Obstacle contour detection method, device, equipment and medium

By combining vision sensors and ultrasonic sensors, we use correlation analysis and Kalman filtering technology to identify obstacles, solving the problems of slow obstacle detection speed and low accuracy in the prior art, and improving the accuracy and real-time performance of obstacle profile detection , enhances the safety of vehicle driving.

CN118587684BActive Publication Date: 2024-11-19IMOTION AUTOMOTIVE TECH (SUZHOU) CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411065015.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-05
Publication Date
2024-11-19
Estimated Expiration
2044-08-05

AI Technical Summary

Technical Problem

The existing obstacle detection technology adopts a variety of high-cost sensors, resulting in slow detection speed and low accuracy, affecting the safety of vehicle driving.

Method used

The combination of vision sensors and ultrasonic sensors is adopted to identify dot-shaped and linear obstacles through correlation analysis and Kalman filtering technology, improve the accuracy and range of obstacle profile detection, and reduce hardware costs and data volume.

Benefits of technology

It improves the accuracy and real-timeness of obstacle profile detection, reduces hardware costs and data volume, and enhances the safety of vehicle driving.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118587684B_ABST
    Figure CN118587684B_ABST
Patent Text Reader

Abstract

The present application discloses an obstacle contour detection method, device, equipment and medium, which relates to the field of obstacle detection, including: performing correlation analysis on the current obstacles and historical obstacles around the vehicle to obtain correlation analysis results; the types of obstacles include obstacles determined based on visual sensors and obstacles determined based on ultrasonic sensors, and the shapes of obstacles include point obstacles and linear obstacles obtained by connecting point obstacles; based on the correlation analysis results, the contours of the current obstacle and the historical obstacles are connected to obtain the current obstacle contour; based on the existence probability of each point obstacle in the current obstacle contour, the current obstacle contour is updated. By introducing point and line obstacles, the present application improves the diversity of correlation analysis, making the obstacle contour detection more accurate, and uses visual sensors and ultrasonic sensors to improve the accuracy and range of obstacle contour detection to ensure the safety of vehicle driving.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of obstacle detection, and in particular to an obstacle contour detection method, device, equipment and medium. Background Art

[0002] In the actual application of obstacle detection, considering the requirements of cost-effectiveness, practicality and real-time performance, if a combination of expensive lidar and various types of cameras is used, it will not only bring about high cost problems, but also the more sensors are used, the more data is actually used for obstacle detection. Correspondingly, the obstacle detection speed is slower, which has a negative impact on the real-time performance of obstacle detection. In addition, the accuracy of obstacle detection in this way is also low. Obstacle detection with poor real-time performance and low accuracy will seriously affect the safety of vehicle driving. Summary of the invention

[0003] In view of this, the purpose of the present invention is to provide an obstacle contour detection method, device, equipment and medium, which improves the diversity of correlation analysis by introducing point obstacles and linear obstacles, making the obstacle contour detection more accurate, and uses visual sensors and ultrasonic sensors to improve the accuracy and range of obstacle contour detection to ensure the safety of vehicle driving. The specific scheme is as follows:

[0004] In a first aspect, the present application provides an obstacle contour detection method, comprising:

[0005] Performing correlation analysis on current obstacles and historical obstacles around the vehicle to obtain correlation analysis results; wherein the types of obstacles include first obstacles determined based on a visual sensor and second obstacles determined based on an ultrasonic sensor, and the shapes of obstacles include point obstacles and / or linear obstacles obtained by connecting point obstacles;

[0006] Connecting the contours of the current obstacle and the historical obstacle based on the association analysis result to obtain the current obstacle contour;

[0007] The current obstacle contour is updated based on the existence probability of each of the point-shaped obstacles in the current obstacle contour.

[0008] Optionally, the correlation analysis of the current obstacles around the vehicle and the historical obstacles to obtain the correlation analysis result includes:

[0009] Performing a correlation analysis on the current obstacle and the hit obstacles in the historical obstacles that meet the preset stability condition to obtain a corresponding correlation analysis result;

[0010] Among them, the preset stability condition is that the covariance of the hit obstacle is smaller than the covariance of the current obstacle; the covariance is inversely proportional to the stability.

[0011] Optionally, the correlation analysis of the current obstacles around the vehicle and the historical obstacles to obtain the correlation analysis result includes:

[0012] Calculating the correlation between the current obstacle and the historical obstacle based on the Mahalanobis distance using the preset correlation rule, and determining the comparison result between the maximum correlation among the correlations and the preset correlation threshold; wherein the preset correlation rule includes a first correlation rule between point obstacles, a second correlation rule between point obstacles and linear obstacles, and a third correlation rule between linear obstacles;

[0013] Based on the comparison result, an association analysis result between the current obstacle and the historical obstacle is determined; wherein the association analysis result includes an association type.

[0014] Optionally, the calculating the correlation between the current obstacle and the historical obstacle based on the Mahalanobis distance by using a preset association rule includes:

[0015] The first association rule is used and based on the Mahalanobis distance, and the correlation between the current point obstacle and the historical point obstacle is calculated according to the updated covariance; wherein the updated covariance is the covariance obtained by updating the covariance of the current point obstacle and the historical point obstacle by using Kalman filtering.

[0016] Optionally, the calculating the correlation between the current obstacle and the historical obstacle based on the Mahalanobis distance by using a preset association rule includes:

[0017] Using the second association rule and based on the Mahalanobis distance, respectively calculate the first initial association degree between the current point obstacle and two end points of the historical linear obstacle, and calculate the first initial association degree based on the distance ratio to obtain the association degree between the current point obstacle and the historical linear obstacle;

[0018] Among them, the distance ratio is a ratio determined based on the distances from the target projection point to the two end points on the historical linear obstacle respectively; the target projection point is the projection point with the shortest projection distance from the current point obstacle to the historical linear obstacle.

