle corresponding to the historical obstacle position are the same obstacle. At this time, it is necessary to update the historical obstacle position according to the obstacle position to determine the clearest obstacle position that is currently available for reference.

[0072] S60: When the obstacle position fails to match with all the historical obstacle positions, the obstacle position is added to the historical position set as a new historical obstacle position. Understandably, when the obstacle position fails to match with the historical obstacle position, it means that the obstacle corresponding to the obstacle position and the historical obstacle corresponding to the historical obstacle position are not the same obstacle. At this time, the obstacle position can be stored in the historical position set as a new historical obstacle position in association with the corresponding obstacle type, so as to facilitate subsequent comparison and matching.

[0073] In one embodiment, the step S40, i.e., matching the obstacle position with the historical obstacle positions in the historical position set, includes:

[0074] When the type detection results corresponding to the obstacle position and the historical obstacle position are both first detection results, the first center point of the obstacle position and the second center point of the historical obstacle position are obtained; since the type detection results corresponding to the obstacle position and the historical obstacle position are both first detection results, the historical obstacle and the obstacle detected in the obstacle image at the current moment are both target vehicles containing target wheels. Therefore, according to the above embodiment, the first center point of the obstacle position refers to the center point of the first rectangular box corresponding to the obstacle position (that is, the first rectangular box surrounded by the length sideline and the two target vertical lines in the above text), and the second center point of the historical obstacle position refers to the second center point of the first rectangular box corresponding to the historical obstacle position.

[0075] When the distance between the first center point and the second center point is less than or equal to the preset distance threshold, it is determined that the obstacle position matches the historical obstacle position successfully; understandably, the preset distance threshold can be set according to demand, such as 1m. At this time, the smaller the distance between the first center point and the second center point, the closer the obstacle corresponding to the obstacle position is to the historical obstacle corresponding to the historical obstacle position. When the distance between the first center point and the second center point is less than or equal to the preset distance threshold, it means that the obstacle corresponding to the obstacle position is the same obstacle as the historical obstacle corresponding to the historical obstacle position. At this time, the obstacle position matches the historical obstacle position successfully.

[0076] When the distance between the first center point and the second center point is greater than a preset distance threshold, it is determined that the obstacle position fails to match the historical obstacle position. Understandably, the greater the distance between the first center point and the second center point, the farther the obstacle corresponding to the obstacle position is from the historical obstacle corresponding to the historical obstacle position. When the distance between the first center point and the second center point is greater than a preset distance threshold, it means that the obstacle corresponding to the obstacle position is not the same obstacle as the historical obstacle corresponding to the historical obstacle position, and at this time, the obstacle position fails to match the historical obstacle position.

[0077] In one embodiment, the step S40, i.e., matching the obstacle position with the historical obstacle positions in the historical position set, includes:

[0078] When the type detection results corresponding to the obstacle position and the historical obstacle position are not all the first detection results, a target comparison area is determined according to the obstacle position, and a historical comparison area corresponding to the historical obstacle position is obtained; wherein, the type detection results corresponding to the obstacle position and the historical obstacle position are not all the first detection results, indicating that at least one of the obstacle corresponding to the obstacle position and the historical obstacles corresponding to the historical obstacle position is a target vehicle including a target wheel (that is, the target vehicle including the target wheel corresponds to the first detection result), for example, when the obstacle corresponding to the obstacle position and the historical obstacles corresponding to the historical obstacle position are not the target vehicle including the target wheel, the obstacle position and the historical obstacle position are both the target vehicles mentioned in the above embodiment. Fig. 9 When the obstacle corresponding to the obstacle position and one of the historical obstacles corresponding to the historical obstacle position is a target vehicle including a target wheel, then the obstacle position and one of the historical obstacle positions are the target vehicle mentioned in the above embodiment. Fig. 9 The second side line L3 shown in FIG. 1 and the other side line L4 shown in FIG. 1 are the side lines L3 shown in FIG. 1 and L3 ... Figure 7 The target rectangular area surrounded by the target vertical line L2 and the length side line L1 is shown in . Understandably, the target comparison area is a rectangular area determined according to the above-mentioned obstacle position, and it needs to be different according to the first detection result corresponding to the obstacle. The historical comparison area is a target comparison area that has been determined corresponding to a historical time point. Therefore, the historical comparison area can be pre-stored in the storage area corresponding to the historical obstacle position, and can be retrieved at any time when needed in this step. Of course, the historical comparison area can also be generated in real time with reference to the generation method of the target comparison area, which is not limited here.

