Research On Parking System For Full Video Parking-lot

Aiming at the problem that the current parking management system is not real-time,accurate and intelligent,this paper proposed a parking detection system based on video image processing.The system is mainly composed of two parts: the available vehicle number detection system and the parking space status detection system.The available vehicle number detection system is based on TMS320DM642.The Gaussian model improved background difference method is used for vehicle detection,and the virtual detection belt is used for vehicle entry and exit judgment to count the number of available parking spaces in the closed parking lot.After the simulation experiment,the success rate of the test can reach 100%.The parking space condition detection system relies on the fisheye lens to monitor the parking space status.The video image of the fisheye lens is distorted,so the distortion correction is performed before the image processing.This paper proposed a fisheye image correction algorithm based on the radiance expansion.In order to solve the problem of the large workload of the traditional manual marking parking space method,a combination of Hough line detection and corner point detection was proposed,and the specific coordinates of the parking space endpoint were detected from the video image.Through experiments,the accuracy and rate of HOG+SVM,R-CNN,Fast R-CNN,YOLO and other vehicle identification methods were compared.This paper selected YOLO to identify vehicles.The moving object in the video is detected by a combination of Gaussian background modeling and background difference.When a moving object appears in the video,the frame image is identified,and the vehicle or non-vehicle can be identified.When the vehicle stops moving,the coordinates of the detection frame of the output are combined with the coordinates of the end point of the parking space to determine in which parking space the coordinate of the center point of the detection frame is located,thereby realizing real-time detection and display of the full-empty state of the parking space.After testing,the correct rate of detection can reach more than 92%.