The Key Technology Research Of Target Detection And Tracking

Compute vision is developing rapidly In recent year, which fuses multiplie technologies, such as image processing, pattern recognition, artificial intelligence, artificial neural network, automatic control and so on. As one of the most important research subjects, moving target tracking can get the importance information which interest people, such as the location, the shape and the speed of the object, through detecting, extraction and recognition from the video sequence. What’s more, further analysis can realize the understanding of the behavior, so this technology is used widely in military and civil applications.so the research in this field is significant in both theory and application.This thesis discusses the background and significance of target tracking. In a complete target tracking system we should pick up the target using the technology of target detection, and background subtraction becomes a hot research topic in target detection due to its low computational complexity, strong robustness to noise. As an important part of backguound subtraction, backtground modeling is becoming a re-search focus in resent years. In this thesis we choose codebook after comparing sever-al methods, and improve it based on the original method, the improved method has a stronger robust to the change of illuminance by mapping the original method from RGB to YUV color space, on the other hand, and the weighted coefficient is Intro-duced to deal with Ghost phenomenon. At the same time, the speed of backgrouond modeling and target detection are both improved.By combing the Kalman and Camshift, Which choose color as characteristic and Bhattacharyya coefficient as the similarity function to describe the similarity of the two model, using Kalman to estimate the location of the target in the next frame can reduce the scope of search and enhance the tracking accuracy. Experiments show that this method can realize a stable tracking effect.