[0079] Obtain an intersection-and-union ratio between the target comparison area and the historical comparison area; wherein the intersection-and-union ratio refers to the ratio of the intersection area divided by the union area, wherein the intersection area refers to the area of ​​intersection between the target comparison area corresponding to the obstacle position and the historical comparison area corresponding to the historical obstacle position; and the union area refers to the area after the target comparison area corresponding to the obstacle position and the historical comparison area corresponding to the historical obstacle position are merged.

[0080] When the intersection-and-union ratio is greater than a preset intersection-and-union threshold, it is determined that the obstacle position matches the historical obstacle position successfully; specifically, the preset intersection-and-union threshold can be set according to demand, for example, set to 0.2; the larger the intersection-and-union ratio, the closer the obstacle corresponding to the obstacle position is to the historical obstacle corresponding to the historical obstacle position; when the intersection-and-union ratio is greater than the preset intersection-and-union threshold, it means that the obstacle corresponding to the obstacle position is the same obstacle as the historical obstacle corresponding to the historical obstacle position, and at this time the obstacle position matches the historical obstacle position successfully.

[0081] When the intersection-and-union ratio is less than or equal to the preset intersection-and-union threshold, it is determined that the obstacle position fails to match the historical obstacle position. Understandably, the smaller the intersection-and-union ratio, the farther the obstacle corresponding to the obstacle position is from the historical obstacle corresponding to the historical obstacle position. When the intersection-and-union ratio is less than or equal to the preset intersection-and-union threshold, it means that the obstacle corresponding to the obstacle position is not the same obstacle as the historical obstacle corresponding to the historical obstacle position, and at this time, the obstacle position fails to match the historical obstacle position.

[0082] In one embodiment, determining the target comparison area according to the obstacle position includes:

[0083] When the obstacle position is the first detection result, the obstacle position is determined as the target comparison area. That is, in this embodiment, since the obstacle position corresponding to the first detection result is Figure 7 The target rectangular area surrounded by the target vertical line L2 and the length sideline L1 shown in , therefore, the target rectangular area can be directly determined as the target comparison area. At this time, the type detection result corresponding to the historical obstacle position is not the first detection result, therefore, the historical comparison area corresponding to the historical obstacle position refers to a rectangular area generated according to the historical obstacle position corresponding to the historical obstacle (that is, the second sideline L3 corresponding to the historical obstacle) (can be generated by referring to the target comparison area corresponding to the obstacle position that is not the first detection result in the next embodiment).

[0084] In one embodiment, the current obstacle information includes obstacle detection data; the obstacle detection data includes an obstacle detection frame for selecting obstacles corresponding to different obstacle types; Fig. 9As shown, the obstacle position refers to the second connecting line L3 between the third positioning point C2 and the fourth positioning point D2 in the driving scene top view after the third positioning point C2 and the fourth positioning point D2 are marked in the driving scene top view; the driving scene top view is based on Figure 8 The obstacle image shown in is generated; the third positioning point C2 and the fourth positioning point D2 refer to the obstacle detection frame (ie Figure 8 The second detection frame 12) in the two end points of the bottom edge;

[0085] Further, determining the target comparison area according to the obstacle position includes:

[0086] When the obstacle position is not the first detection result, the target comparison area is determined by taking the second line L3 as the length center line; the length of the target comparison area is equal to the length of the second line L3, and the ratio between the width of the comparison obstacle area and the length of the second line L3 is a preset ratio. Specifically, the preset ratio can be set according to demand, such as being set to one-half. Among them, the second line L3 is the center line of the target comparison area in the length direction, that is, the second line L3 is parallel to the long side of the target comparison area, and is located at the center position between the two long sides of the target comparison area. The two long sides of the target comparison area are symmetrically arranged with the second line L3 as the symmetry axis. At the same time, the lengths of the two long sides of the target comparison area are equal to the second line L3, and the ratio between the lengths of the two short sides of the target comparison area (that is, the width of the target comparison area) and the length of the second line L3 is a preset ratio. In this way, the target comparison area can be regarded as Figure 8 The obstacle T2 in Fig. 9 The approximate area occupied by the driving scene in the overhead view.

[0087] In one embodiment, if Figure 5 As shown, in the step S50, the operation of updating the successfully matched historical obstacle position according to the obstacle position includes:

[0088] S501, dividing the driving scene top view into several areas along the length direction of the driving vehicle; the driving scene top view is generated according to the current obstacle information; the obstacle position and the successfully matched historical obstacle position are marked in the driving scene top view; that is, in this embodiment, the driving scene top view can be equally divided into an odd number of areas (such as 7 areas) to facilitate determining its central area and other areas on both sides of the central area; wherein the central area refers to the most central area among all the areas. It can be understood that the length direction of the driving vehicle is the up and down direction in the driving scene top view, the upper end in the driving scene top view corresponds to the front end of the driving vehicle, and the lower end in the driving scene top view corresponds to the rear end of the driving vehicle. In this embodiment, since the obstacle image is captured by a fisheye camera, the obstacle image captured by the fisheye camera has the characteristics of small distortion at the center and large distortion at the edges on both sides. Therefore, in the top view of the driving scene converted from the obstacle image, the closer to the upper and lower ends, the more missing position-related data there are. Therefore, missing data may make the detection position inaccurate; therefore, in the top view of the driving scene, the closer to the center area, the better the imaging quality, and the more accurate the position-related data will be. Since the obstacle position and the successfully matched historical obstacle position correspond to the same obstacle, it is possible to further determine how to update the historical obstacle position by determining the area to which the obstacle position and the successfully matched historical obstacle position belong.

[0089] S502, when it is determined that the obstacle position and the successfully matched historical obstacle position belong to the same area, the successfully matched historical obstacle position is replaced with the obstacle position in the historical position set. Understandably, since the obstacle position and the successfully matched historical obstacle position correspond to the same obstacle, if the obstacle position and the successfully matched historical obstacle position belong to the same area, it means that the obstacle position and the successfully matched historical obstacle position have similar data missing degrees. At this time, since the obstacle position is the latest obstacle position measured at the current moment, the successfully matched historical obstacle position in the historical position set can be directly replaced with the obstacle position. In addition, the number of detections of the historical obstacle (that is, the obstacle that the obstacle position and the historical obstacle position both correspond to) can be increased by one.

[0090] In one embodiment, after step S501, dividing the top view of the driving scene into a plurality of regions includes:

[0091] When it is determined that the obstacle position and the successfully matched historical obstacle position belong to different areas, the first distance between the area to which the obstacle position belongs and the central area, and the second distance between the area to which the historical obstacle position belongs and the central area are determined; the central area refers to the area located at the center of all the areas; it can be understood that since the obstacle position and the successfully matched historical obstacle position correspond to the same obstacle, if the obstacle position and the successfully matched historical obstacle position do not belong to the same area, it means that the degree of data missing of the obstacle position and the successfully matched historical obstacle position may be different. At this time, it is necessary to take the position with a relatively small degree of data missing as the current position corresponding to the obstacle. Specifically, first, the first distance between the area to which the obstacle position belongs and the central area, and the second distance between the area to which the historical obstacle position belongs and the central area are determined. For example, if there are 7 areas in total, if the area to which the historical obstacle position belongs is 3 and the central area is 4, the second distance is 4-3=1; if the area to which the obstacle position belongs is 6 and the central area is 4, the second distance is 6-4=2.

