Moving obstacle tracking method and device and storage medium

Abstract

The invention relates to a moving obstacle tracking method and device and a storage medium, and the method comprises the steps: carrying out the moving target detection of a current video frame, obtaining at least one moving target region, and extracting the contour feature points of the region; calculating optical flow vectors of the contour feature points of the at least one moving target area, carrying out clustering, and determining the contour feature points contained in each moving obstacle according to a clustering result; calculating the image coordinates of each moving obstacle according to the image coordinates of the contour feature points contained in each moving obstacle, and calculating the moving speed of each moving obstacle according to the optical flow vectors of the contour feature points contained in each moving obstacle; and inputting the image coordinate and the moving speed of each moving obstacle into a preset Kalman filtering model for filtering processing to obtain the image coordinate and the moving speed of each moving obstacle after filtering, thereby ensuring that information such as the position and the speed of the moving obstacle does not suddenly change.

Description

Technical Field

[0001] The present application relates to the field of vehicle perception technology, and in particular to a mobile obstacle tracking method and device, and a storage medium thereof. Background Art

[0002] When tracking moving obstacles around the vehicle, if the sensor's measurement results are used directly, the obstacle measurement results may change suddenly due to errors, which is unacceptable for vehicle perception. Therefore, it is necessary to track based on the sensor's measurement results to ensure that the position, speed and other information of the moving obstacle will not change suddenly. Summary of the invention

[0003] The purpose of the present application is to propose a moving obstacle tracking method and device, and a computer-readable storage medium to ensure that the position, speed and other information of the moving obstacle will not change suddenly.

[0004] To achieve the above object, an embodiment of the present application provides a moving obstacle tracking method, the method comprising:

[0005] Get the current video frame output by the vehicle's surround view camera;

[0006] Performing moving target detection on the current video frame to obtain at least one moving target area, and extracting contour feature points of the at least one moving target area;

[0007] Calculate the optical flow vector of the contour feature points of the at least one moving target area;

[0008] Clustering the optical flow vectors of the contour feature points of the at least one moving target area, and determining the contour feature points contained in each moving obstacle according to the clustering result;

[0009] Calculating the image coordinates of each moving obstacle in the current video frame according to the image coordinates of the contour feature points contained in each moving obstacle in the current video frame, and calculating the moving speed of each moving obstacle according to the optical flow vector of the contour feature points contained in each moving obstacle;

[0010] The image coordinates and moving speed of each moving obstacle in the current video frame are input into a preset Kalman filter model for filtering processing to obtain the filtered image coordinates and moving speed of each moving obstacle in the current video frame.

[0011] In some embodiments, the method further comprises:

[0012] The filtered image coordinates of each moving obstacle in the current video frame are converted to a vehicle coordinate system and output.

[0013] In some embodiments, performing moving target detection on the current video frame to obtain at least one moving target area specifically includes:

[0014] The video frames of two historical moments adjacent to the current moment output by the vehicle surround view camera are obtained, and based on the three-frame difference method and the video frames of the two historical moments, moving target detection is performed on the current video frame to obtain at least one moving target area.

[0015] In some embodiments, clustering the optical flow vectors of the contour feature points of the at least one moving target area specifically includes:

[0016] The optical flow vectors of the contour feature points of the at least one moving target area are clustered using majority clustering, and all contour feature points belonging to the same moving obstacle are clustered in one cluster.

[0017] In some embodiments, the image coordinates of each moving obstacle in the current video frame are specifically: the image coordinates of the geometric center of the contour area formed by the contour feature points included in each moving obstacle in the current video frame.

[0018] In some embodiments, the moving speed of each moving obstacle is specifically: a weighted average value of optical flow vectors of contour feature points included in each moving obstacle.

[0019] In some embodiments, the method further comprises:

[0020] Grayscale processing is performed on the current video frame to obtain a grayscale image, and median filtering is performed on the grayscale image. The filtered grayscale image is used for moving target detection to obtain at least one moving target area.

