Static vehicle identification method and device, storage medium and computer equipment

Abstract

The invention discloses a static vehicle identification method and device, a storage medium and computer equipment, and the method comprises the steps: obtaining multiple frames of vehicle visual data corresponding to a current vehicle, and determining respective static vehicle prediction data of each frame of vehicle visual data, the static vehicle prediction data comprising a visual prediction position of at least one static vehicle; for each frame of static vehicle prediction data, based on the visual prediction position and a plurality of preset areas corresponding to the current vehicle, determining an area where each static vehicle is located relative to the current vehicle, and determining a prediction weight of each static vehicle according to a preset area weight corresponding to the area; and based on the corresponding visual prediction position and the prediction weight of each static vehicle in each frame of static vehicle prediction data, performing fusion prediction on the visual prediction position of each static vehicle to obtain a visual fusion position of each static vehicle. According to the invention, the accuracy of static vehicle identification can be improved.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and in particular to a static vehicle identification method and device, storage medium, and computer equipment. Background Art

[0002] Autonomous driving environment perception mainly relies on sensors such as cameras and lidar to obtain data, and then uses algorithms to process, extract, and model the sensor data to obtain information about the vehicle's surrounding environment. Cameras are still an indispensable sensor in the current L3 / L4 autonomous driving system.

[0003] In parking scenarios, autonomous vehicles use perception algorithms to detect surrounding parked vehicles and obtain available parking spaces. Due to small spaces, many blind spots, and timing jitter, the static vehicle recognition solution in parking scenarios has the problem of low detection accuracy. How to improve the recognition accuracy of static vehicles in the surrounding environment has become one of the core issues in autonomous parking scenarios. Summary of the invention

[0004] In view of this, the present application provides a static vehicle recognition method and device, storage medium, and computer equipment to predict multiple frames of static vehicle prediction data, and use the prior condition that the static vehicle has been predicted to be a static vehicle to fuse the visual prediction positions of the static vehicle in each frame, thereby suppressing jitter. At the same time, different weights are designed for different areas around the vehicle, which helps to improve the accuracy of visual prediction position fusion.

[0005] According to one aspect of the present application, a method for identifying a static vehicle is provided, the method comprising:

[0006] Acquire multiple frames of vehicle visual data corresponding to the current vehicle, and determine static vehicle prediction data for each frame of vehicle visual data, wherein the static vehicle prediction data includes a visual prediction position of at least one static vehicle;

[0007] For each frame of static vehicle prediction data, based on the visual prediction position and a plurality of preset areas corresponding to the current vehicle, determine the area where each static vehicle is located relative to the current vehicle, and determine the prediction weight of each static vehicle according to the preset area weight corresponding to the area;

[0008] Based on the visual prediction position and prediction weight corresponding to each static vehicle in each frame of static vehicle prediction data, the visual prediction position of each static vehicle is fused and predicted to obtain the visual fusion position of each static vehicle.

[0009] Optionally, after performing fusion prediction on the visual predicted positions of each static vehicle to obtain the visual fusion positions of each static vehicle, the method further includes:

[0010] Get the vehicle radar data of the current vehicle;

[0011] Determine the radar data fusion area of ​​the static vehicle based on the visual fusion position of the static vehicle, and obtain target radar data falling into the radar data fusion area from the vehicle radar data;

[0012] The visual fusion position of the static vehicle is adjusted according to the target radar data to obtain the target predicted position of the static vehicle.

[0013] Optionally, determining a radar data fusion area of ​​the static vehicle based on the visual predicted position of the static vehicle, and acquiring target radar data falling within the radar data fusion area from the vehicle radar data, includes:

[0014] Based on the visual predicted position of the static vehicle, multiple predicted edges of the static vehicle are determined, and for any predicted edge, a frame including at least part of the predicted edge is selected as a radar data fusion region corresponding to the predicted edge;

[0015] Target radar data falling into each of the radar data fusion areas is respectively acquired from the vehicle radar data.

[0016] Optionally, adjusting the visual fusion position of the static vehicle according to the target radar data to obtain the target predicted position of the static vehicle includes:

[0017] For the target radar data corresponding to any of the radar data fusion regions, fitting the target radar data to obtain a fitting edge corresponding to the radar data fusion region;

[0018] The visual fusion position of the static vehicle is adjusted based on the fitting edge corresponding to the radar data fusion area to obtain the target predicted position of the static vehicle.

[0019] Optionally, for any predicted edge, selecting a frame including at least a portion of the predicted edge as a radar data fusion region corresponding to the predicted edge includes:

[0020] For any predicted edge, based on the length of the predicted edge and a preset length ratio, determine the frame selection length of the predicted edge;

[0021] Taking the middle point of the predicted edge as the center, taking the portion of the frame selection length as the frame selection position of the predicted edge, and selecting the frame containing the frame selection position as the radar data fusion area corresponding to the predicted edge.

