Research On Remote Sensing Image Processing Based On Information Fusion

As one of the most precise guidance techniques today, imaging guidance is widely used in terminal guidance of various missiles. Computer identification location technique is the key one in imaging guidance. Based on the backgrounds of imaging guidance, remote sensing technique, and information fusion technique, we have studied the methods of the remote sensing image processing in this paper. The main conclusions are as follows.(1) We have researched the remote sensing images systematically. We summarized the identification features, appraisement standards, as well as the target characteristics of the remote sensing images. Because of the most notable characteristic of remote sensing is huge data, and as the only means to afford dynamic and global observations until now, the research for the remote sensing image processing technique is very important on many military or civil applications.(2) We have analyzed the principles and methods for image fusion toward applications on remote sensing in details. Through integrated usage of multiple-source information, data fusion can get more objective and intrinsic knowledge of certain target. After comparing the methods of image fusion among pixel level, feature level and decision level, we got the different features in these three levels. Moreover, we pointed out the existent problems and development tendency in future.(3) After comparing many traditional fusion methods in spatial or frequency domain, in the paper a self-adaptive image fusion method based on dynamic weighting of regional features is proposed. By fusion experiments of SAR and TM images, it appears that the proposed method has apparent advantages in reservation of information and boundary as well as neighboring coherence.(4) Based on the traditional method of image restoration by maximum entropy, some disadvantages are pointed out. We supplied a new idea to solve the maximum entropy problem by sequence quadratic programming. At the same time, we explored a new method to solve general positive definite quadratic programming, and got the convergence of the method in finite steps.(5) The researches on edge detection for the images of remote sensing have been made systematically. Some properties of common edge detection algorithms have been compared. We provided a new edge detection algorithm based on multiresolution analysis. From experiments, the algorithm can get effective edge information, and improve edge location precision while restraining noise efficiently.(6) We have studied the methods of image texture segmentation systematically. The texture feature is the important feature of the remote sensing images. We discussed the techniques of image texture segmentation in details, and compared the differences. Finally, we pointed out the main problems in existing methods.(7) Mathematical morphology is one of the new subjects studying the structural feature of digital image morphology and fast parallel processing technique. The structural analysis and feature extraction is made by morphological transformation in object image. Firstly, we showed the basic functions of dilation and erosion in mathematical morphology. Afterwards, we extracted the edge in binary images by using these functions. The experiments testified that these two functions could extract the edge in binary images effectively. Finally, some airport features in aerial images are extracted by using mathematical morphology method.(8) Establishing expert system is the only way to understand remote sensing image. In the paper, we have researched remote sensing image understanding. We got a simple model to recognize the airport in aerial images by using knowledge graph. The project "Knowledge Graphs" was started in 1992 as a joint activity of the faculty of sociology of the University of Groningen and the faculty of applied mathematics of the University of Twente in the Netherlands. The initial goal is the investigation of the possibility to use graphs for the representation of kno...