| Visual analysis of moving objects involves much professional knowledge such as pattern analysis, image processing, artificial intelligence, and it's the research hotspot in computer vision. Visual analysis of moving objects includes detection and tracking of moving object. Furthermore, it also includes the classification, recognition and behavior comprehension of the moving objects and so on. This paper presents a set of detection and tracking algorithm for moving objects, and it's able to detect moving objects under the complex scene and track them steadily, and it's also able to classify several kinds of moving objects.The main work and research results of this paper are as follows.(1) A moving object detection algorithm based on background model of adaptive optimal feature is presented in this paper, which models background for many features, and then selects the optimal feature adaptively, and utilizes background model of the optimal feature for motion detection. Many experimental results have demonstrated that it can remove the background noise effectively and detect moving objects robustly.(2) A multi-object tracking algorithm based on particle filter of color space analysis is proposed in this paper, taking the analysis of the color space modeling for muti-object, extracting robust color feature of every object for modeling, and introducing a corresponding method for solving the covering problem. Many experimental results have demonstrated that the proposed algorithm solves the covering problem of multi-object tracking and leads to good tracking results.(3)Based on the algorithm of motion detection and tracking, this paper researches the algorithm based on muti-class SVM for classification of moving objects and realizes to classify some specific kinds of moving objects by extracting some new features with other efficiency features for objects modeling.(4) Combined the above methods, a number of experiments are designed to verify the robustness and correctness of the algorithms. |
PDF Full Text Request