Research On The Technology Of Moving Object Detection And Tracking In Video Sequence

In the video sequence,moving object detection and tracking is one of the key tech-niques which implements intelligent video surveillance system(IVSS). The real videosurveillance system applied environment is usually complex, full of variety. The real-ization of moving object detection and tracking with veracity, real-time performanceand robustness is the direction for current IVSS study.This paper is under the develop request of the Guangxi science research and tech-nology develop project, which is named Research and Implementation of Object Track-ing System Based on Image Sensor Array(Gui Ke Ji 0731020). The related algorithmof moving object detection and tracking, which is carried out by single camera understatic scene, is studied. A project to realize moving object detection and tracking bysingle camera is presented, at the same time, an experiment platform based on movingobject detection and tracking by multi-camera is built and relevant experiment is didwith another classmate. The main work is as follows:1. Taking into account of the disadvantage of frame difference and backgroundsubtraction in moving object detection field, relevant improvement and perfection forbackground subtraction is made. A way to extract and update background is proposedin improved algorithm, morphological filter and area threshold restricting conditionis developed to deal with the results of the object segmentation, and a reliable objectcentroid for the subsequence tracking is calculated by using image moment. The resultsof the experiment show that improved moving object detection algorithm can fast,accurately and more completely detect moving object.2. Lucas Kanade optical flow is based on sparse features tracking method, andit has the disadvantage of easily losing tracking object and low robustness. Templatematching searches object template in the full image with excessive calculation andbad real-time performance. To overcome the disadvantage of two methods above inobject tracking field, relevant improvement and perfection is made. Object centroid,which is used as characteristic point, is tracked by using Lucas Kanade optical flow.The calculation of template matching is reduced by diamond search algorithm, whichsearchs object template in the adjacent area of the tracked object centroid. Using theobject centroid corrected and located by template matching as the next frame trackingfeature points can reduce the error, which is caused by using single Lucas Kanadeoptical flow tracking object centroid, realize reliable tracking, and offer reliable datafor the subsequent multi-camera object handoff.3. The experiment platform based on object detection and tracking by multi-camera is designed and carried out, morever, this platform can achieve such functions as: moving object detection, object tracking, object moving trajectory generation andobject handoff.