Research And Realization Of Multi-Sources Remote Sensing Data On Object Oriented

With the rapid development of remote sensing technology, remote sensing for a global multi-level and multi-angle monitoring has been widely used in mapping, land resources, agroforestry, urban planning, geology and mineral resources exploration, military and other industries and domains. The rapid acquisition of massive high-resolution remote sensing images and the in-depth application of GIS, set higher requirements to extract geographic information accurately from Multi-source remote sensing data.The traditional pixel-based image classification method, because of neglect of information about the rich semantic space and the obvious texture structure of the high spatial resolution images, and basically not considered a variety of remote sensing information sources complementary integration, whose classification result is difficult to meet the scientific and engineering application needs. The new object-oriented image analysis method, making fuzzy logic classification based on features of both spectral and spatial of the multi-source remote sensing images, has already demonstrated an obvious advantage. At present no such technology based on the multi-source image analysis system, which blocked the use of image in various area.In this thesis, this object-oriented classification technology to multi-source remote sensing image has been studied systematically and deeply, and on this basis the multi-source remote sensing information extraction system has been developed. Main research results are as follows:1. The multi-scale segmentation technology of multi-source remote sensing image has been researched and realized. In this part, the construction the objects of multi-source image scale space has been solved mainly, and the image object network structure of multi-scale layer has been formed, it is that the image pixels were divided into regions with different brightness value at different scale layers, and then different object polygons were extracted. On the basis of studying various image segmentation algorithms, and combining with high-resolution imaging features and multi-source data integration characteristics, the fractal net evolution image segmentation approach has been researched and achieved, which laid the base for object-oriented classification of multi-source remote sensing data.2. The description method of image object features has been studied. In this part, the simulation and description of geographical entities features based on image objects has been solved mainly, and the quantitative expression model of the spectrum, texture, shape and other geological features of image objects has been researched and established. The quantified expressions of image object features based on semantic context information have been researched. The structural and visual expression of each description feature of the image object has been studied, and the reusable description base of image object features has been built. So the sufficient alternative bases of classification features have been provided for the object-oriented fuzzy classification.3. The objected-oriented fuzzy classification tecchnology has been researched and realized. In this part, fuzzy system was introduced to object-oriented image classification on the basis of studying the knowledge-based fuzzy systems. Three parts of fuzzy classification system - the fuzzification of classification rule, fuzzy rule base and defuzzification - have been researched and realized focused. The fuzzification methods and means of multidimensional image object features have been solved. The accuracy evaluation system to object-oriented fuzzy classification has been built by studying several methods of accuracy assess of fuzzy classification.4. The object-oriented multi-source remote sensing data classification system has been designed and realized. In this part, the design of system structure and system function module has been elaborated. Aim to the more extensive use of high spatial resolution image information extraction and classification based on multi-source remote sensing data, two examples have been shown respectively to prove the analysis result of this system.Finally, the research work has been summed up and the further research direction and some problems have been pointed out.