[0019] Optionally, the calculating the correlation between the current obstacle and the historical obstacle based on the Mahalanobis distance by using a preset association rule includes:

[0020] Calculating a second initial correlation degree between each point obstacle on the current linear obstacle and each point obstacle on the historical linear obstacle based on the Mahalanobis distance by using the third correlation rule, and determining a correlation degree between the current linear obstacle and the historical linear obstacle based on the second initial correlation degree;

[0021] Alternatively, the third association rule is used to calculate the third initial association degree between each point obstacle on the current linear obstacle and the historical linear obstacle based on the Mahalanobis distance, and the association degree between the current linear obstacle and the historical linear obstacle is determined based on the third initial association degree.

[0022] Optionally, the connecting the contours of the current obstacle and the historical obstacle based on the association analysis result to obtain the current obstacle contour includes:

[0023] Using Kalman filtering and based on the association analysis result, the position and covariance of the current obstacle are updated to obtain an updated current obstacle;

[0024] The updated current obstacle and the historical obstacle contours are connected based on the association analysis result to obtain the current obstacle contour.

[0025] Optionally, updating the current obstacle contour based on the existence probability of each of the point-shaped obstacles in the current obstacle contour includes:

[0026] Setting an initial existence probability for a point obstacle in the current obstacle that does not exist in the historical obstacles, and updating the existence probability of the point obstacle in the historical obstacles based on whether the historical obstacle exists in the current obstacle;

[0027] Point obstacles whose existence probability in the current obstacle contour is less than a preset probability threshold are eliminated to update the current obstacle contour.

[0028] In a second aspect, the present application provides an obstacle contour detection device, comprising:

[0029] A correlation analysis module is used to perform correlation analysis on the current obstacles around the vehicle and the historical obstacles to obtain correlation analysis results; wherein the types of obstacles include a first obstacle determined based on a visual sensor and a second obstacle determined based on an ultrasonic sensor, and the shapes of obstacles include point obstacles and / or linear obstacles obtained by connecting point obstacles;

[0030] A contour connection module, used for connecting the contours of the current obstacle and the historical obstacle based on the association analysis result to obtain the current obstacle contour;

[0031] The contour updating module is used to update the current obstacle contour based on the existence probability of each point obstacle in the current obstacle contour.

[0032] In a third aspect, the present application provides an electronic device, including:

[0033] Memory, used to store computer programs;

[0034] A processor is used to execute the computer program to implement the aforementioned obstacle contour detection method.

[0035] In a fourth aspect, the present application provides a computer-readable storage medium for storing a computer program, wherein the computer program implements the aforementioned obstacle contour detection method when executed by a processor.

[0036] In the present application, the current obstacle around the vehicle is analyzed for correlation with the historical obstacle to obtain the correlation analysis result; wherein the types of obstacles include a first obstacle determined based on a visual sensor and a second obstacle determined based on an ultrasonic sensor, and the shape of the obstacle includes a point obstacle and / or a linear obstacle obtained by connecting point obstacles; based on the correlation analysis result, the contours of the current obstacle and the historical obstacle are connected to obtain the current obstacle contour; based on the existence probability of each point obstacle in the current obstacle contour, the current obstacle contour is updated. It can be seen that the present application uses visual sensors and ultrasonic sensors to connect the contours of obstacles around the vehicle by taking advantage of the long sensing distance of the visual sensor and the high sensing accuracy of the ultrasonic sensor, thereby improving the accuracy and range of obstacle contour detection; and the present application uses fewer types of vehicle sensors to reduce the hardware cost required for obstacle contour detection, reduce the data required for obstacle contour detection, and improve the speed of obstacle contour detection, thereby improving the safety of vehicle driving according to the detected obstacle contour with higher accuracy, longer range, and higher real-time performance. Furthermore, the present application introduces point obstacles and linear obstacles, so that correlation analysis between point obstacles, correlation analysis between linear obstacles, and correlation analysis between point obstacles and linear obstacles can be performed, thereby improving the diversity of correlation analysis and making the obstacle contour detection more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0038] Figure 1 A flow chart of an obstacle contour detection method disclosed in this application;

[0039] Figure 2 A schematic diagram of obstacle detection disclosed in this application;

[0040] Figure 3 A schematic diagram of a drivable area boundary disclosed in this application;

[0041] Figure 4 A schematic diagram of obstacle contour connection based on extended association disclosed in this application;

[0042] Figure 5 A relevance analysis flow chart disclosed in this application;

[0043] Figure 6 Another correlation analysis flow chart disclosed in this application;

[0044] Figure 7 A flowchart of obstacle contour detection disclosed in this application;

[0045] Figure 8 This is a schematic diagram of the structure of an obstacle contour detection device disclosed in this application;

[0046] Fig. 9 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0047] 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 only 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.

[0048] In the actual application of obstacle detection, if a combination of expensive laser radar and various types of cameras is used, it will not only bring about the problem of high cost, but also the more sensors are used, the more data actually used for obstacle detection will be, and accordingly, the slower the obstacle detection speed will be, which will have a negative impact on the real-time performance of obstacle detection and further affect the safety of vehicle driving. To this end, the present application provides an obstacle contour detection method, which improves the diversity of correlation analysis by introducing point obstacles and line obstacles, making the obstacle contour detection more accurate, and uses visual sensors and ultrasonic sensors to improve the accuracy and range of obstacle contour detection to ensure the safety of vehicle driving.

[0049] See also Figure 1As shown, an embodiment of the present invention discloses an obstacle contour detection method, comprising:

[0050] Step S11: performing a correlation analysis on the current obstacles around the vehicle and the historical obstacles to obtain a correlation analysis result; wherein the types of obstacles include a first obstacle determined based on a visual sensor and a second obstacle determined based on an ultrasonic sensor, and the shapes of obstacles include point obstacles and / or linear obstacles obtained by connecting point obstacles.

[0051] In this embodiment, obstacles around the vehicle are continuously sensed by visual sensors and ultrasonic sensors, and the perceived obstacles are analyzed for relevance by the contour detection system, so that subsequent obstacle contour connections can be made based on the results of the relevance analysis. Among them, the relevance analysis can be performed based on the Mahalanobis distance, which can not only measure the degree of relevance, but also indicate the extension or association direction of point obstacles or linear obstacles. Therefore, the results of the relevance analysis include the degree of relevance and the type of relevance; the types of relevance include Kalman relevance, extended relevance, aggregate relevance, no relevance, etc. Kalman relevance is used to update the location attributes of obstacles, extended relevance is used to update new obstacle contours, aggregate relevance is used to update important joint point information of obstacles, and no relevance indicates that the current obstacle is a newly detected obstacle.

