Research And Application On Moving Objects Tracking Algorithm Based On Binocular Vision

Binocular vision tracking and measuring is a technique which based onbinocular vision technology to detect specific moving object for recognition andtracking, and three-dimensional coordinate measuring. The technology can be widelyused in military research, industrial and civilian areas. In the game of sportsvideo,automatic tracking and ranging ball is a popular direction. For the purposes toreal-time3D tracking and measuring ball in sport project,this paper is the study ofreal-time binocular vision Tracking and measuring technology.In this paper,the ball object real-time binocular vision tracking and measuringwork is divided into two parts. The first part is to study the real-time object detectionand tracking technology and binocular vision calibration technology. The second partof the research results using the first part of the analysis to select the bestcombination of programs designed to achieve real-time3D tracking and ranging theballs application system. The main work done as follows:1) based on the comparativeanalysis of the common background modeling detection algorithm, proposed aspecific moving object detection algorithm.It is combined of HSV color conversionpretreatment and regularly updated background model,making the system has beengreatly improved detection performance;2) Testing the template matching objecttracking, camshift tracking and kalman filtering tracking solutions, after analyzingthe advantages and disadvantages of the three object tracking algorithm, papercombined the object detection results presented a fast update of the region of interestROI kalman filter object tracking algorithm.This work makes real-time systemssignificantly improve;3) Comparative analysis of the various common cameracalibration technique.Select and implement a moderate precision and flexibility ofZhang Zhengyou calibration method, more accurate calibration of the binocularcamera platform;4) Comparative analysis method to locate the center of the objectimage, the paper proposes an ellipse fitting target binary promising method to quicklylocate the center，and then match the center of the object to achieve the goal ofreal-time three-dimensional measurement.Experimental results of the design and implementation of the various modules ofthe system are as follows: On a normal webcam connected PC platform,binocularcamera can display real-time detection of of table tennis complete outline clearobjectives;For thrown of table tennis,the system can measure real-time tracking andobtaining the three-dimensional coordinates at a rate of30frames,and binocular distance results in the appropriate distance actual error is below5%,reaches thesystem measurement accuracy requirements.The results show that the system isexpected to complete the design goals, achieve real-time tracking ranging purposes.