Research On The Key Technologies Of Detection For Vehicles Intersection Staying Based On Video Information

Along with the growth of car ownership,the problem of traffic congestion is increasingly prominent.It seriously affects the people's daily travel and restricts the development of social economy.Although our country invested a lot of manpower and material resources in the solution of traffic congestion each year,the problem is still serious.In our country congestion traffic situation,a common and serious phenomenon is the vehicle intersection detention phenomenon,this phenomenon mainly refers to at the junction exit has already occurred congestion congestion,the successor vehicle still continues to drive into the intersection caused by a traffic overflow phenomenon.Stranded at the intersection of vehicles will sometimes hinder the normal traffic inother directions,aggravating congestion,if not in time to evacuate,the phenomenon is very easy to spread to other adjacent intersection caused traffic lattice lock,resulting in the entire road network traffic paralysis.This paper based on the actual intersection surveillance video,based on the summary of domestic and foreign research status,from the vehicle intersection detention phenomenon prevention,the detention vehicle detection,the detention information extraction and upload as well as the vehicle intersection detention information system establishment four aspects to carry on the research,the main contents are as follows:(1)This paper studies the mechanism of traffic stop at intersections,puts forward a control method to prevent traffic stop at intersections,detects whether queuing vehicles are about to block intersections by setting up a sense of location detector in the lower section of intersections,and uses the electronic warning screen in intersections to remind traffic entering intersections and to prevent the occurrence of traffic stop phenomenon at intersections.(2)For stranded vehicles which do not follow the warning screen instruction to enter the intersection by force and cause traffic overflow,according to the characteristics of the process of vehicle holdup phenomenon,this paper proposes a detection algorithm of intersections holdup vehicles based on video virtual detector group,sets up virtual detector group by using intersections surveillance video information,detects the presence and holdup of vehicles in intersections by processing the brightness information,and uploads the detected phenomenon to the server to further process.(3)In the process of extracting the information records of the stranded vehicles in the video segment,this paper analyzes the phenomenon of the vehicle license plate being blocked by each other in the case of the multiple vehicles ' detention,and then determines the area that the vehicle will not be occluded in the video according to the characteristics of the video and the traffic characteristics The vehicle license plate in the video section is extracted in the area,and the car that is extracted to the license plate is taken as the vehicle which is stranded at the junction,finally the information of the vehicle plate which conforms to the traffic requirement of the intersection is generated to upload the detention information system database.This paper mainly studies the extraction method of the license plate region in the video image of the intersection surveillance,improves a method of the license plate region image extraction based on the candidate area selection of the license plate,this method first uses the selective search algorithm to extract the candidate area of the license plate of the video image.Based on the principle of extracting target candidate region from selective search,this paper improves the algorithm's segmentation and extraction speed of the candidate area image by treating the image with a certain degree of Gauss Blur,compared with the original method,The improvement measures in this paper can improve the processing speed by twice times under the condition that the accuracy rate is similar.Finally,this paper designs and trains a vehicle license plate classifier based on convolution neural network,which can be used to select the license plate image from candidate image,and has higher recognition accuracy compared with SVM classifier in the original method.(4)This paper uses ASP.NET technology and SQLserver database management system software to design a vehicle intersections detention information management system based on B/S architecture.Owners and traffic police can query and manage the detention information through the system website.The preventive measures and detection algorithms proposed in this paper are of great significance to reduce the occurrence of traffic overflow and to improve the supervision efficiency of traffic managers for vehicles stranded at intersections.