[0022] Optionally, the adjusting the visual fusion position of the static vehicle based on the fitting edge corresponding to the radar data fusion area to obtain the target predicted position of the static vehicle includes:

[0023] Determining a fusion edge corresponding to the radar data fusion area based on a fitting edge and a predicted edge corresponding to the radar data fusion area;

[0024] A radar data fusion border of the static vehicle is generated based on the multiple fusion edges of the static vehicle, and the static vehicle is moved from the visual fusion position to the radar data fusion border to obtain a target predicted position of the static vehicle.

[0025] Optionally, the preset area weight is set based on a perceptual evaluation error of the preset area, and the preset area weight corresponding to the preset area is negatively correlated with the perceptual evaluation error.

[0026] Optionally, after determining the static vehicle prediction data for each frame of vehicle visual data, the method further includes:

[0027] Based on the static vehicle recognition range corresponding to the current vehicle, deleting the static vehicle whose visual predicted position is outside the static vehicle recognition range;

[0028] The posture information of the current vehicle is obtained, and based on the posture information, the visual predicted position of the static vehicle is converted into a world coordinate system.

[0029] Optionally, the static vehicle prediction data further includes a vehicle existence probability of the static vehicle;

[0030] The method of performing fusion prediction on the visual prediction positions of the static vehicles based on the corresponding visual prediction positions and prediction weights of the static vehicles in each frame of static vehicle prediction data to obtain the visual fusion positions of the static vehicles includes:

[0031] Based on the vehicle existence probability corresponding to each static vehicle in each frame of static vehicle prediction data and the prediction probability weight, the vehicle existence probability of each static vehicle is fused and predicted, and the static vehicle is filtered according to the obtained vehicle existence probability fusion prediction result of each static vehicle;

[0032] Based on the visual prediction positions and prediction weights corresponding to the remaining static vehicles in each frame of static vehicle prediction data, a fusion prediction is performed on the visual prediction position of the static vehicle to obtain the visual fusion position of the static vehicle.

[0033] According to another aspect of the present application, a static vehicle identification device is provided, the device comprising:

[0034] A visual perception module, used to obtain multiple frames of vehicle visual data corresponding to the current vehicle, and determine static vehicle prediction data for each frame of vehicle visual data, wherein the static vehicle prediction data includes a visual prediction position of at least one static vehicle;

[0035] A region determination module, for determining, for each frame of static vehicle prediction data, a region where each static vehicle is located relative to the current vehicle based on the visual prediction position and a plurality of preset regions corresponding to the current vehicle, and determining a prediction weight for each static vehicle according to a preset region weight corresponding to the region;

[0036] The position fusion module is used to perform fusion prediction on the visual prediction positions of each static vehicle based on the corresponding visual prediction positions and prediction weights of each static vehicle in each frame of static vehicle prediction data to obtain the visual fusion positions of each static vehicle.

[0037] Optionally, the device further comprises:

[0038] A radar data acquisition module is used to acquire vehicle radar data of the current vehicle; based on the visual fusion position of the static vehicle, determine the radar data fusion area of ​​the static vehicle, and acquire target radar data falling into the radar data fusion area from the vehicle radar data;

[0039] The position fusion module is further used to adjust the visual fusion position of the static vehicle according to the target radar data to obtain the target predicted position of the static vehicle.

[0040] Optionally, the radar data acquisition module is further used to:

[0041] Based on the visual predicted position of the static vehicle, multiple predicted edges of the static vehicle are determined, and for any predicted edge, a frame including at least part of the predicted edge is selected as a radar data fusion region corresponding to the predicted edge;

[0042] Target radar data falling into each of the radar data fusion areas is respectively acquired from the vehicle radar data.

[0043] Optionally, the position fusion module is further used to:

[0044] For the target radar data corresponding to any of the radar data fusion regions, fitting the target radar data to obtain a fitting edge corresponding to the radar data fusion region;

[0045] The visual fusion position of the static vehicle is adjusted based on the fitting edge corresponding to the radar data fusion area to obtain the target predicted position of the static vehicle.

[0046] Optionally, the radar data acquisition module is further used to:

[0047] For any predicted edge, based on the length of the predicted edge and a preset length ratio, determine the frame selection length of the predicted edge;

[0048] Taking the middle point of the predicted edge as the center, taking the portion of the frame selection length as the frame selection position of the predicted edge, and selecting the frame containing the frame selection position as the radar data fusion area corresponding to the predicted edge.

