CN-121982338-A - Parking detection method, parking detection device and computer storage medium

CN121982338ACN 121982338 ACN121982338 ACN 121982338ACN-121982338-A

Abstract

The application provides a parking detection method, a parking detection device and a computer storage medium. The method comprises the steps of determining a calibrated berth area and an interested area, determining the quality score of each video frame, obtaining a detection frame of each detection target in the video frame, determining candidate targets in the interested area according to the coincidence condition of the detection frame and the interested area, dividing the candidate targets into different area matrixes according to the located areas, determining the matrix scores of the area matrixes respectively according to the candidate targets in the area matrixes, integrating the matrix scores of all the area matrixes, determining the quality score of the video frame, selecting the video frame with the highest quality score as a target video frame, and detecting the parking behavior of the vehicle aiming at the target video frame. By the parking detection method, the optimal frame for parking behavior detection is selected, so that the equipment calculation power consumption is reduced, and the detection accuracy is improved.

Inventors

ZHAO HONGYU
WU YUN
HU WANGBIN
LIU BINLONG
XU FANGYUAN

Assignees

浙江大华技术股份有限公司

Dates

Publication Date: 20260505
Application Date: 20251208

Claims (11)

1. A parking detection method, characterized in that the parking detection method comprises: Determining a calibrated berth area and a region of interest; The quality score for each video frame is determined by: acquiring a detection frame of each detection target in the video frame; according to the coincidence condition of the detection frame and the region of interest, determining candidate targets in the region of interest; Dividing the candidate targets into different area matrixes according to the areas; determining matrix scores of the area matrix according to candidate targets in the area matrix respectively; Synthesizing matrix scores of all area matrixes, and determining quality scores of the video frames; Selecting the video frame with the highest quality score as a target video frame; a parking behavior of the vehicle is detected for the target video frame.
2. The parking detection method according to claim 1, wherein, The determining a candidate target in the region of interest according to the coincidence condition of the detection frame and the region of interest comprises the following steps: Determining the overlapping area of a detection frame of each detection target and the region of interest; Determining the overlapping proportion of the detection targets according to the overlapping area and the area of the detection frame; And determining the detection target with the overlapping proportion higher than a preset overlapping threshold value as a candidate target in the region of interest.
3. The parking detection method according to claim 2, wherein, The area matrix comprises a berth area matrix, a horizontal suspicious area matrix and a vertical suspicious area matrix; The dividing the candidate targets into different area matrixes according to the areas, including: and dividing the candidate target into corresponding area matrixes according to the area where the overlapping area is.
4. The parking detection method according to claim 1, wherein, The determining the matrix score of the area matrix according to the candidate targets in the area matrix respectively comprises the following steps: Determining the target type of a candidate target in the berth area matrix; Determining position information of the candidate target; determining a weight parameter of the candidate target according to the target type and the position information; And determining the matrix score of the berth area matrix according to the weight parameters of all candidate targets in the berth area matrix.
5. The method for detecting a parking according to claim 4, wherein, The determining the location information of the candidate target includes: determining a maximum distance according to the resolution of the video frame; Determining a pixel distance from a detection frame center point of the candidate target to the bottom of the video frame; and determining the position information of the candidate target according to the maximum distance and the pixel distance.
6. The parking detection method according to claim 1, wherein, The determining the matrix score of the area matrix according to the candidate targets in the area matrix respectively comprises the following steps: determining the detection type of a candidate target in the suspicious region matrix; Determining the candidate target of which the detection type is vehicle as a vehicle target; Determining a position parameter according to the historical change condition of the distance between the center of the detection frame of the vehicle target and the center of the berth; determining a distance parameter according to the distance between the center of the detection frame of the vehicle target and the berth boundary; determining a driving direction vector according to the position change condition of the vehicle target; Determining an angle parameter according to the included angle between the running direction vector and the normal vector of the road surface; determining a weight parameter of the vehicle target according to the position parameter, the distance parameter and the angle parameter; and determining the matrix score of the suspicious berth area matrix according to the weight parameters of all candidate targets in the suspicious area matrix.
7. The parking detection method according to claim 6, wherein, The suspicious region matrix comprises a horizontal suspicious region matrix and a vertical suspicious region matrix, wherein the horizontal suspicious region and the vertical suspicious region are regions except the berth region in the region of interest; The region of interest is obtained by expanding each side of the calibrated berthing region.
8. The parking detection method according to claim 7, wherein, The determining the angle parameter according to the included angle between the driving direction vector and the normal vector of the road surface comprises the following steps: Determining an angle interval in which the included angle is positioned for the vehicle target of the horizontal suspicious region matrix, wherein the first angle interval is smaller than the second angle interval; when the included angle is located in a first angle interval, determining an angle parameter of the included angle according to a positive correlation; when the included angle is located in the second angle interval, determining an angle parameter of the included angle according to a negative correlation; And/or the number of the groups of groups, The determining the angle parameter according to the included angle between the driving direction vector and the normal vector of the road surface comprises the following steps: and determining the angle parameter of the included angle according to the negative correlation relation for the vehicle target of the vertical suspicious region matrix.
9. The parking detection method according to claim 6, wherein, The determining the weight parameter of the vehicle target according to the position parameter, the distance parameter and the angle parameter comprises the following steps: Determining a weight parameter of the vehicle target according to the position parameter, the distance parameter, the angle parameter and the license plate parameter; the size of the license plate parameter is positively correlated with the size of the license plate detection integrity.
10. A parking detection device, comprising a memory and a processor coupled to the memory; wherein the memory is for storing program data and the processor is for executing the program data to implement the park detection method as claimed in any one of claims 1 to 9.
11. A computer storage medium for storing program data which, when executed by a computer, is adapted to carry out the park detection method as claimed in any one of claims 1 to 9.