[0092] When the first distance is less than or equal to the second distance, the successfully matched historical obstacle position is replaced with the obstacle position in the historical position set; understandably, when the first distance is less than the second distance, it means that the historical obstacle position is farther from the central area, so the historical obstacle position data is more missing. Therefore, the obstacle position with a relatively small degree of data missing can be replaced with the historical obstacle position, and the obstacle position can be used as the new position data of the above-mentioned historical obstacle. When the first distance is equal to the second distance, it means that the distance between the historical obstacle position and the obstacle position from the central area is equal. At this time, it means that the obstacle position and the successfully matched historical obstacle position have similar data missing degrees. At this time, since the obstacle position is the latest obstacle position measured at the current moment, the successfully matched historical obstacle position in the historical position set can be directly replaced with the obstacle position. In addition, the number of detections of the historical obstacle (that is, the obstacle position and the historical obstacle position corresponding to each other) can be increased by one.

[0093] When the first distance is greater than the second distance, the successfully matched historical obstacle position is kept unchanged in the historical position set. Understandably, when the first distance is greater than the second distance, it means that the historical obstacle position is farther from the central area, so the historical obstacle position data is relatively less missing and the imaging quality is better. Therefore, the historical obstacle position is the position obtained after conversion according to the current posture information. Therefore, the historical obstacle position can be used as the latest position data of the historical obstacle.

[0094] It is understandable that in the present invention, multiple obstacles may be photographed at the same time during the driving process of the vehicle. However, the position of each obstacle is determined by the above-mentioned obstacle data processing method, so that the arrangement of all obstacles around the driving vehicle can be determined, thereby providing a position reference for the driving navigation and automatic parking of the driving vehicle.

[0095] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0096] In one embodiment, a controller is provided. The controller may be a server, and its internal structure diagram may be as follows: Fig.10 As shown. The controller includes a processor, a memory, a network interface and a database connected via a system bus. The processor of the controller is used to provide computing and control capabilities. The memory of the controller includes a readable storage medium and an internal memory. The readable storage medium stores an operating system, computer-readable instructions and a database. The internal memory provides an environment for the operation of the operating system and computer-readable instructions in the readable storage medium. When the computer-readable instructions are executed by the processor, an obstacle data processing method is implemented. The readable storage medium provided in this embodiment includes a non-volatile readable storage medium and a volatile readable storage medium.

[0097] In one embodiment, a controller is provided, comprising a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned obstacle data processing method when executing the computer-readable instructions.

[0098] The present invention further provides a vehicle, comprising the above controller. Specific definitions of the vehicle can be found in the above definitions of the obstacle data processing method and the controller, which will not be repeated here.

[0099] In the above-mentioned vehicle of the present invention, the controller first determines the type detection result according to the obstacle image of the obstacle, and then, when the type detection result is determined to be a first detection result for characterizing that the obstacle is a target vehicle including a target wheel, the obstacle position is determined using the wheel position information corresponding to the two wheels located on the same side of the target vehicle. In this way, the obstacle position corresponding to the target vehicle (i.e., the obstacle) is located using the wheel position information, which will reduce the error caused by image distortion, thereby improving the positioning accuracy of the obstacle position.

[0100] In one embodiment, a computer-readable storage medium is provided, on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, the steps of the above obstacle data processing method are implemented.