[0049] Optionally, the position fusion module is further used to:

[0050] Determining a fusion edge corresponding to the radar data fusion area based on a fitting edge and a predicted edge corresponding to the radar data fusion area;

[0051] A radar data fusion border of the static vehicle is generated based on the multiple fusion edges of the static vehicle, and the static vehicle is moved from the visual fusion position to the radar data fusion border to obtain a target predicted position of the static vehicle.

[0052] Optionally, the preset area weight is set based on a perceptual evaluation error of the preset area, and the preset area weight corresponding to the preset area is negatively correlated with the perceptual evaluation error.

[0053] Optionally, the device further comprises: a data processing module, configured to:

[0054] Based on the static vehicle recognition range corresponding to the current vehicle, deleting the static vehicle whose visual predicted position is outside the static vehicle recognition range;

[0055] The posture information of the current vehicle is obtained, and based on the posture information, the visual predicted position of the static vehicle is converted into a world coordinate system.

[0056] Optionally, the static vehicle prediction data further includes a vehicle existence probability of the static vehicle;

[0057] The position fusion module is further used for:

[0058] Based on the vehicle existence probability corresponding to each static vehicle in each frame of static vehicle prediction data and the prediction probability weight, the vehicle existence probability of each static vehicle is fused and predicted, and the static vehicle is filtered according to the obtained vehicle existence probability fusion prediction result of each static vehicle;

[0059] Based on the visual prediction positions and prediction weights corresponding to the remaining static vehicles in each frame of static vehicle prediction data, a fusion prediction is performed on the visual prediction position of the static vehicle to obtain the visual fusion position of the static vehicle.

[0060] According to another aspect of the present application, a storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned static vehicle identification method is implemented.

[0061] According to another aspect of the present application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor implements the above-mentioned static vehicle identification method when executing the program.

[0062] By means of the above technical solution, the present application provides a static vehicle recognition method and device, storage medium, and computer equipment, which perform visual perception calculations on multiple frames of vehicle visual data to determine the static vehicle prediction data of each frame, and determine the prediction weights of the visual prediction positions of the static vehicles of each frame based on the respective preset area weights of different preset areas of the current vehicle, thereby weightedly fusing the visual prediction positions of the static vehicles of each frame to obtain the visual fusion position of the static vehicle. The present application predicts multiple frames of static vehicle prediction data, and fuses the visual prediction positions of the static vehicles of each frame using the prior condition that the static vehicle has been predicted, which plays a role in suppressing jitter, and at the same time designs different weights for different areas around the vehicle, which helps to improve the accuracy of the fusion of the visual prediction positions.

[0063] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0065] Figure 1 A schematic diagram showing a flow chart of a static vehicle identification method provided in an embodiment of the present application is shown;

[0066] Figure 2 A schematic diagram of dividing a preset area of ​​a vehicle provided in an embodiment of the present application is shown;

[0067] Figure 3 A schematic flow chart of another static vehicle identification method provided in an embodiment of the present application is shown;

[0068] Figure 4 A schematic diagram showing a selection of radar data fusion area for a static vehicle provided in an embodiment of the present application is shown;

[0069] Figure 5 A schematic diagram of the structure of a static vehicle identification device provided in an embodiment of the present application is shown;

[0070] Figure 6A schematic diagram of the device structure of a computer device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0071] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0072] In this embodiment, a static vehicle identification method is provided. Figure 1 As shown, the method includes:

[0073] Step 101, obtaining multiple frames of vehicle visual data corresponding to the current vehicle, and determining static vehicle prediction data for each frame of vehicle visual data, wherein the static vehicle prediction data includes a visual prediction position of at least one static vehicle.

[0074] The embodiment of the present application is used to identify stationary vehicles (i.e., static vehicles) around a vehicle, and is particularly suitable for vehicle parking scenarios. Prediction weights are set for different perception areas near the vehicle, and the jitter of the static vehicle recognition results is removed and the accuracy of the static vehicle recognition results is improved by weighted fusion of multiple frames of static vehicle recognition results. First, using visual perception technology, static vehicle recognition is performed on multiple frames of vehicle visual data collected by sensors such as cameras during the current vehicle driving process, and the static vehicle prediction position corresponding to each frame of vehicle data is determined, that is, the visual prediction position of the static vehicle.

[0075] Step 102, for each frame of static vehicle prediction data, based on the visual prediction position and multiple preset areas corresponding to the current vehicle, determine the area where each static vehicle is located relative to the current vehicle, and determine the prediction weight of each static vehicle according to the preset area weight corresponding to the area.

[0076] Step 103 , based on the visual prediction positions and prediction weights corresponding to the static vehicles in each frame of static vehicle prediction data, the visual prediction positions of the static vehicles are fused and predicted to obtain the visual fusion positions of the static vehicles.

