CN-117079261-B - Campus vehicle cross-domain tracking method, system, equipment and storage medium

CN117079261BCN 117079261 BCN117079261 BCN 117079261BCN-117079261-B

Abstract

The invention discloses a campus vehicle cross-domain tracking method, a system, equipment and a storage medium, which relate to the technical field of image detection and identification and comprise the steps of acquiring a vehicle image acquired by a numbered camera, detecting a vehicle image, performing license plate detection, performing inclination correction on the obtained license plate image, performing super-resolution reconstruction on the corrected license plate image, identifying to obtain a license plate character sequence, constructing a vehicle information set on the license plate character sequence, corresponding image acquisition time and camera number, screening the license plate character sequence matched with a license plate of a target vehicle in a specified time period, and determining a target vehicle running track according to the camera number and the image acquisition time corresponding to the matched license plate character sequence. The manual supervision pressure is reduced, and the detection and processing efficiency of traffic events in the campus is improved.

Inventors

ZHENG LAIBO
YU CHANG
WANG ZHENGFAN
LIU XING
LI LINYAN
CHEN MIAOMIAO

Assignees

山东大学

Dates

Publication Date: 20260508
Application Date: 20230904

Claims (9)

1. The campus vehicle cross-domain tracking method is characterized by comprising the following steps of: acquiring a vehicle image acquired by a numbered camera; Performing license plate detection on a vehicle image, performing inclination correction on the obtained license plate image, and recognizing the corrected license plate image after super-resolution reconstruction to obtain a license plate character sequence, wherein the inclination angle in the vertical direction and the inclination angle in the horizontal direction are estimated according to the gradient information of the license plate image so as to perform inclination correction; constructing a vehicle information set for the license plate character sequence, the corresponding image acquisition time and the corresponding camera number; screening license plate character sequences matched with license plates of the target vehicles in a specified time period in the vehicle information set, determining the positions of the target vehicles according to camera numbers corresponding to the matched license plate character sequences, and determining the driving tracks of the target vehicles by combining corresponding image acquisition time; Wherein, the process of estimating the vertical direction inclination angle and the horizontal direction inclination angle includes: dividing a license plate image into a plurality of non-overlapping sub-blocks, and calculating a characteristic value and a corresponding characteristic vector of each sub-block image according to image gradients; Traversing the sampling points, calculating the angles of the feature vectors at the sampling points, and adding the angles into an angle statistics array; According to the preset angle range, the angle with the largest frequency is selected from the angle statistics array to be used as the vertical direction inclination angle and the horizontal direction inclination angle.
2. The campus vehicle cross-domain tracking method of claim 1, wherein a gradient vector in a local neighborhood of a pixel point is calculated, a covariance matrix is calculated based on the gradient vector, and a characteristic value and a corresponding characteristic vector are obtained by decomposing a characteristic value of the covariance matrix, wherein the characteristic vector represents a gradient change direction, and the characteristic value represents a change strength in a specific direction.
3. The campus vehicle cross-domain tracking method of claim 1 wherein the vertical angle range is 30 degrees to 150 degrees, and the vertical inclination angle is determined by counting the angle with the largest number of occurrences in the angle statistics array within the vertical angle range.
4. The campus vehicle cross-domain tracking method of claim 1 wherein the horizontal direction angle range includes a horizontal left direction angle range and a horizontal right direction angle range, which are respectively 0 to 30 degrees and 150 degrees to 180 degrees; Determining a horizontal left-direction inclination angle X1 by counting the angle with the largest occurrence number in the angle counting array in the horizontal left-direction angle range; Determining a horizontal right inclination angle X2 by counting the angle with the largest occurrence number in the angle counting array in the horizontal right angle range; if X1 is larger than 180-X2, taking X1 as a horizontal direction inclination angle, otherwise taking X2 as a horizontal direction inclination angle.
5. The method of cross-domain tracking of campus vehicles of claim 1, wherein the corrected license plate images are super-resolution reconstructed by adopting an improved super-resolution generation countermeasure network model, wherein the improved super-resolution generation countermeasure network model removes batch normalization layers in residual blocks, and a 1X 1 convolution layer is added after an activation function of the residual blocks and before residual connection.
6. The campus vehicle cross-domain tracking method of claim 1, wherein searching is performed in a vehicle information set according to a specified time period, license plate character sequences in the specified time period are screened, license plate character sequences, in which 5 or more than 5 characters in 7 or 8 characters are completely consistent with a target vehicle, are continuously searched in the screened license plate character sequences, corresponding image acquisition time and camera numbers are determined, and the camera numbers are numbered according to the position where a camera is located and the orientation of the camera.
7. A campus vehicle cross-domain tracking system, comprising: an acquisition module configured to acquire a vehicle image acquired by the numbered camera; the license plate recognition module is configured to detect a license plate of a vehicle image, perform inclination correction on the obtained license plate image, and recognize the corrected license plate image after super-resolution reconstruction to obtain a license plate character sequence, wherein the inclination angle in the vertical direction and the inclination angle in the horizontal direction are estimated according to gradient information of the license plate image so as to perform inclination correction; Wherein, the process of estimating the vertical direction inclination angle and the horizontal direction inclination angle includes: dividing a license plate image into a plurality of non-overlapping sub-blocks, and calculating a characteristic value and a corresponding characteristic vector of each sub-block image according to image gradients; Traversing the sampling points, calculating the angles of the feature vectors at the sampling points, and adding the angles into an angle statistics array; according to a preset angle range, selecting the angle with the largest frequency from the angle statistics array as a vertical direction inclination angle and a horizontal direction inclination angle; The license plate storage module is configured to construct a vehicle information set for the license plate character sequence, the corresponding image acquisition time and the corresponding camera number; the track tracking module is configured to screen license plate character sequences matched with license plates of the target vehicles in a specified time period in the vehicle information set, determine the positions of the target vehicles according to camera numbers corresponding to the matched license plate character sequences, and determine the running track of the target vehicles in combination with corresponding image acquisition time.
8. An electronic device comprising a memory and a processor and computer instructions stored on the memory and running on the processor, which when executed by the processor, perform the method of any one of claims 1-6.
9. A computer readable storage medium storing computer instructions which, when executed by a processor, perform the method of any of claims 1-6.

