Vehicle Detection Algorithms Based On Video Technique

The importance of acquiring traffic information timely is self-evident. But trafficinformation collection methods can not satisfy the demand of traffic control andmanagement at present. With the development of the computer technology, it'swidely used for the technology based on the video in traffic. Compared with othersvideo detection has the following advantages: wider detection scope, more flexiblesetup, and easier adjustment, convenient install and maintenance, richer traffic flowparameters. Therefore it has much meaning for detecting moving vehiclesparameters by researching the crucial video technology.This paper is completed based on the project "Research on detecting technologyon the mixed traffic flow characters by image recognition". We has an in-depthstudy on the contents of moving vehicles detecting and analyzes the mainlyalgorithms and technologies in detail such as background updating, shadowremoving, vehicle segmentation and tracing. Then we introduce main principleabout the system designs and present a vehicle detection method based on fictitiousloop, which can acquire vehicle parameters rapidly through detecting the fictitiousline changes in gray. At last it has realized in Visual C ++7 .0 platform. The resultsindicate that the algorithm is feasible to some extent and can detect the targetparameters rapidly in the scene.This paper is divided into five chapters. The content of each chapter is asfollows:First chapter is Introduction, which introduces research background,significance, present situation and development as well as the content of the paper.Second chapter is background updating and shadow removing. In the chapter weintroduce the research development firstly and present a method of backgroundupdating based on the improved EKF, which fully integrates the image informationin pixel-level, region-level, frame-level and can adapt to the background changesunder all conditions. Of course the shadow may be very serious and has negativefluence upon coming step's precision. So in the next portion we analysismechanism of generation and the former research fruits and put forward aself-adaptability shadow detection algorithm based on the RGB invariance. It getsgood effect during the experiences of detection and analysis of roadway segments,and highways, which proves the method have better robust and self-adaptability.Third chapter is the research on moving vehicle segmentation and tracingalgorithms. It is also the kernel of extracting the moving vehicles successfully. Atfirst we introduce based-threshold, based-region and based-edge imagesegmentation methods and their applicability respectively. Then we present awatershed segmentation method used the markers, which integrate the movingvehicle templates can be acquired in the last chapter. The segmentation effect isvery well. In the end we use an EKF to trace the moving vehicles in twodimensions and introduce the tracing process and the experimental effect in detail.Fourth chapter is the video detecting system design and realization based onthe fictitious loop. We firstly introduce main principle about the system design.Accord to meet the demand of timing;we adopt Object-Oriented method based onthe studied algorithms. The vehicle detection method based on fictitious loop,which can rapidly acquire vehicle parameters such as vehicle numbers, velocity,rate of occupy, vehicle size through detecting the fictitious line changes in gray, ispresented. At last it has realized in Visual C ++7 .0 platform. The results indicatethat the algorithm is feasible to some extent and can detect the target parametersrapidly.The fifth chapter summarizes the full text and prospects the work of next step.