[0077] Furthermore, based on the multiple preset areas around the current vehicle that are pre-divided, determine which of the multiple preset areas the visual prediction position of the static car in each frame falls into, that is, determine the area where the static car is located relative to the current vehicle. According to the preset area weights corresponding to the different preset areas, determine the prediction weight corresponding to each frame of static car prediction data. Therefore, after determining the prediction weight of each static car in each frame of static car prediction data, the visual prediction position of each static car can be weighted fused based on the prediction weight of each static car in different frames, and the weighted fusion result can be used as the visual fusion position of each static car. In this way, the visual prediction positions of static cars in different frames can be fused, the jitter of the static car recognition results can be removed, and the accuracy of the static car recognition results can be improved.

[0078] like Figure 2 As shown, the surrounding area of ​​the current vehicle (self-vehicle) is divided into 4 preset areas, namely A, B, C, and D. Area A, area B, and area C are respectively high-weighted areas, medium-weighted areas, and low-weighted areas, and corresponding preset area weights (for example, 0.8, 0.6, and 0.4, respectively) are set for different weight areas, and area D is an area that does not participate in fusion. For example, the static vehicle prediction data corresponding to the first frame of vehicle visual data includes static car 1, static car 2, etc., where static car 1 falls into area A, and the preset area weight 0.8 corresponding to area A is used as the prediction weight of static car 1 in the first frame. The static vehicle prediction data corresponding to the second frame of vehicle visual data includes static car 1, static car 2, etc., where static car 1 falls into area B, and the preset area weight 0.6 corresponding to area B is used as the prediction weight of static car 1 in the second frame. The static vehicle prediction data corresponding to the third frame of vehicle visual data includes static vehicle 1, static vehicle 2, etc., where static vehicle 1 falls into region D, which is a non-fusion region. Therefore, when the visual prediction position fusion is performed on static vehicle 1, the visual prediction position of static vehicle 1 in the third frame is not selected. When the visual prediction position fusion is performed on static vehicle 1, the prediction weights of static vehicle 1 in different frames are normalized, and the normalized prediction weights of static vehicle 1 in each frame are used to perform weighted fusion on the visual prediction positions of static vehicle 1 in each frame, and the fusion result is used as the visual fusion position of static vehicle 1.

[0079] In an embodiment of the present application, optionally, the preset area weight is set based on a perceptual evaluation error of the preset area, and the preset area weight corresponding to the preset area is negatively correlated with the perceptual evaluation error.

[0080] In this embodiment, the preset area weights corresponding to different preset areas can be set based on manual experience, or the error evaluation results of different areas of the vehicle perception evaluation can be used to design the weights. The areas with small errors have high weights, and the areas with large errors have low weights (or are not used), thereby improving the overall accuracy.

[0081] By applying the technical solution of this embodiment, visual perception calculations are performed on multiple frames of vehicle visual data to determine the static vehicle prediction data of each frame, and based on the preset area weights of different preset areas of the current vehicle, the prediction weights of the visual prediction positions of the static vehicles in each frame are determined, thereby weighted fusion of the visual prediction positions of the static vehicles in each frame is performed to obtain the visual fusion position of the static vehicle. The present application predicts multiple frames of static vehicle prediction data, and fuses the visual prediction positions of the static vehicles in each frame using the prior condition that the static vehicle has been predicted, which plays a role in suppressing jitter. At the same time, different weights are designed for different areas around the vehicle, which helps to improve the accuracy of visual prediction position fusion.

[0082] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, another static vehicle identification method is provided, such as Figure 3 As shown, the method includes:

[0083] Step 201, obtaining multiple frames of vehicle visual data corresponding to the current vehicle, and determining static vehicle prediction data for each frame of vehicle visual data, wherein the static vehicle prediction data includes a visual prediction position of at least one static vehicle.

[0084] Step 202: based on the static vehicle recognition range corresponding to the current vehicle, delete the static vehicles whose visually predicted positions are outside the static vehicle recognition range.

[0085] Step 203, obtaining the posture information of the current vehicle, and converting the visual predicted position of the static vehicle into a world coordinate system based on the posture information.

[0086] In the embodiment of the present application, taking the parking scene as an example, generally only other vehicles near the current vehicle will affect the driving of the vehicle. After performing visual perception calculations on the vehicle visual data of each frame and obtaining the static vehicle prediction data of each frame, it is possible to determine whether the predicted static vehicle is within the effective area of ​​the current vehicle (i.e., the static vehicle recognition range), delete the static vehicles that are not within the range, reduce the subsequent calculation amount, and improve the calculation efficiency. In addition, it is also necessary to convert the visual prediction positions of static vehicles in different frames to the same coordinate system to facilitate subsequent fusion calculations. For the convenience of calculation, the visual prediction positions of each frame of the static vehicle can be converted to the world coordinate system based on the posture information of each frame of the current vehicle.

