Target Detection And Tracking Method Based On Feature Points

Target detection and tracking is a hot and difficult subject in computer vision. It was widely used in traffic control, human-machine interaction, precision guidance, photoelectric navigation, etc. There are two main aspects of the key technologies in target tracking system which includes the target detection and tracking. And the target detection is more crucial.In this dissertation, the methods of feature points detection and tracking are studied against clutteried background. Meanwhile, a novel approach of feature point tracking is presented in the dissertation. The main research works and contributions are summarized as follows:(1) Several classic feature point detection methods are discussed in this dissertation, such as Moravec comer detection, Harris corner detection and SUSAN corner detection algorithm. And then computer simulation is conducted to analyze these methods' advantages, disadvantages and applicability.(2) A novel approach of feature point detection, called the scale invariant feature transform (SIFT), is introduced in this paper. At the same time, for the problem of its low efficiency, the method of calculating extreme values in scale space is improved. The computing time decreases obviously by using the proposed method.(3) The proposed KLT (Kanade-Lucas-Tomasi) tracking method based on the SIFT feature point is introduced in this paper. Simulation results show that the proposed approach is able to detect the feature points effectively in the video and image sequences, and it provides a reliable basis for the next target tracking steps.(4) The moving target tracking procedure is established by the the proposed feature point detection approach. Then, many diffirent target tracking experiments in variant video scenes are completed based on the effective feature detection and extraction method combined with Kalman filter.Experimental results show that this proposed approach has higher stability and tracking accuracy than the conventional method.(5) Finally, the simulated calculation by using the VC++ 6.0 and Matlab 7.1 tools shows that the proposed tracking algorithm has a good tracking effect in the practical applications.