Research On Method Of Target Detection And Tracking In Video

With hardware and software technology developing fast, kinds of research and exploitation on image recognition is becoming a hot spot in recent years. The smart video tracking synthesizes subjects of mode recognition, machine vision, image processing, artificial intelligence, etc. So it becomes a research method with great application value. And tracking and recognition of moving objective are the two key parts of it, so they are important research tasks.Firstly, this paper compares methods of traditional edge drawing and detecting, and points out the advantage of colorful edge drawing. Then, detection ways basing on adjacent frame difference and basing on background difference are introduced. Experiments prove the former method operates more easily, faster. Besides, it has perfect result when background is still. So this paper does moving target drawing using the adjacent frame difference method.As to the moving objective tracking, this paper puts forward a statistical method by block matching points, basing on researching kinds of target tracking ways. This method uses theory of block-matching and point's statistics. It can track target fast, with high accuracy and other superiority. Basing on it, this paper puts forward a program of humanoid contour tracking. It tracks target using the method of blocking-match and point's statistical together with snake algorithm.Using the proposed program, this paper designs a warning zone intrusion system. It improves the performance of the traditional method of block-tracking point's statistics, with higher sensitivity for target detection. And then, this paper builds up a recognition simulation system to recognize foreign body in application-specific scenarios. It uses the method of point's statistics and snake algorithm to recognize the foreign body automatically, reducing the human input greatly.The experiments show that, the algorithm of target detection and tracking has good robustness, and carries through a significative exploration at intelligent target tracking and behavior recognition.