Outdoor parking lot unoccupied parking space detection method based on deep learning

Abstract

The invention discloses an outdoor parking lot unoccupied parking space detection method based on deep learning, and the method comprises the steps that a picture is collected through a high-altitude camera located above a parking lot, a PSPNet network model is improved for extracting a parking space line in the image, and the linear detection and contour extraction are carried out, the positions of all parking spaces in the image are obtained in combination with a corresponding parking space line filling method, a YOLOv3 network model is improved to detect all vehicles in the image, the positions of the vehicles are recorded, the parking spaces and the vehicles are subjected to Hungarian matching, and whether the parking spaces are occupied is judged according to whether the parking spaces are successfully matched and the distance between the two successfully matched parties; and the number of unoccupied parking spaces in the image is counted. The number and the positions of the unoccupied parking spaces of the outdoor parking lot can be accurately judged, the robustness is high, and a vehicle owner can conveniently find the parking spaces for parking.

Description

technical field [0001] The invention relates to the technical field of detecting vacant parking spaces in outdoor parking lots, in particular to a method for detecting vacant parking spaces in outdoor parking lots based on deep learning. Background technique [0002] Nowadays, with the improvement of living standards, the use of vehicles is becoming more and more extensive, and the transportation is becoming more and more convenient. Traveling by car has become the first mode of transportation for people, especially for densely populated areas such as stations and shopping malls, it is usually necessary to get off the car and walk on foot, so The car must be parked in the parking lot. If you do not know whether there is a vacant space in the parking lot and enter directly, it will inevitably waste a lot of time. However, most of the existing parking space statistics methods are set up at the entrance, and the entering vehicles are detected and counted one by one. Compared wi...

AI Technical Summary

Problems solved by technology

[0002] Nowadays, with the improvement of living standards, the use of vehicles is becoming more and more extensive, and the transportation is becoming more and more convenient. Traveling by car has become the first mode of transportation for people, especially for densely populated areas such as stations and shopping malls, it is usually necessary to get off the car and walk on foot, so It is necessary to park the car in the parking lot. If you do not know whether there is a vacant space in the parking lot and enter directly, it will inevitably waste a lot of time. However, most of the existing parking space statistics methods are set up at the entrance, and the entering vehicles are detected and counted one by one, and the realization of the design is carried out. Compared with the determined total number of parking spaces to determine the number of vacant parking spaces, the cost of detection equipment is high, and only the number of vacant parking spaces can be obtained, but the location information of vacant parking spaces cannot be obtained, and the owner needs to search one by one, which is inefficient

However, some existing methods of using high-altitude cameras to detect vacant parking spaces need to manually calibrate the position information of each parking space in the image, and detect the status of each parking space one by one. If the position or angle of the camera changes, it needs to be re-calibrated, which is more cumbersome. , poor adaptability

Images