Description

Campus vehicle cross-domain tracking method, system, equipment and storage medium Technical Field The invention relates to the technical field of image detection and identification, in particular to a campus vehicle cross-domain tracking method, a system, equipment and a storage medium. Background The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art. With the continuous expansion of the high school rule models, the number of students, coaches and visitor vehicles is gradually increased, and the traffic management in campuses is faced with more serious challenges. In particular, during peak hours of rapid increases in traffic flow in campuses, traditional vehicle management methods, such as manual patrol and manual recording, have been difficult to meet the requirements for accurate identification and real-time tracking of large-scale vehicles. Meanwhile, the number of cameras in the campus is numerous, but monitoring management staff is relatively scarce, so that monitoring resources are not matched with management requirements, and real-time monitoring of all traffic problems is difficult to achieve. The vehicle cross-domain tracking technology can realize seamless tracking of vehicles in different areas and different cameras in the campus, can overcome the problem of insufficient human resources, and can monitor traffic conditions in the campus in a more intelligent and efficient manner, so that accurate identification of the vehicles is realized and a history track of the vehicles in a period of time is drawn. The cross-domain tracking of vehicles in a campus scene relates to target identification in different time periods and different campus areas, and the problems of visual angle change, illumination change, shielding, appearance change and the like need to be overcome. The local area of the license plate image in the real campus traffic scene can be influenced by interference factors such as illumination change and shielding, and due to factors such as installation positions and vehicle vibration, the license plate can be inclined in the horizontal direction and can also be rotated and distorted, and different inclination degrees of the license plate image can exist in different areas, so that the problems are not considered in the prior art. Disclosure of Invention In order to solve the problems, the invention provides a cross-domain tracking method, a system, equipment and a storage medium for campus vehicles, which are used for solving the problem of reduced license plate recognition accuracy caused by visual angle change by adopting a license plate inclination correction algorithm based on a direction field for an extracted license plate image, reducing the influence of illumination change and low image resolution on the license plate recognition accuracy through super-resolution reconstruction, and drawing a historical track of a target vehicle in a period of time by combining a license plate character sequence recognition result, corresponding image acquisition time and a camera number, thereby reducing manual supervision pressure and improving the detection and processing efficiency of traffic events in the campus. In order to achieve the above purpose, the present invention adopts the following technical scheme: In a first aspect, the present invention provides a campus vehicle cross-domain tracking method, including: acquiring a vehicle image acquired by a numbered camera; Performing license plate detection on a vehicle image, performing inclination correction on the obtained license plate image, and recognizing the corrected license plate image after super-resolution reconstruction to obtain a license plate character sequence, wherein the inclination angle in the vertical direction and the inclination angle in the horizontal direction are estimated according to the gradient information of the license plate image so as to perform inclination correction; constructing a vehicle information set for the license plate character sequence, the corresponding image acquisition time and the corresponding camera number; And screening license plate character sequences matched with license plates of the target vehicles in a specified time period in the vehicle information set, determining the positions of the target vehicles according to camera numbers corresponding to the matched license plate character sequences, and determining the driving tracks of the target vehicles by combining corresponding image acquisition time. Alternatively, the estimating the vertical tilt angle and the horizontal tilt angle includes: dividing a license plate image into a plurality of non-overlapping sub-blocks, and calculating a characteristic value and a corresponding characteristic vector of each sub-block image according to image gradients; Traversing the sampling points, calculating the angles of the feature vectors at the sam