[0087] Step 204, for each frame of static vehicle prediction data, based on the visual prediction position and multiple preset areas corresponding to the current vehicle, determine the area where each static vehicle is located relative to the current vehicle, and determine the prediction weight of each static vehicle based on the preset area weight corresponding to the area.

[0088] Step 205, based on the vehicle existence probability corresponding to each static vehicle in each frame of static vehicle prediction data and the prediction probability weight, the vehicle existence probability of each static vehicle is fused and predicted, and the static vehicles are filtered according to the obtained vehicle existence probability fusion prediction results of each static vehicle.

[0089] Step 206, based on the visual prediction positions and prediction weights corresponding to the remaining static vehicles in each frame of static vehicle prediction data, a fusion prediction is performed on the visual prediction position of the static vehicle to obtain the visual fusion position of the static vehicle.

[0090] In this embodiment, the static vehicle prediction data obtained by visual perception calculation also includes the vehicle existence probability of each static vehicle. Before performing the position fusion calculation, the vehicle existence probability of the static vehicle of each frame can also be fused and calculated to improve the accuracy of static vehicle recognition. Specifically, after determining the prediction weight of the static vehicle of each frame and normalizing the prediction weight, the vehicle existence probability of the static vehicle of each frame can also be weighted and fused according to the normalized result of the prediction weight of the static vehicle of each frame to obtain the fusion prediction result of the vehicle existence probability of each static vehicle. For example, the vehicle existence probability of the static vehicle 1 in the first frame is 80% and the prediction weight is 0.8, and the vehicle existence probability of the static vehicle 1 in the second frame is 70% and the prediction weight is 0.6. Then the fusion prediction result of the vehicle existence probability of the static vehicle 1 in the first and second frames is 80%*[0.8 / (0.8+0.6)]+70%*[0.6 / (0.8+0.6)]=75.7%. Specifically, a vehicle existence probability threshold can be set. If the vehicle existence probability fusion prediction result is greater than the threshold, it is considered that a static vehicle exists. Otherwise, it is considered that a static vehicle does not exist, so as to filter static vehicles. Then, the remaining static vehicles are fused and calculated based on the visual prediction position to determine the visual fusion position of the static vehicle.

[0091] Step 207, obtaining vehicle radar data of the current vehicle; determining the radar data fusion area of ​​the static vehicle based on the visual fusion position of the static vehicle, and obtaining target radar data falling into the radar data fusion area from the vehicle radar data.

[0092] In this embodiment, the visual fusion position of the static vehicle can be further corrected based on the ultrasonic radar data collected by the vehicle. Specifically, the vehicle radar data of the current vehicle can be obtained for discretization calculation, and the visual fusion position of the static vehicle can be used to determine the radar data fusion area close to the static vehicle, so that the vehicle radar data falling into the radar data fusion area is obtained from the vehicle radar data as the target radar data, so as to use this part of the target radar data to correct the visual fusion position of the static vehicle, thereby reducing the interference of irrelevant radar data on the static vehicle position correction.

[0093] In the embodiment of the present application, optionally, step 207 includes:

[0094] Step 207-1, based on the visual predicted position of the static vehicle, determining multiple predicted edges of the static vehicle, and for any predicted edge, selecting a frame including at least part of the predicted edge as a radar data fusion area corresponding to the predicted edge;

[0095] Step 207 - 2 , respectively obtaining target radar data falling into each of the radar data fusion areas from the vehicle radar data.

[0096] In this embodiment, for any static vehicle, the four sides of the static vehicle, i.e., predicted sides, can be determined according to the visual predicted position of the static vehicle, and based on each predicted side, a frame containing at least part of the predicted side can be selected as the radar data fusion area of ​​the predicted side. Specifically, the four sides (i.e., predicted sides) of the static vehicle can be determined according to the visual fusion position of the static vehicle, and part or all of them can be selected for each side, and a frame containing the selected part or all of the sides can be framed out, and the frame selected for each frame can be used as the radar data fusion area corresponding to the side. Thus, the vehicle radar data falling into a certain radar data fusion area is used as the target radar data of the predicted side corresponding to the radar data fusion area, so as to correct the static vehicle position. Since the target radar data finally selected is a radar detection point that is relatively close to the side of the static vehicle, using this part of data to correct the static vehicle position can effectively reduce the influence of irrelevant radar detection points on the static vehicle position correction, so that the final correction result is more accurate.

[0097] In an embodiment of the present application, optionally, step 207-1 includes: for any predicted edge, determining the frame selection length of the predicted edge based on the length of the predicted edge and a preset length ratio; taking the midpoint of the predicted edge as the center, taking part of the frame selection length as the frame selection position of the predicted edge, and selecting the border containing the frame selection position as the radar data fusion area corresponding to the predicted edge.