Description

Parking detection method, parking detection device and computer storage medium Technical Field The application relates to the technical field of intelligent analysis and processing, in particular to a parking detection method, a parking detection device and a computer storage medium. Background As the total number of motor vehicles in towns increases, the contradiction between the supply and demand of berths and vehicles is also increasing. Along with the rapid development of the intelligent Internet of things technology, the management of road side parking gradually becomes an important module of the intelligent city Internet of things. By constructing intelligent camera points at the high position of the intersection, collecting YUV data and carrying out algorithm identification analysis, and uploading parking information to the platform in real time, the intelligent traffic management system can be realized. However, in the current scheme for analyzing the behavior of vehicles entering and exiting from berths, in order to ensure the accuracy of parking behavior determination and the integrity of detection information, a complete process of parking detection is required for all video frames, so that the calculation power consumption of equipment for deploying a detection algorithm is excessive, and the efficiency of parking detection is reduced. Disclosure of Invention In order to solve the technical problems, the application provides a parking detection method, a parking detection device and a computer storage medium. In order to solve the technical problems, the application provides a parking detection method, which comprises the following steps: Determining a calibrated berth area and a region of interest; The quality score for each video frame is determined by: acquiring a detection frame of each detection target in the video frame; according to the coincidence condition of the detection frame and the region of interest, determining candidate targets in the region of interest; Dividing the candidate targets into different area matrixes according to the areas; determining matrix scores of the area matrix according to candidate targets in the area matrix respectively; Synthesizing matrix scores of all area matrixes, and determining quality scores of the video frames; Selecting the video frame with the highest quality score as a target video frame; a parking behavior of the vehicle is detected for the target video frame. Wherein, according to the coincidence condition of the detection frame and the region of interest, determining the candidate target in the region of interest comprises: Determining the overlapping area of a detection frame of each detection target and the region of interest; Determining the overlapping proportion of the detection targets according to the overlapping area and the area of the detection frame; And determining the detection target with the overlapping proportion higher than a preset overlapping threshold value as a candidate target in the region of interest. The regional matrix comprises a berth regional matrix, a horizontal suspicious regional matrix and a vertical suspicious regional matrix; The dividing the candidate targets into different area matrixes according to the areas, including: and dividing the candidate target into corresponding area matrixes according to the area where the overlapping area is. Wherein the determining the matrix score of the area matrix according to the candidate objects in the area matrix respectively includes: Determining the target type of a candidate target in the berth area matrix; Determining position information of the candidate target; determining a weight parameter of the candidate target according to the target type and the position information; And determining the matrix score of the berth area matrix according to the weight parameters of all candidate targets in the berth area matrix. Wherein the determining the location information of the candidate object includes: determining a maximum distance according to the resolution of the video frame; Determining a pixel distance from a detection frame center point of the candidate target to the bottom of the video frame; and determining the position information of the candidate target according to the maximum distance and the pixel distance. Wherein the determining the matrix score of the area matrix according to the candidate objects in the area matrix respectively includes: determining the detection type of a candidate target in the suspicious region matrix; Determining the candidate target of which the detection type is vehicle as a vehicle target; Determining a position parameter according to the historical change condition of the distance between the center of the detection frame of the vehicle target and the center of the berth; determining a distance parameter according to the distance between the center of the detection frame of the vehicle target and the berth boundary; determinin