Remote Sensing Image Classification, Fuzzy Pattern Recognition And Decision Tree Method,

As a part of the Research of Processing Flat and Application in Remote Sensing Data supported by the National Hi-New Technology Research and Development Plan ("863" Plan ) Project, the thesis has made a deep research on some key techniques and algorithms of remote sensing image classification. In allusion to the shortcoming of traditional supervised and unsupervised classification, fuzzy pattern recognition is inducted into classification and improved fuzzy c-means algorithm is propounded obtaining better classified effect. Based on the decision tree's models of some special application which is propounded by the former researcher, the thesis builds a general-utility symbol warehouse of knowledge regulations, and designs a general-utility decision tree classification flat that giving convenience to build decision tree model and to classify. At last a perfect system of Remote Sensing Classification is designed and implemented as a body with "image classification, post-processing and accuracy assessment". The main research work is as follows:1) After analyzing the methods of traditional supervised and unsupervised classification, the thesis presents classified thought and algorithmic flow to six kinds of traditional classification algorithms. Pointing out the shortcoming of traditional classification method based on statistics pattern recognition and combining with the method of fuzzy mathematics, fuzzy pattern recognition is inducted into the remote sensing image classification. On the foundation of fuzzy c-means algorithm's research, the thesis propounds an improved fuzzy c-means algorithm that is fit for remote sensing image classification. It can be obtained better speed, higher accuracy of classification and better effect images by using the improved algorithm which is combining neighbor statistics method to improve the pixel subjection degree function.2) Based on the decision tree's models of some special application which is propounded by the former researcher, the thesis builds a general-utility symbol warehouse of knowledge regulations, and designs a general-utility decision tree classification flat that giving convenience to build decision tree model and to do decision tree classification on remote sensing application.3) At last based on the above researches, a perfect system is designed and...