Research Of The Technology Of LANDSET7 ETM+ Remote Sensing Image Classfication And Recognition

The classification of Remote Sensing image is one of the important research content in Remote Sensing image processing. The classified precision influences the application level and practical value of the Remote Sensing data directly. How to classify some land use types and satisfy a certain precision is a key problem in the study of Remote Sensing image. It has an important significance. It can provide foundational information for researching in disaster evaluation, town enlarging, environment changing and so on. But at present the most used method in recognizing Remote Sensing image is judging by people. The method does not fit the demand for real time operation, and need more manpower, material resources.According to the project of "Measure and Monitor of Sand Resource in Inner Mongolia Autonomous Region", this paper takes the LANDSET7 ETM+ Remote Sensing image of Bayannaoer city in Aug 2002 as source data. Its major tasks are to research the texture analysis and neuro-fuzzy network using in Remote Sensing image classification. According to the actual situation of the selected areas, we divide the land cover/use into six classes: mountain, city, water, vegetation, sand, gobi. In the experiment based on texture analysis, we use the energy, contrast, entropy, correlation, inverse difference moment as feature values, and adopts FCM as classifier. In the experiment based on neuro-fuzzy network, we used TM1,TM3,TM4,TM5,TM7 as feature values, let "winner takes all" as classification rule.According to the method metioned above,we use the Remote Sensing image classification system developed in vc++ environment and NEFCLASS system to do the experiment.The result shows that both methods we used can obtain better outcome comparing to the maximum likelihood method.