The Theories Of Feature Extraction & Selection And The Application In Image Classification

In this paper, some problems in image texture are studied making use of the theories of feature extraction and selection. Further more, the author carries out experiments on texture classification. The main content is as follows: The foundation of feature base 15 texture features are gained through gray-level histogram, co-occurrence matrix and run-length matrix. Feature extraction and selection In this section, different methods on feature extraction and selection are proposed, such as method of transformation based on classifiability criterion, method of discrete K-L transformation, method of Fisher transformation, method of direct feature choosing and so on. Post-classification process In order to make the classified image more practical, we need to do something else to improving the quality of classified image. The post-classification process includes median filtering, sieve, eliminate and the transform between grid and vector. The program design of real-time inspection in land use This program can inspect the changed areas in land use automatically by computer.