The Research Of Detection And Tracking Of Moving Object

Identifying moving objects from a video sequence is a fundamental and critical task in manycomputer-vision applications. Surveillance video captured by static camera is becoming more andmore widespread in the modern society, and it has brought huge change for people's life. Itmainly applies in many domains such as identity recognition and intelligence guarding in security,vehicles license plate recognition and traffic flow analysis in intelligence traffic system, cellmovement rule analysis in medicine examination, hostile target track in military combat, trainingdata analysis in sports training, etc.This article has discussed and studied the key technology and algorithm of target detectionand tracking. In movement target detection aspect, this article studies and analyses massivealgorithms. For the peculiar circumstance, if the target is person, first we process the originalpicture with the method of light compensation and filter. Then we use skin model to detect themovement target fast. If the target is any-other, this article studies and analyses the methods ofbackground subtraction, neighboring frame subtraction, optical flow, background estimation, etc.According to the situation of the pictures which the quality is not very good, we use the methodof accumulate projection to determine the approximate position of the target and improve themethod of accumulate projection to adapt the reality situation. We proposed a method ofdetermining the target center of gravity by the method of rectangular search. In the aspect ofobject tacking, we have studied the method of Snake algorithm withdrawing the object outline,template match and Kalman filter's movement object estimate. In view of the question which thetraditional neighboring frame subtraction and background subtraction are not very good if thereare multitudinous object, a method of withdrawing object when there are two targets has beenproposed. The experiments have proved that this method is very useful to the multitudinousobject detection. A method of combining the accumulate projection and the connectiveexamination has been proposed. And we have realized the movement object tracking using theSnake model and template model. Simulator examples and coding have carried for all kinds ofplans and algorithms. Tests and revision have carded many times for the last using algorithms.And these tests have achieved the anticipated effect.The research of image has extraordinary broad application foreground,, the paper's researchis a beneficial explore in the application of image analysis and intelligence.