Research On Video Object Extraction And Tracking Technology

Since the application of intelligent video surveillance system is getting to be moreand more widely, with the scale is continually expanding, video object extraction andtracking technology has been paid more attention as the two of it’ s key technologies,and attracting a large number of experts and scholars to participate in the study athome and abroad.This paper aims to research the video object extraction and trackingtechnology with static background, the main contents include the following:In the video object extraction, this paper first introduced the typical movingobject extraction algorithm at present, then analyzed and compared the merit anddemerit of various algorithms. On this basis, used the method of based on backgroundsubtraction, which combined with the Surendra background update algorithm todetect moving target. In order to adopt the changes of light in the scene，introducedthe Otsu dynamic threshold algorithm in the detection process, and added a incrementwhich reflect the changes of light as a actual threshold, thus we can detect movingtarget more effectively.In the shadow detection, on the basis of summarized the characteristics of RGBand HSV color space model, chose the shadow detection algorithm based on HSVcolor model which is closer to human physiological characteristics of visualperception to eliminate the shadow, and got a good result.In the video object tracking, first, introduced several commonly used movingtarget tracking algorithms, then after analyzed the applications and merit and demeritof various algorithms, focus on studied a target tracking algorithm which based onfeature-Camshift algorithm, finally, for the lack of this algorithm，used aalgorithm which based on combine Camshift algorithm with Kalman filter to trackingmoving targets, and it can solve the occlusion problem effectively.The experimental results showed that the video object extraction and trackingalgorithm that this paper used can detect and track moving targets effectively, can geta good adaptability of the changes of light and target occlusion in the scene, and canachieve a better result of detection and tracking.Finally, summed up the overall works of the paper，discussed the shortcomingsand the future research direction.