[0101] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a readable storage medium, and the readable storage medium includes a non-volatile readable storage medium and a volatile readable storage medium. When the computer-readable instructions are executed, they may include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0102] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units or modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional units or modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0103] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. An obstacle data processing method, characterized in that: include: Determine a type detection result based on an obstacle image of the obstacle; When it is determined that the type detection result is a first detection result, determining the obstacle position of the obstacle according to the wheel position information corresponding to the target wheel; the first detection result indicates that the obstacle image contains a target vehicle having a target wheel; The target wheels include two wheels located on the same side of the target vehicle; The method of determining the obstacle position of the obstacle according to the wheel position information corresponding to the target wheel includes: determining a target rectangular area surrounded by two target vertical lines and a length sideline as the obstacle position of the obstacle; the length sideline refers to a line segment obtained by extending a first connecting line between a first positioning point and a second positioning point in a driving scene top view toward both ends by a preset length; the two target vertical lines refer to two vertical lines of a preset height extending in a direction away from a center point of the driving scene top view with two end points of the length sideline as vertical feet respectively; the target rectangular area represents an area corresponding to the obstacle position occupied by the target vehicle in the driving scene top view; the driving scene top view is generated according to an obstacle image collected in real time by at least one camera installed on the driving vehicle; the first positioning point and the second positioning point refer to the bottom midpoints of two first detection frames converted to corresponding coordinate points in the driving scene top view; and the two first detection frames refer to two wheels in the obstacle image used to frame the same side of the target vehicle.

2. The obstacle data processing method according to claim 1, characterized in that: The determining of the type detection result according to the obstacle image of the obstacle includes: Collecting obstacle images in real time through at least one camera installed on the driving vehicle, and determining current obstacle information based on all the obstacle images collected at the current moment; The type detection result is determined according to the current obstacle information.

3. The obstacle data processing method according to claim 2, characterized in that: The determining the current obstacle information according to all the obstacle images collected at the current moment includes: All the obstacle images collected at the current moment are input into a preset obstacle recognition model, and current obstacle information output by the preset obstacle recognition model is obtained.

4. The obstacle data processing method according to claim 2, characterized in that: The current obstacle information includes at least one obstacle type and obstacle detection data associated with each obstacle type; The determining the type detection result according to the current obstacle information includes: When it is determined that the obstacle type includes a target vehicle type and a wheel type, and when it is determined based on the obstacle detection data that the wheel type and the target vehicle type meet a preset association condition, the type detection result is determined to be the first detection result; the target vehicle type refers to the vehicle type corresponding to the target vehicle.

5. The obstacle data processing method according to claim 4, characterized in that: The determining, according to the obstacle detection data, that the wheel type and the target vehicle type meet a preset association condition comprises: Acquire vehicle position information and wheel position information from the obstacle detection data, wherein the vehicle position information is associated with the target vehicle model, and the wheel position information is associated with the wheel type; When it is determined that the vehicle position information includes the wheel position information corresponding to the target wheel, it is determined that the wheel type and the target vehicle type meet a preset association condition.

6. The obstacle data processing method according to claim 5, characterized in that: The wheel position information includes a first detection frame for selecting a wheel corresponding to the wheel type; the vehicle position information includes a second detection frame for selecting the target vehicle; The determining that the vehicle position information includes the wheel position information corresponding to the target wheel includes: When the second detection frame includes two of the first detection frames, it is determined that the vehicle position information includes the wheel position information corresponding to the target wheel.

7. The obstacle data processing method according to claim 2, characterized in that: The determining the obstacle position of the obstacle according to the wheel position information corresponding to the target wheel includes: Wheel position information corresponding to the target wheel is determined from the current obstacle information, and the obstacle position of the target vehicle is determined based on the wheel position information, wherein the wheel position information includes two first detection frames for selecting two wheels located on the same side of the target vehicle.

8. The obstacle data processing method according to claim 7, characterized in that: The step of determining the obstacle position of the target vehicle according to the wheel position information comprises: Marking a first positioning point and a second positioning point in a top view of the driving scene; the top view of the driving scene is generated according to the obstacle image; the first positioning point and the second positioning point refer to coordinate points corresponding to the midpoints of the bottom edges of the two first detection frames converted to the top view of the driving scene; The obstacle position of the obstacle in the driving scene overhead view is determined according to the first positioning point and the second positioning point in the driving scene overhead view.

