Abstract:
The expert classification system is a hierarchy of rules,or a decision tree,that describes the conditions under which a set of low level constituent information gets abstracted into a set of high level informational classes.The remote sensing Expert Classifier provides a rules based approach to multi spectral image classification,post classification refinement,and GIS modeling.Dazhongdian township in Northwest Yunnan province was selected as a study case,the vegetation classification method by use of expert classification system provided by ERDAS IMAGE software was explored.In the process of rule base building,besides the remote sensing multi spectral image band data was organized as variables,aspect,slope and elevation obtained from TIN of terrain data by GIS function also being considered as variables.In addition,NDVI,which is known effected for separating vegetation from the others,was also taken part in te rule base builing.Based on the above variables preparation,the multi spectral feature was analyzed and the main types in the study area were defined.The conditions and rule base was set up.The vegetation classification was organized and implemented in Expert Classifier.As a result,four different types of which three vegetation types and one no vegetation wee mapped.The vegetation types are classified as forest,mountain meadow,and agriculture land.Based on the case study,the primary results show that:① Compared with the conventional classification technique that uses only multi spectral information,expert classification can use many more relevance geographic information.② Based on the feature analysis of knowledge in study area,by use of knowledge engineer model,expert can build up knowledge base that included condition variables and rules,so that classification can be effectively organized and implemented.