[0098] In this embodiment, if Figure 4As shown in the figure, the visual fusion position of the static car is represented by a dotted frame. For each predicted edge, a portion in the middle of the predicted edge can be selected to determine the border of this predicted edge, that is, the four solid frames in the figure. Specifically, the frame selection length of the predicted edge can be determined according to the length of the predicted edge and the preset length ratio. For example, if the preset length ratio is 60% and the length of the predicted edge is 5 meters, the length of this predicted edge is 3 meters. Then, with the middle point of the predicted edge as the center, a 3-meter-long portion is selected as the frame selection position of this predicted edge, so that a 3-meter-long frame is used to frame the 3-meter-long portion in the middle of the predicted edge to form the radar data fusion area of ​​this predicted edge. Among them, the width of the frame can be a set value; it can also be a certain proportion of the length of the frame, for example, the width of the frame is 60% of the length of the frame.

[0099] Step 208: adjusting the visual fusion position of the static vehicle according to the target radar data to obtain the target predicted position of the static vehicle.

[0100] In this embodiment, after determining the target radar data corresponding to each predicted edge, the position of each predicted edge can be fitted according to the target radar data, and then the visual fusion position of the static vehicle is adjusted and calibrated using the predicted edge position fitted by the radar data to obtain the final target predicted position of the static vehicle.

[0101] In the embodiment of the present application, optionally, step 208 includes:

[0102] Step 208-1, for the target radar data corresponding to any of the radar data fusion regions, fitting the target radar data to obtain a fitting edge corresponding to the radar data fusion region;

[0103] Step 208-2: adjusting the visual fusion position of the static vehicle based on the fitting edge corresponding to the radar data fusion area to obtain the target predicted position of the static vehicle.

[0104] In this embodiment, for the target radar data corresponding to each predicted edge, that is, the target radar data corresponding to each radar data fusion area, the target radar data corresponding to a predicted edge can be fitted to a straight line, that is, fitted to the fitted edge corresponding to each predicted edge of the static car, the predicted edge is the edge of the static car obtained by visual perception calculation and fusion of visual prediction, and the fitted edge is the edge of the static car selected based on the predicted edge and then fitted using the selected radar data.

[0105] In the embodiment of the present application, the visual fusion position of the static car can be adjusted directly based on the fitted edge, or the visual fusion position of the static car can be adjusted by combining the fitted edge with the predicted edge.

[0106] In an optional manner, step 208-2 may be: generating a radar data fusion border of the static vehicle based on the fitting edge corresponding to the radar data fusion area; moving the static vehicle from the visual fusion position to the radar data fusion border, and determining the target predicted position of the static vehicle.

[0107] In this embodiment, the four fitting edges can be used as an edge of the radar data fusion border of the static vehicle respectively, the four edges can be connected into a radar data fusion border, and the visual fusion position of the static vehicle can be adjusted to the radar data fusion border to correct the visual fusion position of the static vehicle.

[0108] In another optional manner, step 208-2 may be: determining the fusion edge corresponding to the radar data fusion area based on the fitting edge and the predicted edge corresponding to the radar data fusion area; generating the radar data fusion bounding box of the static vehicle based on the multiple fusion edges of the static vehicle, and moving the static vehicle from the visual fusion position to the radar data fusion bounding box to obtain the target predicted position of the static vehicle.

[0109] In this embodiment, each fitted edge and the corresponding predicted edge can be fused to obtain a fused edge. Specifically, the average of the fitted edge and the predicted edge can be taken to obtain the fused edge. The four fused edges are then respectively used as an edge of the radar data fusion border of the static vehicle. The obtained four edges are connected into a radar data fusion border, and the visual fusion position of the static vehicle is adjusted to the radar data fusion border to correct the visual fusion position of the static vehicle.

[0110] Further, as Figure 1 The specific implementation of the method, the embodiment of the present application provides a static vehicle identification device, such as Figure 5 As shown, the device comprises:

[0111] A visual perception module, used to obtain multiple frames of vehicle visual data corresponding to the current vehicle, and determine static vehicle prediction data for each frame of vehicle visual data, wherein the static vehicle prediction data includes a visual prediction position of at least one static vehicle;

[0112] A region determination module, for determining, for each frame of static vehicle prediction data, a region where each static vehicle is located relative to the current vehicle based on the visual prediction position and a plurality of preset regions corresponding to the current vehicle, and determining a prediction weight for each static vehicle according to a preset region weight corresponding to the region;

[0113] The position fusion module is used to perform fusion prediction on the visual prediction positions of each static vehicle based on the corresponding visual prediction positions and prediction weights of each static vehicle in each frame of static vehicle prediction data to obtain the visual fusion positions of each static vehicle.

