Moving Object Detecting And Tracking Based On The Video Image

With the rapid development of science and technology, it is applied to various fields ofnavigation and guidance, terrain matching, face recognition, Video Surveillance, medicaldiagnosis, military and other, and has the important practical significance and extensivepractical value. Therefore, a high detection rate and good real-time moving target analysisalgorithm is particularly necessary. This paper focus on the video images of the moving objectdetecting and tracking method under the background of vehicles against the demonstrationsystem.This paper first analyses the process of target detection and tracking system based on theAdaboost algorithm, describes a rectangle feature extraction method, analyses the reason oflong training time for this algorithm. In ensuring the accuracy of the premise, select rectanglefeature template and optimize the feature of selection to improve the training speed, and selectthe appropriate method for image processing. Second, select the Multi-angle tank modelsample and trained cascade classifier in the detection process. Carry out a detailed analysis ofthe composition of the cascade Adaboost algorithm and testing processes. Select the bestwindow search strategy and reduce the detection time. Third, introduce the colorcharacteristics to Classify for tracking vehicle in Multi-objective real-time tracking stage.Output target coordinate position, track the target trajectory and assign an ID number. Set upthe PC software and verified by experiment. During the experiment, analyze and compare theresults of the single-objective and multi-target detection. Detection rate is beyond 92.7%,Multi-target detection rate is 28.9ms/frame. The experiment shows that the algorithm ofdetection and tracking performance is effective and real-time.