Research And System Implementation Of Photo Detection Method For Illegally Parked Vehicles

With the number of vehicles increasing in our country,illegal parking problem is also on the increase.Traffic police can't penalize all the illegal parking behavior because of some objective factors,such as their work time and manpower,while fixed camera or monitor equipment also cannot catch all violations due to its blind area.The supervision way of reporting illegal parking phenomenon based on smart phone allows people to take photos of traffic violations anytime and anywhere,which can make up for the problems of police resources limited and monitors having blind areas.But it can't avoid that there are invalid or duplicate reports and photos in the illegal parking information which is reported by the public.The currently existing report system usually filter them manually.It not only increases the audit's workload and reduces the efficiency of processing reports,but also wastes store resources of the system.According to the problem that there are invalid or duplicate report photos in the process of reporting vehicle illegal parking,this paper researches the vehicle detection and recognition method for static image,proposes a kind of vehicle recognition method which is called vehicle image recognition method based on local learning(VIRLL)and detection methods for report photos of vehicle illegal parking.Then it designs and implements the report photo of vehicle illegal parking detection system.In this paper,the main research work is as follows:1.This paper proposes a vehicle image recognition method based on local learning which switches the vehicle detection based on static image to a binary classification problem of target and background using super pixels as the unit.It is based on center selection policy for local learning(CSPLL)and Bhattacharyya distance,divides the sample into several subsets and trains a local classifier for each subset to detect whether each super pixel belongs to vehicle,thus confirms the area of target vehicle.The simulation results show that this method has better recognition rate and recognition effect on the problem of static vehicle image recognition.2.For the effectiveness and repeatability problem of vehicle illegal parking report photos,designs the detecting precept of report photo's effectiveness for vehicle illegal parking and the multi-information combined repeatability removal precept of vehicle report photo.The effectiveness detecting precept takes use of the license plate recognition,VIRLL and Orientation sensor data of mobile to test the photo's effectiveness.The repeatability removal precept uses the license plate information,location information,reporter information,time,and image similarity between two photos to filter and deal with the report information and photos.3.Design and implement a report photo detecting system of vehicle illegal parking which includes Android App client and back-end management platform.App client mainly contains reporting and uploading module as well as report view module while the back-end management platform includes some modules such as report management,effectiveness detection,timed repeatability removal and illegal parking distribution.It can detect effectiveness and remove repeatability of the report photos of vehicle illegal parking taken by the public using smart phone and filter out part invalid or duplicate reports,which greatly reduces the workload of human review.