[0052] It should be noted that the current obstacles around the vehicle can be obstacles currently determined based on visual sensors or obstacles currently determined based on ultrasonic sensors. As for the historical obstacles around the vehicle, they can be obstacles historically determined based on visual sensors or obstacles historically determined based on ultrasonic sensors, or obstacle outlines detected historically. For example, a correlation analysis can be performed between obstacles currently determined based on visual sensors around the vehicle and obstacles historically determined based on visual sensors / ultrasonic sensors, or between obstacles currently determined based on ultrasonic sensors around the vehicle and obstacles historically determined based on visual sensors / ultrasonic sensors, or between obstacles currently determined based on visual sensors / ultrasonic sensors around the vehicle and obstacle outlines detected historically, without specific limitation here.

[0053] Among them, for the first obstacle determined based on the visual sensor, the distortion-corrected image obtained by the visual sensor when detecting the target obstacle around the vehicle is spliced ​​into a top view, and the boundary of the drivable area around the vehicle is detected from the top view, and then the boundary of the drivable area is filtered to obtain the first obstacle including point obstacles and / or linear obstacles. It should be noted that in order to be able to perceive the vehicle's surrounding environment information in all directions, multiple visual sensors of the same type can be installed on the vehicle, so that the fields of view of the multiple visual sensors overlap with each other, thereby achieving all-round 360-degree perception of the vehicle's surrounding environment information.

[0054] Taking the fisheye camera as an example, the fisheye camera has a relatively large detection field of view of about 180 degrees. By installing four fisheye cameras on the front, back, left and right sides of the vehicle, the vehicle's surrounding environment information can be fully sensed. Figure 2 As shown, the original image captured by the fisheye camera is obtained, and the original image is subjected to distortion correction to obtain a distortion-corrected image. Then, obstacle detection is performed on the distortion-corrected image, and an obstacle detection frame and obstacle type are marked on the distortion-corrected image to obtain the target obstacle; for example, obstacle types include car, pillar, block, etc. Further, as Figure 3 As shown in FIG, the distortion-corrected images are stitched into a top view, and the drivable area (freespace) detection is performed on the top view to obtain the drivable area boundary around the vehicle, for example Figure 3 The blue line in is the boundary of the drivable area. Considering that there is a non-obstacle boundary in the boundary of the drivable area, it is necessary to further filter the boundary of the drivable area to obtain the first obstacle.

[0055] Specifically, the pseudo-boundaries and area boundary points in the boundary of the drivable area are filtered to obtain the first obstacle; wherein the pseudo-boundaries include image boundaries and / or non-obstacle boundaries caused by the limited field of view of the visual sensor; the area boundary points are boundary points whose boundary detection probability is less than the preset obstacle probability. It can be understood that the boundary of the drivable area is actually a boundary composed of several area boundary points, and each area boundary point corresponds to a boundary detection probability. If the boundary detection probability is less than the preset obstacle probability, it indicates that the area boundary point may not be the point corresponding to the obstacle. Therefore, it is necessary to filter out the area boundary points whose boundary detection probability is less than the preset obstacle probability from the boundary of the drivable area. Furthermore, considering that the shooting range of the visual sensor is limited, it is necessary to filter out the image boundaries in the boundary of the drivable area, for example Figure 3The small uppermost boundary in the blue boundary is not the obstacle boundary, but the image boundary. It should also be noted that during the actual shooting process, the visual sensor may be blocked by some obstacles, which may limit the visual sensor's field of view. As a result, some non-obstacle boundaries may appear due to the limited visual sensor field of view, such as Figure 3 The elongated blue line on the upper left side of the blue boundary requires filtering out the non-obstacle boundary in the drivable area boundary due to the limited field of view of the visual sensor.

[0056] For the non-obstruction boundary caused by the limited field of vision of the visual sensor, the first vector between two adjacent area boundary points in the drivable area boundary can be calculated, for example, the first vector between area boundary point A and area boundary point B is calculated, and the second vector from the visual sensor to area boundary point A is calculated; then the vector angle between the first vector and the second vector is determined. If the vector angle is greater than the preset angle, it indicates that the area boundary between area boundary point A and area boundary point B is the non-obstruction boundary caused by the limited field of vision of the visual sensor. The preset angle can be set to 180 degrees, or it can be set to other angles but it needs to be greater than 90 degrees.

[0057] It should also be noted that after the target obstacle is detected from the distortion-corrected image, the corresponding associated obstacle can be determined from the first obstacle based on the position information of the target obstacle, and the associated obstacle is marked with the attribute information of the target obstacle; wherein the attribute information includes the type and height of the obstacle. It can be understood that the grounded boundary line is determined from the detection frame of the target obstacle, and the pixel coordinates of the midpoint of the boundary line are determined, and then the pixel coordinates are converted to the coordinate system of the first obstacle using the camera calibration parameters, so that the corresponding associated obstacle is determined from the first obstacle based on the converted coordinates, and the associated obstacle is marked with the attribute information of the target obstacle. wherein the camera calibration parameters include camera height, camera angle, etc.

[0058] Among them, for the second obstacle determined based on the ultrasonic sensor, the sound wave data sent by the ultrasonic sensor is obtained, and the obstacle distance information is determined based on the sound wave data; then the initial point obstacle is determined based on the obstacle distance information between different ultrasonic sensors using the triangulation positioning algorithm, and the initial point obstacle is clustered using the clustering algorithm, and the point obstacles in the same cluster obtained by clustering are connected to obtain the second obstacle including point obstacles and / or linear obstacles. It should be noted that it is also necessary to filter the unstable ultrasonic obstacles in the second obstacle to improve the accuracy of obstacle contour detection.

