Research On Texture Features Based Recognition And Tracking Technology

Object recognition and tracking technology which is based on video, is become more and more important in military guidance system, visual navigation system, intelligent monitoring system, medical diagnostics system and weather analysis system and so on. In recent years,local descriptors are widely used in image processing, and it has played an important role in promoting the development of object detection technology. This paper descripe a method to improve the traditional object recognition and tracking algorithm which based on feature with local binary pattern alogrithm.This paper studies the basic principle of object recognition and tracking technology and some commonly used target recognition and tracking methods that based on the feature of image.In the study,three frame difference method is used to detect the moving object. And a correlation function is used to calculate the similarity of the template in the matching algorithm.Then this paper combine the correlation function and texture feature to run the matching algorithm. In order to make this algorithm running more fast, this paper use kalman linear prediction alogrithm to reduce the searching range. But the reasult of the experiment show that the template sensitive to block. So this paper propose a new meathod to tracking the object which has been partial obscured.The new method combine RGB color and tecture feature to describe the object and use meanshift algorithm to track the object. The result of experiment show that, although this method can track the object which has been partial obscured, the speed of this algorithm is very slow. At last this paper propose a new tracking system using both of those method. And the result of the experiment show that this new tracking system perform perfect in the condition of lightting change and partical obscured.