[0114] Optionally, the device further comprises:

[0115] A radar data acquisition module is used to acquire vehicle radar data of the current vehicle; based on the visual fusion position of the static vehicle, determine the radar data fusion area of ​​the static vehicle, and acquire target radar data falling into the radar data fusion area from the vehicle radar data;

[0116] The position fusion module is further used to adjust the visual fusion position of the static vehicle according to the target radar data to obtain the target predicted position of the static vehicle.

[0117] Optionally, the radar data acquisition module is further used to:

[0118] Based on the visual predicted position of the static vehicle, multiple predicted edges of the static vehicle are determined, and for any predicted edge, a frame including at least part of the predicted edge is selected as a radar data fusion region corresponding to the predicted edge;

[0119] Target radar data falling into each of the radar data fusion areas is respectively acquired from the vehicle radar data.

[0120] Optionally, the position fusion module is further used to:

[0121] For the target radar data corresponding to any of the radar data fusion regions, fitting the target radar data to obtain a fitting edge corresponding to the radar data fusion region;

[0122] The visual fusion position of the static vehicle is adjusted based on the fitting edge corresponding to the radar data fusion area to obtain the target predicted position of the static vehicle.

[0123] Optionally, the radar data acquisition module is further used to:

[0124] For any predicted edge, based on the length of the predicted edge and a preset length ratio, determine the frame selection length of the predicted edge;

[0125] Taking the middle point of the predicted edge as the center, taking the portion of the frame selection length as the frame selection position of the predicted edge, and selecting the frame containing the frame selection position as the radar data fusion area corresponding to the predicted edge.

[0126] Optionally, the position fusion module is further used to:

[0127] Determining a fusion edge corresponding to the radar data fusion area based on a fitting edge and a predicted edge corresponding to the radar data fusion area;

[0128] A radar data fusion border of the static vehicle is generated based on the multiple fusion edges of the static vehicle, and the static vehicle is moved from the visual fusion position to the radar data fusion border to obtain a target predicted position of the static vehicle.

[0129] Optionally, the preset area weight is set based on a perceptual evaluation error of the preset area, and the preset area weight corresponding to the preset area is negatively correlated with the perceptual evaluation error.

[0130] Optionally, the device further comprises: a data processing module, configured to:

[0131] Based on the static vehicle recognition range corresponding to the current vehicle, deleting the static vehicle whose visual predicted position is outside the static vehicle recognition range;

[0132] The posture information of the current vehicle is obtained, and based on the posture information, the visual predicted position of the static vehicle is converted into a world coordinate system.

[0133] Optionally, the static vehicle prediction data further includes a vehicle existence probability of the static vehicle;

[0134] The position fusion module is further used for:

[0135] Based on the vehicle existence probability corresponding to each static vehicle in each frame of static vehicle prediction data and the prediction probability weight, the vehicle existence probability of each static vehicle is fused and predicted, and the static vehicle is filtered according to the obtained vehicle existence probability fusion prediction result of each static vehicle;

[0136] Based on the visual prediction positions and prediction weights corresponding to the remaining static vehicles in each frame of static vehicle prediction data, a fusion prediction is performed on the visual prediction position of the static vehicle to obtain the visual fusion position of the static vehicle.

[0137] It should be noted that for other corresponding descriptions of the functional units involved in the static vehicle identification device provided in the embodiment of the present application, reference can be made to Figures 1 to 4 The corresponding description in the method will not be repeated here.

[0138] The present application also provides a computer device, which may be a personal computer, a server, a network device, etc. Figure 6As shown, the computer device includes a bus, a processor, a memory and a communication interface, and may also include an input and output interface and a display device. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store location information. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the steps in each method embodiment are implemented.

[0139] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0140] In one embodiment, a computer-readable storage medium is provided. The computer-readable storage medium may be non-volatile or volatile, and stores a computer program thereon. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0141] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0142] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0143] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0144] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0145] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for identifying a static vehicle, characterized in that: The method comprises: Acquire multiple frames of vehicle visual data corresponding to the current vehicle, and determine static vehicle prediction data for each frame of vehicle visual data, wherein the static vehicle prediction data includes a visual prediction position of at least one static vehicle; For each frame of static vehicle prediction data, based on the visual prediction position and a plurality of preset areas corresponding to the current vehicle, determine the area where each static vehicle is located relative to the current vehicle, and determine the prediction weight of each static vehicle according to the preset area weight corresponding to the area; Based on the visual prediction position and prediction weight corresponding to each static vehicle in each frame of static vehicle prediction data, the visual prediction position of each static vehicle is fused and predicted to obtain the visual fusion position of each static vehicle.