[0021] Corresponding to the above method, the embodiment of the present application further provides a mobile obstacle tracking device, including:

[0022] A video frame acquisition module is used to acquire the current video frame output by the vehicle's surround view camera;

[0023] A target detection module, configured to perform moving target detection on the current video frame to obtain at least one moving target area, and extract contour feature points of the at least one moving target area;

[0024] An optical flow calculation module, used to calculate the optical flow vectors of the contour feature points of the at least one moving target area;

[0025] A clustering module, used to cluster the optical flow vectors of the contour feature points of the at least one moving target area, and determine the contour feature points contained in each moving obstacle according to the clustering result;

[0026] A motion information calculation module, used to calculate the image coordinates of each moving obstacle in the current video frame according to the image coordinates of the contour feature points contained in each moving obstacle in the current video frame, and calculate the moving speed of each moving obstacle according to the optical flow vector of the contour feature points contained in each moving obstacle;

[0027] The Kalman filter module is used to input the image coordinates and moving speed of each moving obstacle in the current video frame into a preset Kalman filter model for filtering processing to obtain the image coordinates and moving speed of each moving obstacle in the current video frame after filtering.

[0028] In some embodiments, the apparatus further comprises:

[0029] The coordinate conversion module is used to convert the filtered image coordinates of each moving obstacle in the current video frame into a vehicle coordinate system and output it.

[0030] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the mobile obstacle tracking method as described above is implemented.

[0031] The present invention provides a method and device for tracking a moving obstacle, and a computer-readable storage medium, which have the following beneficial effects:

[0032] The embodiment of the present application obtains several moving target areas by performing moving target detection on the video frame (surround view image) output by the vehicle's surround view camera, and performs contour feature point detection on the several moving target areas. Since a moving obstacle may include multiple moving target areas detected previously, the embodiment of the present application uses the optical flow vectors of the contour feature points to perform clustering to determine how many moving obstacles exist in the video frame, as well as all the contour feature points contained in each moving obstacle. Based on the image coordinates and optical flow vectors of all the contour feature points contained in each moving obstacle, the image coordinates and moving speed of the moving obstacle can be quickly calculated. Furthermore, the image coordinates and moving speed of the moving obstacle are used as model inputs of the Kalman filter model, and the module is used to perform Kalman filtering, and the model output is used as the final detection result. Compared with the measurement results directly using the sensor, it can better ensure that the position, speed and other information of the moving obstacle will not change suddenly, thereby improving the accuracy of the mobile obstacle detection.

[0033] Details and advantages not described in detail in the embodiments of the present application are described in detail in the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0035] Figure 1 This is a flow chart of a moving obstacle tracking method in an embodiment of the present application.

[0036] Figure 2 This is a structural diagram of a mobile obstacle tracking device in an embodiment of the present application. DETAILED DESCRIPTION

[0037] The detailed description of the drawings is intended as an illustration of the current embodiment of the present application, and is not intended to represent the only form in which the present application can be implemented. It should be understood that the same or equivalent functions can be accomplished by different embodiments intended to be included in the spirit and scope of the present application.

[0038] See also Figure 1 An embodiment of the present application provides a moving obstacle tracking method, comprising:

[0039] Step S10, obtaining the current video frame output by the vehicle surround view camera.

[0040] Specifically, the vehicle surround view camera detects the vehicle's surrounding environment in real time and outputs the detected video data, wherein the video data consists of a series of continuous video frames, where a video frame refers to a single still image in a video sequence, and the current video frame is the video frame corresponding to the current moment. The video data output by the vehicle surround view camera is sampled at each moment to obtain the corresponding video frame, and the current video frame is the vehicle surround view image, also called a panoramic image. By performing image processing on the current video frame, relevant information of mobile obstacles in the current vehicle's surrounding environment can be obtained. In the method of this embodiment, steps S20 to S60 are performed on each video frame.

[0041] Step S20: Perform moving target detection on the current video frame to obtain at least one moving target area, and extract contour feature points of the at least one moving target area.

[0042] Specifically, after detection, a video frame may contain one or at least two moving target areas, and the moving target area may be the entire or local area of ​​a moving obstacle. A video frame may contain one or at least two moving obstacles. Contour feature points are extracted for each moving target area, and each moving target area contains multiple contour feature points. In this embodiment, it is preferred but not limited to use the Dlouglas-Peucker algorithm to calculate the contour feature points of each moving target area.