[0059] Step S12: Connect the contours of the current obstacle and the historical obstacles based on the association analysis result to obtain the contour of the current obstacle.

[0060] In this embodiment, after the association analysis result between the current obstacle and the historical obstacle is determined, the contours of the current obstacle and the historical obstacle are connected based on the association analysis result to obtain the current obstacle contour.

[0061] Specifically, the Kalman filter is used to update the position and covariance of the current obstacle based on the association analysis result to obtain the updated current obstacle; and the updated current obstacle and the historical obstacle contour are connected based on the association analysis result to obtain the current obstacle contour. Take the association analysis result as an extended association as an example, Figure 4 As shown in the figure, the black dots are historical point obstacles; the red dots are current point obstacles; the circles represent the covariance of the point obstacles, where the covariance is inversely proportional to the stability; the Kalman filter is used to update the position and covariance of the current point obstacle based on the extended association, and the updated current point obstacle and the outline of the historical obstacle are connected to obtain the current obstacle outline; Figure 4 The left image in the figure shows the outline of the historical obstacle. Figure 4 The current obstacle outline is shown in the right figure of .

[0062] Among them, the formula involved in Kalman filtering is as follows:

[0063] x k =Ax K-1 +Bu K-1 ;

[0064] P k =AP K-1 A T +Q;

[0065] K k =P k H T / (HP k H T +R);

[0066] x K =x k +K k (z k -Hx k );

[0067] P K =(IK k H)P k ;

[0068] Where A represents the state transfer matrix; u K-1 represents the control vector of the historical point obstacle; B represents the matrix that converts the control vector into the state; x K-1Indicates historical point obstacles; x k Represents the predicted point obstacle; T represents the transpose of the matrix; P K-1 represents the covariance of historical point obstacles; Q represents the covariance of process excitation noise; P k represents the prediction covariance; H represents the preset conversion matrix; R represents the measurement noise covariance; K k represents the Kalman gain; z k Indicates the current point obstacle actually detected; x K Indicates the updated current point obstacle; P K Represents the updated covariance of the current point obstacle.

[0069] That is, the predicted point obstacles and the predicted covariance of the predicted point obstacles are first estimated through the historical point obstacles and the covariance of the historical point obstacles, and then the Kalman gain is calculated based on the predicted covariance, so as to use the Kalman gain and based on the actually detected current point obstacles, the predicted point obstacles and the predicted covariance, determine the updated current point obstacles and the updated covariance of the current point obstacles, so as to realize the update of the position and covariance of the current obstacle through Kalman filtering.

[0070] Step S13: updating the current obstacle contour based on the existence probability of each point obstacle in the current obstacle contour.

[0071] In this embodiment, since each point obstacle in the current obstacle contour has a corresponding existence probability, and for a point obstacle with a low existence probability, considering that the point obstacle may no longer exist, it is necessary to eliminate the point obstacle from the current obstacle contour, that is, it is necessary to update the current obstacle contour based on the existence probability of each point obstacle in the current obstacle contour. In addition, this embodiment can store all nodes in the current obstacle contour, and can also store key nodes in the current obstacle contour to reduce the storage amount.

[0072] Specifically, first, an initial existence probability is set for each point obstacle in the current obstacle that does not exist in the historical obstacles, and based on whether the historical obstacle exists in the current obstacle, the existence probability of each point obstacle in the historical obstacles is updated; then, the point obstacles in the current obstacle contour whose existence probability is less than a preset probability threshold are eliminated to achieve the update of the current obstacle contour.

[0073] Based on whether the historical obstacle exists in the current obstacle, the existence probability of each point obstacle in the historical obstacle is updated, specifically, the existence probability of the point obstacle in the historical obstacle that exists in the current obstacle is increased, for example, the first preset probability value is increased; and the existence probability of the point obstacle in the historical obstacle that does not exist in the current obstacle is reduced, for example, the second preset probability value is reduced. The first preset probability value and the second preset probability value can be set to the same value or different values, which is not limited here.

[0074] For ultrasonic sensors, the existence probability of each point obstacle in the historical obstacles can be updated according to the obstacle distance information determined based on the sound wave data. Specifically, if there are point obstacles at both 3 meters due east and 8 meters due east in the past, and there is only a point obstacle at 8 meters due east at present, the existence probability of the point obstacle at 3 meters due east is reduced, and the existence probability of the point obstacle at 8 meters due east is increased.

[0075] For the visual sensor, the existence probability of each point obstacle in the historical obstacles can be updated according to the boundary of the drivable area. Specifically, if a point obstacle in the historical obstacles exists in the boundary of the drivable area currently determined based on the visual sensor, the existence probability of the point obstacle is reduced.

[0076] like Figure 5 As shown, in step S11, the correlation analysis of the current obstacles around the vehicle and the historical obstacles is performed to obtain the correlation analysis result, including:

[0077] Step S21: divide the current obstacle and the historical obstacle into influence areas of different vehicle sensors respectively; wherein the vehicle sensors include visual sensors and ultrasonic sensors.

[0078] Step S22: performing correlation analysis on the current obstacles and historical obstacles in the same impact area to obtain correlation analysis results.

[0079] In a specific implementation, in order to avoid the problem of large amount of calculation caused by directly performing correlation analysis on the current obstacle and all historical obstacles, this embodiment first divides the current obstacle and the historical obstacles into the influence areas of different vehicle sensors, and then performs correlation analysis on the current obstacle and the historical obstacles in the same influence area to obtain the correlation analysis result, thereby reducing a certain amount of calculation and improving the efficiency of obstacle contour detection.

[0080] As for dividing the current obstacles and historical obstacles into the influence areas of different vehicle sensors, the current vehicle posture is first determined based on the vehicle body sensor, and the posture information of each vehicle sensor is updated based on the current vehicle posture. Therefore, according to the posture information of each vehicle sensor and the field of view detection range, the current obstacles and historical obstacles are divided into the influence areas of different vehicle sensors.

