An Simi-Automatic Extraction Of Remote Sensing Image

With the rapidly development of space techniques and information techniques, the high resolution secondary planet can supply human with the digital images of high accuracy and large capability. According to abundant detail information and obvious geometry of the high resolution RS images, target extraction from images has become the important means of the space information updating, and has been used extensively in the national economic production and martial target detection. Based on the automatic interpretation of spatial image and the actuality of region, and emphatically solved the problem of typical region( such as large residential area, water area and plant area) semiautomatic extraction and accurate boundary orientation. The key technique is some correlative techniques, such as machine learning,image segmentation and target pattern extraction etc.This paper puts forward a new semi-automatic extraction of polygon features. At the first stage, the interesting region can be abstracted by the foreground and background curves which can be drawn manually. The algorithm uses pixel value from foreground and background curves as training sample of SVM to gain the initial segmentation result. In this foundation, the initial segmented regions are merged according to spectral features and shape features. Finally, a new criterion is proposed to decide the termination of the merging process.Support Vector Machine(SVM ) as a most cutting-edge and most effective method in statistical learning theory has become a new exciting fielding in pattern recognition and machine learning. Analysis and dealing with RS image is a hot research in SVM application field. SVM is a new machine learning method, under small finite samples and found on the principle of structural risk minimum(SRM), manifests many priorities than other method used before in actual questions of finite training samples. In classification of remote sensing image, with the method of SVM classification, image data don't have to degrade the high dimension and the speed & the precision have greatly improved. The paper focuses on the application of RS image target extraction based on SVM. According to the results of experiments, the initial extraction can make us separate the foreground of the image from the background basically with high precisian and speed.The segmented regions are the shape representation of objects, so the quality of segmentation quite impacts the precision of the following analysis, recognition and comprehension. The paper uses a region-merging. This method computes the merging cost of two adjacent regions based on their spectral and shape features. Then segmentation results of different scales can be retrieved by limiting the merging cost. To improve the efficiency of segmentation, the initial region adjacent graph is partitioned. The experiment results show that this method is accurate and efficient.An object-oriented remote sensing image processing platform named ELU is designed and implemented by Remote Sensing Laboratory of Shanghai Jiao Tong University, supported in part by the research project of Content-based Search in Image and Change Detection and Auto-updating of Special Target, which is administered by Shanghai Science and Technology Committee. Irregular polygon features extraction, image segmentation and machine learning method are studied in this thesis supported by the project. The corresponding modules of ELU system is also designed and implemented according to the study.