9. The obstacle data processing method according to claim 2, characterized in that: After determining the type detection result according to the obstacle image of the obstacle, the method further includes: When it is determined that the type detection result is the second detection result or the third detection result, the obstacle position of the obstacle is determined according to the current obstacle information; the second detection result indicates that the obstacle image does not contain the target vehicle; the third detection result indicates that the obstacle image does not contain the target vehicle with the target wheel.

10. The obstacle data processing method according to claim 9, characterized in that: The current obstacle information includes at least one obstacle type; The determining the type detection result according to the current obstacle information includes: When it is determined that the obstacle type does not include the target vehicle type, determining the type detection result as a second detection result; the target vehicle type refers to the vehicle type corresponding to the target vehicle; When it is determined that the obstacle type includes the target vehicle type but does not include the wheel type, the type detection result is determined to be a third detection result.

11. The obstacle data processing method according to claim 9, characterized in that: The current obstacle information includes at least one obstacle type and obstacle detection data associated with each obstacle type; The determining the type detection result according to the current obstacle information includes: When it is determined that the obstacle type includes a wheel type and a target vehicle type, and when it is determined based on the obstacle detection data that the wheel type and the target vehicle type do not meet a preset association condition, the type detection result is determined to be a third detection result; the target vehicle type refers to a vehicle type corresponding to the target vehicle.

12. The obstacle data processing method according to claim 11, characterized in that: The determining, according to the obstacle detection data, that the wheel type and the target vehicle type do not meet a preset association condition includes: Acquire vehicle position information and wheel position information, wherein the vehicle position information refers to obstacle detection data associated with the target vehicle type in the current obstacle information; and the wheel position information refers to obstacle detection data associated with the wheel type in the current obstacle information; When it is determined that the vehicle position information does not include the wheel position information corresponding to the target wheel, it is determined that the wheel type and the target vehicle type do not meet a preset association condition.

13. The obstacle data processing method according to claim 12, characterized in that: The wheel position information includes a first detection frame for selecting a wheel corresponding to the wheel type; the vehicle position information includes a second detection frame for selecting the target vehicle; The determining that the vehicle position information does not include the wheel position information corresponding to the target wheel includes: When the second detection frame does not include the first detection frame, or includes only one first detection frame, it is determined that the vehicle position information does not include the wheel position information corresponding to the target wheel.

14. The obstacle data processing method according to claim 9, characterized in that: The current obstacle information includes obstacle detection data; The obstacle detection data includes an obstacle detection frame for selecting obstacles corresponding to different obstacle types; The determining the obstacle position of the obstacle according to the current obstacle information includes: An obstacle position of the obstacle is determined according to the obstacle detection frame.

15. The obstacle data processing method according to claim 14, characterized in that: The determining the obstacle position of the obstacle according to the obstacle detection frame includes: Marking a third positioning point and a fourth positioning point in a top view of the driving scene; the top view of the driving scene is generated according to the obstacle image; the third positioning point and the fourth positioning point refer to two end points of the bottom edge of the obstacle detection frame converted to corresponding coordinate points in the top view of the driving scene; A second connecting line between the third positioning point and the fourth positioning point in the top view of the driving scene is determined as the obstacle position of the obstacle.

16. The obstacle data processing method according to any one of claims 2 to 15, characterized in that: After determining the obstacle position of the obstacle, the method further comprises: Matching the obstacle position with the historical obstacle position in the historical position set; the historical obstacle position is obtained by converting the historical position of the historical obstacle according to the current vehicle body posture; the historical position refers to the obstacle position corresponding to the historical obstacle determined in a preset historical period; When the obstacle position matches the historical obstacle position successfully, updating the successfully matched historical obstacle position according to the obstacle position; When the obstacle position fails to match with all the historical obstacle positions, the obstacle position is added to the historical position set as a new historical obstacle position.