[0043] Step S30: Calculate the optical flow vectors of the contour feature points of the at least one moving target area.

[0044] Specifically, in this embodiment, the LK optical flow method is used to process the contour feature points of each moving target area to calculate the optical flow vector of each contour feature point;

[0045] Among them, the LK optical flow method needs to meet the following three conditions:

[0046] Condition (1): the grayscale value or color of the image does not change between adjacent video frames, that is, A(x, y, t) = A(x+d x ,y+d y , t+d t ), where A(x, y, t) is the previous video frame, A(x+d x ,y+d y , t+d t ) is the current video frame, x is the horizontal coordinate of the contour feature point, y is the vertical coordinate of the contour feature point, t is the time, d x is the change in the horizontal axis, d y is the change in the ordinate, d t is the frame difference interval time;

[0047] Condition (2): the movement of obstacles in adjacent video frames is small enough, and the image constraint equation can be rewritten using Taylor's formula: Among them, HOT refers to the higher-order Taylor expansion, whose influence can be ignored, V x is the lateral velocity of the contour feature point, V y is the longitudinal velocity of the contour feature point, a x is the lateral velocity of the contour feature point, a y is the longitudinal velocity of the contour feature point;

[0048] Condition (3): each pixel in the video frame and the pixels in its neighborhood (the adjacent 3*3 area) have similar motions. The points in the neighborhood can be used to establish a set of constraint equations, which is an overdetermined set of constraint equations.

[0049] Based on the above three conditions, the least squares method is used to solve the overdetermined equations to obtain the optical flow vector (V x , V y , a x , a y ).

[0050] Step S40: clustering the optical flow vectors of the contour feature points of the at least one moving target area, and determining the contour feature points included in each moving obstacle according to the clustering result.

[0051] Specifically, for the contour feature points contained in the same moving obstacle, the optical flow vectors of these contour feature points should be consistent. Therefore, clustering can be performed based on the optical flow vectors of the contour feature points to obtain all the contour feature points contained in each moving obstacle.

[0052] Step S50, calculating the image coordinates of each moving obstacle in the current video frame according to the image coordinates of the contour feature points contained in each moving obstacle in the current video frame, and calculating the moving speed of each moving obstacle according to the optical flow vectors of the contour feature points contained in each moving obstacle.

[0053] Step S60: input the image coordinates and moving speed of each moving obstacle in the current video frame into a preset Kalman filter model for filtering, and obtain the filtered image coordinates and moving speed of each moving obstacle in the current video frame.

[0054] For example, after the processing of steps S20 to S50, it can be obtained that the number of moving obstacles in the k-th frame image is N, and the j-th moving obstacle is tracked by using Kalman filtering. The specific steps are as follows:

[0055] According to the kinematic relationship, the state equation is established. For the jth moving obstacle in the kth frame image, its state matrix can be expressed as X k =]x k ,y k , V k , V k ]′, the frame difference interval time t is 1, then its kinematic relationship can be written as:

[0056] X k =FX k-1 +Bw k-1

[0057] Among them, the state transfer matrix B = [0.5 0.51 1]′, w k-1 The system noise satisfies Gaussian distribution and has a mean of 0;

[0058] For the observation model, the position information of the moving obstacle in the kth frame can be used as the observation value X k =

[0059] [x k ,y k ]′, then the observation equation can be written as: k =HX k +v k , where the observation matrix v k In order to satisfy the observation noise of Gaussian distribution, it can be understood as the acceleration at the current moment, and the acceleration obtained by the optical flow method is fitted to obtain its Gaussian distribution.

[0060] The state update process of the Kalman filter in this embodiment is as follows:

[0061] Prediction steps:

[0062] A priori state estimates:

[0063] A priori estimated covariance: Q k-1 is the process noise covariance matrix;

[0064] Update steps:

[0065] Kalman gain: K k =P k|k-1 H T (HP k|k-1 H T +R) -1 , R represents the observation noise covariance matrix;

[0066] Revised estimates:

[0067] Corrected estimated covariance: P k|k =HK k HP k|k-1 ;

[0068] When a moving obstacle is detected, the Kalman filter is used to correct the motion trajectory of the current moving obstacle and update it to the predicted value When a moving obstacle is blocked or missed, the Klarman filter can be used to predict the position of the moving obstacle at the current moment. Output the motion information of moving obstacles to achieve real-time tracking of moving obstacles.

