| Because the existing infrastructure is not perfect,and many car owners have poor awareness of parking standards,irregular parking occurs from time to time.It will not only waste parking space resources,but also increase the management cost of parking lots.Seriously it will affect the smooth flow of traffic,causing traffic jams and inducing traffic accidents-However,the existing parking supervision mainly relies on manual management,which has randomness and hysteresis,and consumes a lot of manpower and material resources.With the popularity of video sensors in cities,the acquisition of surveillance videos is becoming easier and easier.It is a feasible way to solve the problem of irregular parking through the parking behavior recognition technology of surveillance video.Compared with the traditional methods,video-based parking detection and recognition method has the following advantages:Video contains abundant spatio-temporal information and attribute information of the target;with the popularity of surveillance cameras,the acquisition of surveillance video becomes more and more convenient;Camera covers a wide range of area,a single camera can simultaneously detect multiple parking spaces.It has monitoring function to facilitate investigation and evidence collection.It can also facilitate the expansion of parking records,vehicle classification,license plate recognition and other functions.This paper studies the vehicle pose analysis based on surveillance video,and makes use of GIS technology,computer vision technology and other technologies to realize accurate identification of irregular parking behavior by processing surveillance video.On one hand,the irregular parking behavior can be dealt and corrected in time to realize the intelligent and standardized management of roadside parking lots and large parking lots,and to improve the efficiency of parking management;on the other hand,it can improve the utilization rate of parking resources,establish a good parking order,solve the practical problems of parking difficulties and parking chaos,and guarantee the safety of urban public transportation.Based on the above research background and significance,this paper takes the vehicle target in the parking surveillance video as the research object,mainly studies the classification rules,parking feature extraction methods and recognition algorithms for irregular parking behavior.The research results are as follows:(1)A classification rule base of irregular parking behavior is constructed.This paper classifies parking lots and parking spaces,analyzes and summarizes the existing local provisions,and builds a systematic rules base for irregular parking behavior based on parking performance and parking characteristics according to the fact of lacking a unified standard for the current behavioral discrimination against irregular parking(2)A parking feature extraction method based on spatialization of video scenes is developed.Vibe’s moving target detection algorithm is improved.The HSV shadow removal method is added.The deep learning SSD network model is used for vehicle detection,Vehicle detection based on deep learning SSD network model solves the problem that the traditional methods based on simple geometric features of vehicles and SVM classification are not ideal.The vehicle trajectory is obtained by CSRT vehicle tracking method,and the vehicle parking judgment is realized by using the speed threshold and the time threshold.Vehicle feature extraction in traffic incident detection is often based on image space,while surveillance video is the result of perspective transformation in the real world,which has the imaging characteristics of near-large and far-small.It causes a certain degree of interference to traffic incident analysis.This paper establishes the mapping relationship between parking scenery and surveillance video,and obtains the real pose characteristics of vehicle parking status.(3)An irregular parking behavior recognition algorithm is proposed to convert the parking behavior recognition problem into the geometric relationship between parking characteristics and parking spaces.The identification algorithm of irregularparking behavior is proposed to achieve accurate and effective detection of various complex and irregular parking behaviors.(4)An irregular parking behavior recognition system is designed and implemented.Based on the above research results,the functions of vehicle detection and tracking,parking feature extraction and irregular parking behavior recognition are integrated to realize automatic parking behavior recognition,which overcomes the randomness and hysteresis of traditional eye-based identification and expands the application scope of intelligent parking. |
