Camouflaged Moving Objects Detection And Tracking For The Wild Environment

The camouflaged moving objects detection and tracking is a difficult research task in the field of computer vision, and the related researches are less at home and abroad. Currently, the proposed methods for object detection and tracking are mainly based on the human visual differences between objects and backgrounds. So the objects detection and tracking may be fail when the objects and backgrounds are similar. The camouflaged moving objects are a special kind of objects whose visual effects and intensity are similar with background. In the wild, the backgrounds are dynamic, and there are still too many interferences, so these factors result in the increase of difficulties of camouflaged moving objects detection and tracking. Concerning the research topic in this article, the research is starting with the summary of digital image processing technologies and the classical objecting detection and tracking methods. With the research deeper and deeper, a novel method is proposed based on the spatial structure for camouflaged moving objects detection and tracking. Combining with the motion estimation of the objects, the camouflaged moving objects task can be accomplished.Camouflaged moving objects are a kind of moving objects with visual camouflage properties, the study of this kind of objects should be based on the existing research results. In the article, the basic technologies of digital image processing, the common methods of objects detection and tracking in three aspects, including background modeling, objects modeling and motion analysis.In the article, an effective camouflaged moving objects detection and tracking method is proposed by using the spatial structure of the image, which is built up based on the existing research results. Firstly, the image is segmented into several regular blocks, and the spatial key points of every block are extracted based on SIFT features. And then, build up the spatial location relationship among the key points, and the statistical histogram of SIFT features are got as the orientation features of different blocks, both spatial location relationship and orientation features constitute the spatial structure of the image. Finally, the camouflaged moving objects can be detected based on the changes of spatial structure in the video sequence. Considering the historical motion data of objects, the proposed method fuse the spatial correlation and time correlation effectively.The experimental results show that the proposed method can detect and track the camouflaged moving objects effectively. But there is also a defect of the proposed method, the objects’ edge cannot be located accurately, because the spatial structure is designed based on the structure information. And the proposed method works well in the application scenarios with the distribution of SIFT features.In addition, other research works are introduced briefly in the last chapter of the article based on the mastered digital image processing methods during the studying of the camouflaged moving objects detection and tracking, including vehicle license plate location, vision servo system of electric robot and liver segmentation for CT images.