Research On Detection And Tracking Of Feature Fusion Jackets

Target detection and tracking is a research focused on the computer vision, which combines research of pattern recognition, image processing and artificial intelligence, is widely used in military guidance, car navigation, security monitoring, video retrieval,intelligent transportation, behavior recognition and astronomy and other fields. Therefore,the research of detection and tracking methods has important theoretical significance and practical value.Based on the careful analysis of domestic and foreign research present situation and the algorithm of target detection and tracking, this paper proposed to improve the current mainstream target detection and tracking method, the main contributions of the paper are as follows:1, Target detection segmentation algorithm combined HSV space with morphologyIn this paper we present an image segmentation method which combines HSV color space information with the operation of mathematical morphology, and extracts foreground by background subtraction which can eliminate the wrong background segmentation. Firstly it extracts HSV color feature which belongs to target object, next uses the feature value to mark ROI areas, then gets connected areas by the operation of mathematical morphology of the binary image which contains tag information, finds the outside rectangle boundary by extracting contours of connected areas, finally uses background subtraction to eliminate background interference. Applies it to detection of life jacket which appears on ferry safety,the experimental results show that for the life jacket image, this method achieves image segmentation quickly and effectively, and counts the number of segmented regions.2, Particle filter based on the improved weight reducing strategy and more features fusionThis paper combines edge feature with Zernike moment feature selectively, takes the complementary advantages of edge feature and internal shape feature, improves the processing capacity of outside interference such as illumination change and occlusion etc,increases the robustness of algorithm. At the same time, by introducing a weight reducing improvement strategies, updates particle swarm of small weights, improves the diversity ofparticles at the same time also increases the particle effectiveness, inhibits the particle degradation problems, improves the tracking accuracy of the proposed algorithm.