17. The obstacle data processing method according to claim 16, characterized in that: The matching of the obstacle position with the historical obstacle positions in the historical position set includes: When the type detection results corresponding to the obstacle position and the historical obstacle position are both first detection results, obtaining a first center point of the obstacle position and a second center point of the historical obstacle position; When the distance between the first center point and the second center point is less than or equal to a preset distance threshold, determining that the obstacle position matches the historical obstacle position successfully; When the distance between the first center point and the second center point is greater than a preset distance threshold, it is determined that the obstacle position fails to match the historical obstacle position.

18. The obstacle data processing method according to claim 16, characterized in that: The matching of the obstacle position with the historical obstacle positions in the historical position set includes: When the type detection results corresponding to the obstacle position and the historical obstacle position are not all the first detection results, determining a target comparison area according to the obstacle position, and acquiring a historical comparison area corresponding to the historical obstacle position; Obtaining an intersection-and-union ratio between the target comparison area and the historical comparison area; When the intersection-over-combination ratio is greater than a preset intersection-over-combination threshold, it is determined that the obstacle position matches the historical obstacle position successfully; When the intersection-over-combination ratio is less than or equal to a preset intersection-over-combination threshold, it is determined that the obstacle position fails to match the historical obstacle position.

19. The obstacle data processing method according to claim 18, characterized in that: The step of determining the target comparison area according to the obstacle position includes: When the obstacle position is the first detection result, the obstacle position is determined as the target comparison area.

20. The obstacle data processing method according to claim 18, characterized in that: The current obstacle information includes obstacle detection data; the obstacle detection data includes an obstacle detection frame for selecting obstacles corresponding to different obstacle types; the obstacle position refers to a second connection line between the third positioning point and the fourth positioning point in the driving scene top view after the third positioning point and the fourth positioning point are marked in the driving scene top view; the driving scene top view is generated according to the obstacle image; the third positioning point and the fourth positioning point refer to two end points of the bottom edge of the obstacle detection frame converted to corresponding coordinate points in the driving scene top view; The determining of the target comparison area according to the obstacle position includes: When the obstacle position is not the first detection result, the second connecting line is used as the length center line to determine the target comparison area; The length of the target comparison area is equal to the length of the second connection line, and the ratio between the width of the target comparison area and the length of the second connection line is a preset ratio.

21. The obstacle data processing method according to claim 16, characterized in that: The updating operation of the successfully matched historical obstacle position according to the obstacle position includes: The driving scene top view is divided into a plurality of regions along the length direction of the driving vehicle; the driving scene top view is generated according to the current obstacle information; the obstacle position and the successfully matched historical obstacle position are marked in the driving scene top view; When it is determined that the obstacle position and the successfully matched historical obstacle position belong to the same area, the successfully matched historical obstacle position is replaced with the obstacle position in the historical position set.

22. The obstacle data processing method according to claim 21, characterized in that: After dividing the top view of the driving scene into a plurality of areas, the following steps are included: When it is determined that the obstacle position and the successfully matched historical obstacle position belong to different areas, determining a first distance between the area to which the obstacle position belongs and a central area, and a second distance between the area to which the historical obstacle position belongs and the central area; the central area refers to an area located most centrally among all the areas; When the first distance is less than or equal to the second distance, replacing the successfully matched historical obstacle position with the obstacle position in the historical position set; When the first distance is greater than the second distance, the historical obstacle positions that have been successfully matched are kept unchanged in the historical position set.

23. A controller, characterized in that: The method comprises a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, wherein the processor implements the obstacle data processing method according to any one of claims 1 to 22 when executing the computer-readable instructions.

24. A vehicle, characterized in that: Comprising a controller as claimed in claim 23.

25. A computer-readable storage medium storing computer-readable instructions, characterized in that: When the computer-readable instructions are executed by a processor, the obstacle data processing method according to any one of claims 1 to 22 is implemented.

Images