[0081] Specifically, taking any one of the vehicle sensors as an example, a corresponding sensor coordinate system is established based on the posture information of the vehicle sensor, and the reference coordinates of the current obstacle and the historical obstacle in the unified reference coordinate system are transformed into the sensor coordinate system to obtain the corresponding sensor coordinates; wherein the unified reference coordinate system can be a world coordinate system or a vehicle body coordinate system when obstacle contour detection is turned on. Then, based on the field of view of the vehicle sensor, the field of view angle detection range in the sensor coordinate system is determined, and the current obstacle and the historical obstacle whose sensor coordinates are within the field of view angle detection range of the vehicle sensor are classified into the influence area of ​​the vehicle sensor.

[0082] In step S11, a correlation analysis is performed on the current obstacle and the historical obstacles around the vehicle to obtain a correlation analysis result; in another specific implementation, a correlation analysis is performed on the current obstacle and the hit obstacle in the historical obstacles that meets the preset stability condition to obtain a corresponding correlation analysis result; wherein the preset stability condition is that the covariance of the hit obstacle is less than the covariance of the current obstacle; and the covariance is inversely proportional to the stability.

[0083] In this embodiment, considering that the stability of each obstacle is different, in order to ensure the accuracy of the correlation analysis result and the accuracy of the obstacle contour detection, the current obstacle and the historical obstacle that is more stable than the current obstacle can be subjected to correlation analysis to obtain the corresponding correlation analysis result. Specifically, a hit obstacle that meets the preset stability condition is determined from the historical obstacles, wherein the covariance of the hit obstacle is smaller than the covariance of the current obstacle, and the smaller the covariance, the higher the stability. Finally, the current obstacle and the hit obstacle are subjected to correlation analysis to obtain the corresponding correlation analysis result.

[0084] like Figure 6 As shown, in step S11, the correlation analysis of the current obstacles around the vehicle and the historical obstacles is performed to obtain the correlation analysis result; including:

[0085] Step S31: Calculate the correlation between the current obstacle and the historical obstacles based on the Mahalanobis distance using the preset association rules, and determine the comparison result between the maximum correlation among the correlations and the preset correlation threshold; wherein the preset association rules include a first association rule between point obstacles, a second association rule between point obstacles and linear obstacles, and a third association rule between linear obstacles.

[0086] Step S32: determining the association analysis result between the current obstacle and the historical obstacle based on the comparison result; wherein the association analysis result includes the association type.

[0087] In another specific implementation, the association degree between the current obstacle and the historical obstacle is calculated based on the preset association rule and the Mahalanobis distance to determine the maximum association degree from each association degree, and then the association analysis result between the current obstacle and the historical obstacle is determined based on the comparison result of the maximum association degree and the preset association degree threshold. Among them, the preset association rule includes a first association rule between point obstacles, a second association rule between point obstacles and linear obstacles, and a third association rule between linear obstacles; that is, this embodiment can calculate the association degree between point obstacles according to the first association rule, the association degree between point obstacles and linear obstacles according to the second association rule, and the association degree between linear obstacles according to the third association rule. In addition, since the Mahalanobis distance can not only calculate the association degree, but also determine the extension or association direction of the point obstacle or linear obstacle, the association analysis result includes not only the association degree but also the association type. It should be noted that if the maximum correlation is greater than the preset correlation threshold, it indicates that there is a correlation between the current obstacle and the historical obstacle, and the corresponding correlation type is determined, such as Kalman correlation, extended correlation, aggregate correlation, etc.; if the maximum correlation is less than or equal to the preset correlation threshold, it indicates that there is no correlation between the current obstacle and the historical obstacle, and the correlation type at this time is no correlation.

[0088] For calculating the correlation between the current obstacle and the historical obstacle based on the preset association rule and the Mahalanobis distance, taking the current obstacle and the historical obstacle as point obstacles, the first association rule needs to be adopted. The specific operation is to first use the Kalman filter to update the covariance of the current point obstacle and the historical point obstacle to obtain the updated covariance, and then use the Mahalanobis distance and calculate the correlation between the current point obstacle and the historical point obstacle based on the updated covariance. The formula of the Mahalanobis distance involved is as follows:

[0089] ;

[0090] Where x represents the current point obstacle; y represents the historical point obstacle; T represents the transpose of the matrix; ∑ represents the updated covariance; D M (x,y) represents the Mahalanobis distance.

[0091] For calculating the correlation between the current obstacle and the historical obstacle based on the preset association rule and the Mahalanobis distance, taking the current obstacle as a point obstacle and the historical obstacle as a linear obstacle as an example, the second association rule needs to be adopted at this time. The specific operation is to first project the current point obstacle onto the historical linear obstacle. If there are multiple projection points, the target projection point with the smallest projection distance is determined from the multiple projection points. Then, the distance ratio is determined based on the distance from the target projection point to the two endpoints on the historical linear obstacle, and the first initial correlation between the current point obstacle and the two endpoints on the historical linear obstacle is calculated based on the Mahalanobis distance. Then, the two first initial correlations are weighted based on the distance ratio to obtain the correlation between the current point obstacle and the historical linear obstacle. Among them, the first initial correlation between the current point obstacle and the two endpoints on the historical linear obstacle calculated based on the Mahalanobis distance can be determined by the above-mentioned method of calculating the correlation between the two point obstacles. Furthermore, for calculating the correlation between the current linear obstacle and the historical point-shaped obstacles, reference may be made to the calculation method of the correlation between the current point-shaped obstacles and the historical linear obstacles, which will not be described repeatedly herein.

[0092] For calculating the correlation between the current obstacle and the historical obstacle based on the preset correlation rule and the Mahalanobis distance, taking the current obstacle and the historical obstacle as linear obstacles, the third correlation rule needs to be adopted. One specific operation is to calculate the second initial correlation between each point obstacle on the current linear obstacle and each point obstacle on the historical linear obstacle by using the Mahalanobis distance in the manner of calculating the correlation between two point obstacles, and perform weighted calculation on the second initial correlation to obtain the correlation between the current linear obstacle and the historical linear obstacle. Another specific operation is to calculate the third initial correlation between each point obstacle on the current linear obstacle and the historical linear obstacle by using the Mahalanobis distance in the manner of calculating the correlation between point obstacles and linear obstacles, and perform weighted calculation on the third initial correlation to obtain the correlation between the current linear obstacle and the historical linear obstacle.