2. The method according to claim 1, characterized in that After the visual prediction positions of the static vehicles are fused and predicted to obtain the visual fusion positions of the static vehicles, the method further includes: Get the vehicle radar data of the current vehicle; Determine the radar data fusion area of ​​the static vehicle based on the visual fusion position of the static vehicle, and obtain target radar data falling into the radar data fusion area from the vehicle radar data; The visual fusion position of the static vehicle is adjusted according to the target radar data to obtain the target predicted position of the static vehicle.

3. The method according to claim 2, characterized in that The step of determining the radar data fusion area of ​​the static vehicle based on the visual predicted position of the static vehicle, and acquiring target radar data falling within the radar data fusion area from the vehicle radar data, includes: Based on the visual predicted position of the static vehicle, multiple predicted edges of the static vehicle are determined, and for any predicted edge, a frame including at least part of the predicted edge is selected as a radar data fusion region corresponding to the predicted edge; Target radar data falling into each of the radar data fusion areas is respectively acquired from the vehicle radar data.

4. The method according to claim 3, characterized in that The step of adjusting the visual fusion position of the static vehicle according to the target radar data to obtain the target predicted position of the static vehicle includes: For the target radar data corresponding to any of the radar data fusion regions, fitting the target radar data to obtain a fitting edge corresponding to the radar data fusion region; The visual fusion position of the static vehicle is adjusted based on the fitting edge corresponding to the radar data fusion area to obtain the target predicted position of the static vehicle.

5. The method according to claim 3, characterized in that: The step of selecting, for any predicted edge, a frame including at least a portion of the predicted edge as a radar data fusion region corresponding to the predicted edge includes: For any predicted edge, based on the length of the predicted edge and a preset length ratio, determine the frame selection length of the predicted edge; Taking the middle point of the predicted edge as the center, taking the portion of the frame selection length as the frame selection position of the predicted edge, and selecting the frame containing the frame selection position as the radar data fusion area corresponding to the predicted edge.

6. The method according to claim 4, characterized in that The step of adjusting the visual fusion position of the static vehicle based on the fitting edge corresponding to the radar data fusion area to obtain the target predicted position of the static vehicle includes: Determining a fusion edge corresponding to the radar data fusion area based on a fitting edge and a predicted edge corresponding to the radar data fusion area; A radar data fusion border of the static vehicle is generated based on the multiple fusion edges of the static vehicle, and the static vehicle is moved from the visual fusion position to the radar data fusion border to obtain a target predicted position of the static vehicle.

7. The method according to any one of claims 1 to 6, characterized in that The preset area weight is set based on the perception evaluation error of the preset area, and the preset area weight corresponding to the preset area is negatively correlated with the perception evaluation error.

8. The method according to any one of claims 1 to 6, characterized in that After determining the static vehicle prediction data for each frame of vehicle visual data, the method further includes: Based on the static vehicle recognition range corresponding to the current vehicle, deleting the static vehicle whose visual predicted position is outside the static vehicle recognition range; The posture information of the current vehicle is obtained, and based on the posture information, the visual predicted position of the static vehicle is converted into a world coordinate system.

9. The method according to any one of claims 1 to 6, characterized in that The static vehicle prediction data also includes the vehicle existence probability of the static vehicle; The method of performing fusion prediction on the visual prediction positions of the static vehicles based on the corresponding visual prediction positions and prediction weights of the static vehicles in each frame of static vehicle prediction data to obtain the visual fusion positions of the static vehicles includes: Based on the vehicle existence probability corresponding to each static vehicle in each frame of static vehicle prediction data and the prediction probability weight, the vehicle existence probability of each static vehicle is fused and predicted, and the static vehicle is filtered according to the obtained vehicle existence probability fusion prediction result of each static vehicle; Based on the visual prediction positions and prediction weights corresponding to the remaining static vehicles in each frame of static vehicle prediction data, a fusion prediction is performed on the visual prediction position of the static vehicle to obtain the visual fusion position of the static vehicle.

10. A static vehicle identification device, characterized in that: The device comprises: A visual perception module, used to obtain multiple frames of vehicle visual data corresponding to the current vehicle, and determine static vehicle prediction data for each frame of vehicle visual data, wherein the static vehicle prediction data includes a visual prediction position of at least one static vehicle; A region determination module, for determining, for each frame of static vehicle prediction data, a region where each static vehicle is located relative to the current vehicle based on the visual prediction position and a plurality of preset regions corresponding to the current vehicle, and determining a prediction weight for each static vehicle according to a preset region weight corresponding to the region; The position fusion module is used to perform fusion prediction on the visual prediction positions of each static vehicle based on the corresponding visual prediction positions and prediction weights of each static vehicle in each frame of static vehicle prediction data to obtain the visual fusion positions of each static vehicle.

11. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

12. A computer device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 9 is implemented.

Images