Research On Detection And Recognition Of Vehicle In Open Parking Lots

Along with the rapid development of economy,the amount of motor vehicles has a rapid growth in China.As a result,the difficulty of parking has perplexed many cities in recent years.The most important reason is the shortage of parking resource,followed by management inefficiencies.Open parking lots management needs to improve management efficiency necessarily.This paper studies the vehicle detection and recognition,in order to improve the utilization efficiency of open-air car park.In this paper,the work of vehicle detection and recognition in the open car park is divided into two parts,based on motion information fast moving targets and using the strong classifiers to recognize target.The two parts complement each other.The main tasks are follows:I.Using improved ViBe algorithm which combines three frame difference algorithm to detect moving object in open parking lots.The ViBe algorithm is a pop algorithm for foreground detection in recent years.ViBe algorithm has the characteristics of good realtime performance,low computation,high speed.However,ViBe algorithm can neither remove the "ghost" region nor inhibit target shadow,these issues affect the algorithm accuracy and reliability.So we put forward some improvement as follows:1.Using HSV color space brightness information to suppress moving objects shadows.2.The combination of ViBe algorithm and three frame difference algorithm builds a moving object mask,and the operation between masks can determine the "ghost" region.The updated ViBe algorithm initialization model can quickly eliminate the "ghost" region.II.Realization of vehicle identification based on HOG feature and SVM classifier.To improve the identify accuracy of support vector machine,this paper analysis the gesture of target and judges its angle of inclination between the vehicle and camera,then adjusts the image of target to get various proportion of the target and extracts the HOG feature.Finally,the HOG feature of the target image is used to identify the vehicle through the matching SVM.This paper separates the region of the moving object by detect target in the surveillance video.According to the different gesture of the moving object,using suitable support vector machine to identify the vehicles.Based on the above operational principle,we realize the vehicle detection and identification design in the open parking lots.In the experiment,the accuracy of vehicle identification that in open-air parking lot can reach 91.5%.The algorithm in the paper can effectively monitor the vehicles enough for open-air parking.