[0093] It should be noted that there are three specific implementation methods for performing correlation analysis on the current obstacles and historical obstacles around the vehicle to obtain the correlation analysis results. One is to perform correlation analysis on the current obstacles and historical obstacles in the same impact area to obtain the correlation analysis results; one is to perform correlation analysis on the current obstacles and the hit obstacles in the historical obstacles that meet the preset stability conditions to obtain the correlation analysis results; and another is to use the preset association rules and calculate the correlation analysis results between the current obstacles and the historical obstacles based on the Mahalanobis distance. The above three methods can be used alone or in combination, and are not limited here.

[0094] It can be seen that the present application adopts visual sensors and ultrasonic sensors to connect the outlines of obstacles around the vehicle by taking advantage of the long sensing distance of visual sensors and the high sensing accuracy of ultrasonic sensors, thereby improving the accuracy and range of obstacle outline detection; and the present application can reduce the hardware cost required for obstacle outline detection and reduce the data required for obstacle outline detection by using fewer types of vehicle sensors, thereby improving the speed of obstacle outline detection, thereby improving the safety of vehicle driving based on the detected obstacle outlines with higher accuracy, longer range and higher real-time performance. Furthermore, the present application introduces point obstacles and linear obstacles, so that correlation analysis between point obstacles, correlation analysis between linear obstacles and correlation analysis between point obstacles and linear obstacles can be performed, thereby improving the diversity of correlation analysis and making the accuracy of obstacle outline detection higher.

[0095] Take the current obstacles around the ego vehicle as obstacles determined by ultrasonic sensors and the historical obstacles around the ego vehicle as obstacles determined by fisheye cameras as an example, see Figure 7 As shown, an embodiment of the present invention discloses an obstacle contour detection method, which is applied to a contour detection system. The contour detection system includes a visual processing module, an ultrasonic processing module and a perception fusion module; wherein the obstacle contour detection method includes:

[0096] The original visual image taken by the fisheye camera is received by the visual processing module, and the original visual image is distorted to obtain a distortion-corrected image; the distortion-corrected image is then spliced ​​into a top-down view, and the top-down view is subjected to a drivable area (freespace) detection to obtain the boundary of the drivable area around the vehicle, which is then sent to the perception fusion module.

[0097] The boundary of the drivable area is processed by the perception fusion module to filter the image boundary, non-obstacle boundary and area boundary points whose boundary detection probability is less than the preset obstacle probability in the boundary of the drivable area, thereby obtaining historical obstacles including point obstacles and / or line obstacles.

[0098] The ultrasonic processing module receives the acoustic wave data sent by the ultrasonic sensor and filters the acoustic wave data to determine the obstacle distance information based on the filtered acoustic wave data. The triangulation positioning algorithm is used to determine the initial point obstacles based on the obstacle distance information between different ultrasonic sensors, and the clustering algorithm is used to cluster the initial point obstacles, and the point obstacles in the same cluster obtained by clustering are connected to obtain the current obstacle including point obstacles and / or linear obstacles.

[0099] The current obstacles including point obstacles and / or line obstacles sent by the ultrasonic processing module are received through the perception fusion module, and unstable point obstacles and / or line obstacles in the current obstacles are filtered to obtain the filtered current obstacles.

[0100] The current obstacle and the historical obstacle are divided into the influence area of ​​different vehicle sensors respectively through the perception fusion module; wherein the vehicle sensor includes a fisheye camera and an ultrasonic sensor; then the current obstacle and the historical obstacle in the same influence area are analyzed for correlation based on the preset association rule and Mahalanobis distance to obtain the corresponding correlation analysis result. Among them, the correlation between the point obstacles can be calculated according to the first association rule between the point obstacles, the correlation between the point obstacles and the linear obstacles can be calculated according to the second association rule between the point obstacles and the linear obstacles, and the correlation between the linear obstacles can be calculated according to the third association rule between the linear obstacles. Further, the position and covariance of the current obstacle are updated based on the association analysis result by using Kalman filtering to obtain the updated current obstacle, and the contours of the updated current obstacle and the historical obstacle are connected based on the association analysis result to obtain the current obstacle contour.

[0101] The perception fusion module sets the initial existence probability for each point obstacle in the current obstacle that does not exist in the historical obstacle, and updates the existence probability of each point obstacle in the historical obstacle based on whether the historical obstacle exists in the current obstacle. Then, the point obstacles in the current obstacle contour whose existence probability is less than the preset probability threshold are eliminated to complete the update of the current obstacle contour, and the key nodes in the current obstacle contour can be stored in a simplified storage manner to reduce the storage volume.

[0102] It can be seen that the present application adopts a fisheye camera and an ultrasonic sensor to connect the outlines of obstacles around the vehicle by taking advantage of the long sensing distance of the fisheye camera and the high sensing accuracy of the ultrasonic sensor, thereby improving the accuracy and range of obstacle outline detection; and the present application can reduce the hardware cost required for obstacle outline detection and reduce the data required for obstacle outline detection by using fewer types of vehicle sensors, thereby improving the speed of obstacle outline detection, thereby improving the safety of vehicle driving based on the detected obstacle outlines with higher accuracy, longer range, and higher real-time performance. Furthermore, the present application introduces point obstacles and linear obstacles, so that correlation analysis between point obstacles, correlation analysis between linear obstacles, and correlation analysis between point obstacles and linear obstacles can be performed, thereby improving the diversity of correlation analysis and making the accuracy of obstacle outline detection higher.

[0103] See also Figure 8 As shown, an embodiment of the present invention discloses an obstacle contour detection device, comprising:

[0104] The correlation analysis module 11 is used to perform correlation analysis on the current obstacles around the vehicle and the historical obstacles to obtain correlation analysis results; wherein the types of obstacles include first obstacles determined based on the visual sensor and second obstacles determined based on the ultrasonic sensor, and the shapes of obstacles include point obstacles and / or linear obstacles obtained by connecting point obstacles;

[0105] A contour connection module 12, configured to connect the contours of the current obstacle and the historical obstacle based on the association analysis result to obtain the current obstacle contour;

[0106] The contour updating module 13 is used to update the current obstacle contour based on the existence probability of each point obstacle in the current obstacle contour.

