Three-dimensional Target Recognition System Based On Remote Sensing Images Of The Model Study

Recognizing the 3-D object in a single-view intensity remote sensing image by using a precise descriptive model is a task valuable for both theoretical research and practical usage, and is served to many fields of defense as well as civil economic construction.Since there are some limitations when the main theories of computer vision are applied to this task, a new technique way is presented in this paper. The view-independent 3-D structural model of an object is firstly constructed, the relative 2-D aspect model is following obtained, which reflects the different appearances of the object in various views. Then, both statistical features and structural features in various levels are utilized to construct the indexing mechanism. Once the primitives are extracted from an input image, a process of indexing-matching is performed to realize recognition speedily and efficiently. In this approach, the difficult part of the task is shifted to the modeling process, which can be done offline, and the ill-posed problem of reconstructing the 3-D structure of an object from a 2-D miage is avoided. Moreover, the question concerning a too large searching space is solved by upgrading the primitives' configuration degree and by means of effective indexing.The research in the paper has covered the three main works needed by an object recognition system, including modeling, low level processing and matching. A representation method of a 3-D object based on general cone theory and a construction method of the relative aspect model is proposed. New potent algorithms of such areas as the extraction of straight line, salient structure, the external contour, and the approximation of polygon are developed. The technique of hierarchical indexing-matching is also put forward, which combines the statistical and structural recognition method and gives attention to both the local and the global features. Experiments show the validity of the new approach when it is applied to synthetic and real images.In addition, the idea and the method of system analysis and design utilizing object-oriented techniques are proposed towards the problems involved in the development of a recognition system...