Object Character Analysis Based On Visual Perception

Hawk can catch the target in complex background from sky. This ability has been the pursuing goal of intelligent tracking system. It is usually necessary to extract static features and dynamic features for moving target detection and tracking by simulating visual neural information processing system through bottom-to-up target attention. As static features are always the foundation of the dynamic features, this thesis mainly focus on the study of static features extraction and target recognition in complex background, and then texture feature is integrated into visual perception calculation model to improve the salience of interested target in saliency map, so as to firstly focus on interest targets and quickly detect target.Firstly, the thesis introduces the current development of visual attention model and a discussion of the structure and features of the human perception system. An improved model is built by adding a features classification to the Itti visual perception model. Through experiments, it will also prove the veracity of the promoted interest-attracking target achieved by the improved model.Secondly, texture feature is extracted by using the Graylevel Co-occurrence Matrix(GLCM) and is classified as natural or man-made object by Surpport Vector Machion(SVM). There are texture feature differences between natural object and the man-made objects. The texture feature of the natural objects is always rough, and that of man-made objects is subtle.Finally, a novel visual perception model is proposed to improve the salience of interested target based on texture extraction and classification, and test results are presented.