Construction of knowledge rule sets for the classification of land cover information for the coastal zone of Peninsular Malaysia

Abstract

ABSTRACT Landsat images have large advantages for studying land cover changes over long time periods and have become the main data source for the extraction of large-scale land cover. However, due to the image resolution limitations, it is difficult to classify second-class land cover types with high accuracy through only traditional automatic classification methods. This study selected the coastal zone of Peninsular Malaysia in Southeast Asia as the experimental area for studying the knowledge rule sets classification method. First, a system for the classification of coastal land cover suitable for Southeast Asia was established based on regional characteristics, which included 8 first-class and 18 second-class types. Then, through the combination of multiple geoscience knowledge rules, a set of knowledge rules for extracting the second-class types was established by integrating multiple sources of information, such as the temporal features, topographic features, texture features, shape features, topological features and spectral features. The experimental results showed that the knowledge rule sets were effective in the extraction of the secondary land types in the coastal zones. The overall classification accuracy was over 85%, and the Kappa coefficient was higher than 0.8. Compared with the traditional supervised classification method, the classification accuracy was obviously improved.