[0107] It can be seen that the present application adopts visual sensors and ultrasonic sensors to connect the outlines of obstacles around the vehicle by taking advantage of the long sensing distance of visual sensors and the high sensing accuracy of ultrasonic sensors, thereby improving the accuracy and range of obstacle outline detection; and the present application can reduce the hardware cost required for obstacle outline detection and reduce the data required for obstacle outline detection by using fewer types of vehicle sensors, thereby improving the speed of obstacle outline detection, thereby improving the safety of vehicle driving based on the detected obstacle outlines with higher accuracy, longer range and higher real-time performance. Furthermore, the present application introduces point obstacles and linear obstacles, so that correlation analysis between point obstacles, correlation analysis between linear obstacles and correlation analysis between point obstacles and linear obstacles can be performed, thereby improving the diversity of correlation analysis and making the accuracy of obstacle outline detection higher.

[0108] In some specific embodiments, the correlation analysis module 11 includes:

[0109] A correlation analysis unit, used for performing correlation analysis on the current obstacle and the hit obstacle in the historical obstacles that meets the preset stability condition, so as to obtain a corresponding correlation analysis result;

[0110] Among them, the preset stability condition is that the covariance of the hit obstacle is smaller than the covariance of the current obstacle; the covariance is inversely proportional to the stability.

[0111] In some specific embodiments, the correlation analysis module 11 includes:

[0112] A correlation calculation submodule, used to calculate the correlation between the current obstacle and the historical obstacles by using a preset correlation rule and based on the Mahalanobis distance, and determine a comparison result between the maximum correlation among the correlations and a preset correlation threshold; wherein the preset association rules include a first association rule between point obstacles, a second association rule between point obstacles and linear obstacles, and a third association rule between linear obstacles;

[0113] An analysis result determination unit is used to determine the association analysis result between the current obstacle and the historical obstacle based on the comparison result; wherein the association analysis result includes an association type.

[0114] In some specific embodiments, the association degree calculation submodule includes:

[0115] The first association calculation unit is used to calculate the association between the current point obstacle and the historical point obstacle according to the updated covariance by using the first association rule and based on the Mahalanobis distance; wherein the updated covariance is the covariance obtained by updating the covariance of the current point obstacle and the historical point obstacle by using Kalman filtering.

[0116] In some specific embodiments, the association degree calculation submodule includes:

[0117] a second association degree calculation unit, configured to calculate, by using the second association rule and based on the Mahalanobis distance, a first initial association degree between the current point obstacle and two end points of the historical linear obstacle, and to calculate the first initial association degree based on the distance ratio, so as to obtain a correlation degree between the current point obstacle and the historical linear obstacle;

[0118] Among them, the distance ratio is a ratio determined based on the distances from the target projection point to the two end points on the historical linear obstacle respectively; the target projection point is the projection point with the shortest projection distance from the current point obstacle to the historical linear obstacle.

[0119] In some specific embodiments, the association degree calculation submodule includes:

[0120] a third association degree calculation unit, configured to calculate, by using the third association rule and based on the Mahalanobis distance, a second initial association degree between each point obstacle on the current linear obstacle and each point obstacle on the historical linear obstacle, and determine a correlation degree between the current linear obstacle and the historical linear obstacle based on the second initial association degree;

[0121] The fourth association degree calculation unit is used to calculate the third initial association degree between each point obstacle on the current linear obstacle and the historical linear obstacle respectively based on the Mahalanobis distance by using the third association rule, and determine the association degree between the current linear obstacle and the historical linear obstacle based on the third initial association degree.

[0122] In some specific embodiments, the contour connection module 12 includes:

[0123] An obstacle updating unit, configured to update the position and covariance of the current obstacle by using Kalman filtering and based on the association analysis result, so as to obtain an updated current obstacle;

[0124] A contour connecting unit is used to connect the updated contours of the current obstacle and the historical obstacle based on the association analysis result to obtain the current obstacle contour.

[0125] In some specific embodiments, the profile updating module 13 includes:

[0126] an existence probability updating unit, configured to set an initial existence probability for a point obstacle in the current obstacle that does not exist in the historical obstacle, and to update the existence probability of the point obstacle in the historical obstacle based on whether the historical obstacle exists in the current obstacle;

[0127] The obstacle elimination unit is used to eliminate point obstacles in the current obstacle contour whose existence probability is less than a preset probability threshold, so as to update the current obstacle contour.

[0128] Furthermore, the present application also discloses an electronic device. Fig. 9 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in the diagram cannot be regarded as any limitation on the scope of use of the present application.

[0129] Fig. 9 The present invention provides a schematic diagram of the structure of an electronic device 20 provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the obstacle contour detection method disclosed in any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0130] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0131] In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0132] The operating system 221 is used to manage and control the hardware devices and computer program 222 on the electronic device 20, and may be Windows Server, Netware, Unix, Linux, etc. In addition to the computer program that can be used to complete the obstacle contour detection method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 may further include a computer program that can be used to complete other specific tasks.

[0133] Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein the computer program, when executed by a processor, implements the aforementioned disclosed obstacle contour detection method. The specific steps of the method can refer to the corresponding contents disclosed in the aforementioned embodiments, and will not be repeated here.