[0069] Based on the above description, it can be seen that the method of this embodiment obtains several moving target areas by performing moving target detection on the video frame output by the vehicle surround view camera, and performs contour feature point detection on the several moving target areas. Since a moving obstacle may contain multiple moving target areas detected previously, the method of this embodiment uses the optical flow vectors of the contour feature points to cluster to determine how many moving obstacles there are in the video frame, as well as all the contour feature points contained in each moving obstacle. Based on the image coordinates and optical flow vectors of all the contour feature points contained in each moving obstacle, the image coordinates and moving speed of the moving obstacle can be quickly calculated. Furthermore, the image coordinates and moving speed of the moving obstacle are used as the model input of the Kalman filter model, and the module is used to perform Kalman filtering, and the model output is used as the final detection result. Compared with the measurement results directly using the sensor, it can better ensure that the position, speed and other information of the moving obstacle will not change suddenly, thereby improving the accuracy of the detection of the moving obstacle.

[0070] In some embodiments, the method further comprises:

[0071] Step S70: convert the filtered image coordinates of each moving obstacle in the current video frame into a vehicle coordinate system and output the result.

[0072] Specifically, the implementation of the vehicle's assisted driving function is mostly based on the vehicle coordinate system. Therefore, in this embodiment, after step S60 obtains the image coordinates of each moving obstacle in the current video frame, the image coordinates will also be converted to the vehicle coordinate system for subsequent calculations.

[0073] Furthermore, the vehicle surround image in this embodiment is captured by a fisheye camera, wherein the viewing angle of the fisheye camera is about 180 degrees, and under the combined effect of the two types of image distortion, the image structure is distorted, and the final image appears circular. Therefore, after the fisheye camera is installed, its various parameters need to be calibrated. The Zhang Dingyou calibration method is used to calculate the camera intrinsic parameter matrix The external parameter matrix N = [R, T], where f is the focal length, C x is the horizontal coordinate of the center of the image, C y is the ordinate of the image center, R is the rotation matrix, T is the translation vector, the camera distortion parameter K can be provided by the camera manufacturer, and the vehicle coordinate system takes the center of the rear axle of the vehicle as the origin. The transformation relationship between the image coordinate system and the vehicle coordinate system can be obtained according to the camera intrinsic parameter matrix, extrinsic parameter matrix, and distortion parameters.

[0074] In some embodiments, the step S20 of performing moving target detection on the current video frame to obtain at least one moving target area specifically includes:

[0075] The video frames of two historical moments adjacent to the current moment output by the vehicle surround view camera are obtained, and based on the three-frame difference method and the video frames of the two historical moments, moving target detection is performed on the current video frame to obtain at least one moving target area.

[0076] Specifically, suppose three consecutive video frames are A i-1 (x,y),A i (x,y),A i+1 (x, y), the three-frame difference method is used to obtain the moving target area of ​​each video frame. First, the difference images of two adjacent frames are obtained respectively, and a threshold T is set to obtain two binary images, a (i-1,i) (x,y) and a (i,i+1) (x, y), secondly, for each pixel point at the corresponding position in the two binary images (that is, the pixel point exists in both binary images), a logical AND operation is performed to obtain the binary image B i-1 (x, y). Third, the connected domains of the binary image are annotated using morphological methods, and the areas with less than 36 pixels are removed to locate the moving target area.

[0077] It should be noted that for moving obstacles such as pedestrians or animals, different parts will present different motion characteristics at the same time. Therefore, the moving target area obtained by the three-frame difference method may be a local area or the entire area of ​​a moving obstacle.

[0078] In some embodiments, the step S30 of clustering the optical flow vectors of the contour feature points of the at least one moving target area specifically includes:

[0079] The optical flow vectors of the contour feature points of the at least one moving target area are clustered using majority clustering, and all contour feature points belonging to the same moving obstacle are clustered in one cluster.

[0080] Specifically, assuming that there are n moving target areas in each video frame, each moving target area has m contour feature points, the mode of the optical flow vectors of all contour feature points is calculated, and all moving target areas are clustered according to the similarity of the mode to obtain several clusters. After clustering, the moving target areas belonging to the same moving obstacle can be screened out, and finally N moving obstacles and the contour feature points of each moving obstacle are output. The specific values ​​of m, n, and N depend on the actual situation.

[0081] In some embodiments, in step S50, the image coordinates of each moving obstacle in the current video frame are specifically: the image coordinates of the geometric center of the contour area formed by the contour feature points contained in each moving obstacle in the current video frame.

[0082] In some embodiments, in step S50, the moving speed of each moving obstacle is specifically: a weighted average of optical flow vectors of contour feature points included in each moving obstacle.

[0083] Specifically, the optical flow vector of each moving obstacle’s contour feature point is expressed as (V x , V y , a x , a y ), the moving speed of each moving obstacle is specifically: the contour feature points contained in each moving obstacle (V x , V y ) weighted average of the contour feature points; for each moving obstacle, the weighted average of the optical flow vectors of all its contour feature points is calculated, and different weights can be given according to the quality or importance of the contour feature points, for example, high-quality feature points have higher weights.

[0084] In some embodiments, the step S10 further includes:

[0085] Grayscale processing is performed on the current video frame to obtain a grayscale image, and median filtering is performed on the grayscale image. The filtered grayscale image is used for moving target detection to obtain at least one moving target area.

[0086] Specifically, grayscale processing and median filtering are to reduce or eliminate interference factors in video frames, mainly including grayscale and image enhancement. First, the video frame is grayscaled using the mean method, and the grayscale of the processed video frame is gray = (double(B)+double(G)+double(R)) / 3; on this basis, the median filtering technology in nonlinear spatial domain filtering is used to process the video frame. The median filtering can retain the edge features of the video frame to the maximum extent, ensure clear edges and eliminate salt and pepper noise in the video frame.

[0087] Corresponding to the method of the above embodiment, see Figure 2 Another embodiment of the present application further provides a mobile obstacle tracking device, comprising:

[0088] A video frame acquisition module, used to acquire a current video frame output by the vehicle surround view camera; specifically, the video frame acquisition module is also used to perform grayscale processing on the current video frame to obtain a grayscale image, and perform median filtering on the grayscale image, and output the filtered grayscale image to the target detection module for performing moving target detection to obtain at least one moving target area;

[0089] The target detection module is used to perform moving target detection on the current video frame to obtain at least one moving target area, and extract contour feature points of the at least one moving target area; specifically, by acquiring video frames at two historical moments adjacent to the current moment output by the vehicle surround view camera, performing moving target detection on the current video frame based on the three-frame difference method and the video frames at the two historical moments to obtain at least one moving target area;

[0090] An optical flow calculation module, used to calculate the optical flow vectors of the contour feature points of the at least one moving target area;

[0091] A clustering module, configured to cluster the optical flow vectors of the contour feature points of the at least one moving target area, and determine the contour feature points contained in each moving obstacle according to the clustering result; specifically, clustering the optical flow vectors of the contour feature points of the at least one moving target area by adopting majority clustering, and clustering all the contour feature points belonging to the same moving obstacle into one cluster;

[0092] A motion information calculation module, used to calculate the image coordinates of each moving obstacle in the current video frame according to the image coordinates of the contour feature points contained in each moving obstacle in the current video frame, and calculate the moving speed of each moving obstacle according to the optical flow vectors of the contour feature points contained in each moving obstacle; the image coordinates of each moving obstacle in the current video frame are specifically: the image coordinates of the geometric center of the contour area formed by the contour feature points contained in each moving obstacle in the current video frame, and the moving speed of each moving obstacle is specifically: the weighted average value of the optical flow vectors of the contour feature points contained in each moving obstacle;

[0093] The Kalman filter module is used to input the image coordinates and moving speed of each moving obstacle in the current video frame into a preset Kalman filter model for filtering processing to obtain the image coordinates and moving speed of each moving obstacle in the current video frame after filtering.

[0094] In some embodiments, the apparatus further comprises:

[0095] The coordinate conversion module is used to convert the filtered image coordinates of each moving obstacle in the current video frame into a vehicle coordinate system and output it.