[0134] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0135] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0136] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0137] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0138] The technical solution provided by the present application is introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for general technicians in this field, according to the idea of ​​the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for detecting an obstacle contour, characterized in that: include: Performing correlation analysis on current obstacles and historical obstacles around the vehicle to obtain correlation analysis results; wherein the types of obstacles include first obstacles determined based on a visual sensor and second obstacles determined based on an ultrasonic sensor, and the shapes of obstacles include point obstacles and / or linear obstacles obtained by connecting point obstacles; Connecting the contours of the current obstacle and the historical obstacle based on the association analysis result to obtain the current obstacle contour; Updating the current obstacle contour based on the existence probability of each of the point-shaped obstacles in the current obstacle contour; The correlation analysis of the current obstacles around the vehicle and the historical obstacles to obtain the correlation analysis result includes: Calculating the correlation between the current obstacle and the historical obstacle based on the Mahalanobis distance using the preset correlation rule, and determining the correlation analysis result based on the comparison result of the maximum correlation among the correlations and the preset correlation threshold; The method of calculating the correlation between the current obstacle and the historical obstacle based on the Mahalanobis distance by using the preset association rule includes: Using the first association rule and based on the Mahalanobis distance, the correlation between the current point obstacle and the historical point obstacle is calculated according to the updated covariance; wherein the updated covariance is the covariance obtained by updating the covariance of the current point obstacle and the historical point obstacle by using Kalman filtering; The method of calculating the correlation between the current obstacle and the historical obstacle based on the Mahalanobis distance by using the preset association rule includes: The first initial correlation between the current point obstacle and the two endpoints of the historical linear obstacle is calculated based on the Mahalanobis distance by using the second association rule, and the first initial correlation is calculated based on the distance ratio to obtain the correlation between the current point obstacle and the historical linear obstacle; wherein the distance ratio is a ratio determined based on the distances from the target projection point to the two endpoints of the historical linear obstacle; and the target projection point is a projection point with the smallest projection distance from the current point obstacle to the historical linear obstacle.

2. The obstacle contour detection method according to claim 1, characterized in that: The correlation analysis of the current obstacles and the historical obstacles around the vehicle to obtain the correlation analysis result includes: Performing a correlation analysis on the current obstacle and the hit obstacles in the historical obstacles that meet the preset stability condition to obtain a corresponding correlation analysis result; Among them, the preset stability condition is that the covariance of the hit obstacle is smaller than the covariance of the current obstacle; the covariance is inversely proportional to the stability.

3. The obstacle contour detection method according to claim 1, characterized in that: The correlation analysis of the current obstacles and the historical obstacles around the vehicle to obtain the correlation analysis result includes: Calculating the correlation between the current obstacle and the historical obstacle based on the Mahalanobis distance using the preset association rule, and determining the comparison result between the maximum correlation among the correlations and the preset correlation threshold; wherein the preset association rule includes the first association rule between point obstacles, the second association rule between point obstacles and linear obstacles, and the third association rule between linear obstacles; Based on the comparison result, an association analysis result between the current obstacle and the historical obstacle is determined; wherein the association analysis result includes an association type.

4. The obstacle contour detection method according to claim 3, characterized in that: The method of calculating the correlation between the current obstacle and the historical obstacle based on the Mahalanobis distance by using the preset association rule includes: Calculating a second initial correlation degree between each point obstacle on the current linear obstacle and each point obstacle on the historical linear obstacle based on the Mahalanobis distance by using the third correlation rule, and determining a correlation degree between the current linear obstacle and the historical linear obstacle based on the second initial correlation degree; Alternatively, the third association rule is used to calculate the third initial association degree between each point obstacle on the current linear obstacle and the historical linear obstacle based on the Mahalanobis distance, and the association degree between the current linear obstacle and the historical linear obstacle is determined based on the third initial association degree.

5. The obstacle contour detection method according to claim 1, characterized in that: The step of connecting the contours of the current obstacle and the historical obstacle based on the association analysis result to obtain the current obstacle contour includes: Using Kalman filtering and based on the association analysis result, the position and covariance of the current obstacle are updated to obtain an updated current obstacle; The updated current obstacle and the historical obstacle contours are connected based on the association analysis result to obtain the current obstacle contour.

6. The obstacle contour detection method according to any one of claims 1 to 5, characterized in that: The updating of the current obstacle contour based on the existence probability of each of the point-shaped obstacles in the current obstacle contour comprises: Setting an initial existence probability for a point obstacle in the current obstacle that does not exist in the historical obstacles, and updating the existence probability of the point obstacle in the historical obstacles based on whether the historical obstacle exists in the current obstacle; Point obstacles whose existence probability in the current obstacle contour is less than a preset probability threshold are eliminated to update the current obstacle contour.

7. An obstacle contour detection device, characterized in that: include: A correlation analysis module is used to perform correlation analysis on the current obstacles around the vehicle and the historical obstacles to obtain correlation analysis results; wherein the types of obstacles include a first obstacle determined based on a visual sensor and a second obstacle determined based on an ultrasonic sensor, and the shapes of obstacles include point obstacles and / or linear obstacles obtained by connecting point obstacles; A contour connection module, used for connecting the contours of the current obstacle and the historical obstacle based on the association analysis result to obtain the current obstacle contour; A contour updating module, used for updating the current obstacle contour based on the existence probability of each of the point-shaped obstacles in the current obstacle contour; The correlation analysis module is specifically used to: calculate the correlation between the current obstacle and the historical obstacle based on the preset correlation rule and the Mahalanobis distance, and determine the correlation analysis result based on the comparison result between the maximum correlation among the correlations and the preset correlation threshold; Wherein, the correlation analysis module includes: A first correlation calculation unit, configured to calculate the correlation between the current point obstacle and the historical point obstacle according to the updated covariance by using the first association rule and based on the Mahalanobis distance; wherein the updated covariance is the covariance obtained by updating the covariance of the current point obstacle and the historical point obstacle by using a Kalman filter; Wherein, the correlation analysis module includes: The second association degree calculation unit is used to calculate the first initial association degree between the current point obstacle and the two endpoints on the historical linear obstacle respectively based on the Mahalanobis distance by using the second association rule, and calculate the first initial association degree based on the distance ratio to obtain the association degree between the current point obstacle and the historical linear obstacle; wherein the distance ratio is a ratio determined based on the distances from the target projection point to the two endpoints on the historical linear obstacle respectively; and the target projection point is a projection point with the smallest projection distance from the current point obstacle to the historical linear obstacle.

8. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the obstacle contour detection method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: Used to store a computer program, which, when executed by a processor, implements the obstacle contour detection method according to any one of claims 1 to 6.