[0096] It should be noted that the mobile obstacle tracking device of this embodiment corresponds to the mobile obstacle tracking method of the above embodiment. Therefore, the parts of the mobile obstacle tracking device of this embodiment that are not described in detail can be obtained by referring to the contents of the mobile obstacle tracking method of the above embodiment, and will not be repeated here.

[0097] Furthermore, if the mobile obstacle tracking device of the above embodiment is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0098] Another embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the mobile obstacle tracking method as described in the above embodiment is implemented.

[0099] Specifically, the computer-readable storage medium may include: any entity or recording medium that can carry the computer program instructions, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0100] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for tracking a moving obstacle, characterized in that: The method comprises: Get the current video frame output by the vehicle's surround view camera; Performing moving target detection on the current video frame to obtain at least one moving target area, and extracting contour feature points of the at least one moving target area; Calculate the optical flow vector of the contour feature points of the at least one moving target area; Clustering the optical flow vectors of the contour feature points of the at least one moving target area, and determining the contour feature points contained in each moving obstacle according to the clustering result; Calculating the image coordinates of each moving obstacle in the current video frame according to the image coordinates of the contour feature points contained in each moving obstacle in the current video frame, and calculating the moving speed of each moving obstacle according to the optical flow vector of the contour feature points contained in each moving obstacle; The image coordinates and moving speed of each moving obstacle in the current video frame are input into a preset Kalman filter model for filtering processing to obtain the filtered image coordinates and moving speed of each moving obstacle in the current video frame.

2. The method according to claim 1, characterized in that The method further comprises: The filtered image coordinates of each moving obstacle in the current video frame are converted to a vehicle coordinate system and output.

3. The method according to claim 1 or 2, characterized in that: The performing moving target detection on the current video frame to obtain at least one moving target area specifically includes: The video frames of two historical moments adjacent to the current moment output by the vehicle surround view camera are obtained, and based on the three-frame difference method and the video frames of the two historical moments, moving target detection is performed on the current video frame to obtain at least one moving target area.

4. The method according to claim 1 or 2, characterized in that: The clustering of the optical flow vectors of the contour feature points of the at least one moving target area specifically includes: The optical flow vectors of the contour feature points of the at least one moving target area are clustered using majority clustering, and all contour feature points belonging to the same moving obstacle are clustered in one cluster.

5. The method according to claim 1 or 2, characterized in that: The image coordinates of each moving obstacle in the current video frame are specifically: the image coordinates of the geometric center of the contour area formed by the contour feature points contained in each moving obstacle in the current video frame.

6. The method according to claim 1 or 2, characterized in that: The moving speed of each moving obstacle is specifically: a weighted average value of optical flow vectors of contour feature points included in each moving obstacle.

7. The method according to claim 1 or 2, characterized in that: The method further comprises: Grayscale processing is performed on the current video frame to obtain a grayscale image, and median filtering is performed on the grayscale image. The filtered grayscale image is used for moving target detection to obtain at least one moving target area.

8. A mobile obstacle tracking device, characterized in that: include: A video frame acquisition module is used to acquire the current video frame output by the vehicle's surround view camera; A target detection module, configured to perform moving target detection on the current video frame to obtain at least one moving target area, and extract contour feature points of the at least one moving target area; An optical flow calculation module, used to calculate the optical flow vectors of the contour feature points of the at least one moving target area; A clustering module, used to cluster the optical flow vectors of the contour feature points of the at least one moving target area, and determine the contour feature points contained in each moving obstacle according to the clustering result; A motion information calculation module, used to calculate the image coordinates of each moving obstacle in the current video frame according to the image coordinates of the contour feature points contained in each moving obstacle in the current video frame, and calculate the moving speed of each moving obstacle according to the optical flow vector of the contour feature points contained in each moving obstacle; The Kalman filter module is used to input the image coordinates and moving speed of each moving obstacle in the current video frame into a preset Kalman filter model for filtering processing to obtain the image coordinates and moving speed of each moving obstacle in the current video frame after filtering.

9. The device according to claim 8, characterized in that The device also includes: The coordinate conversion module is used to convert the filtered image coordinates of each moving obstacle in the current video frame into a vehicle coordinate system and output it.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the moving obstacle tracking method according to any one of claims 1 to 7 is implemented.

Images