Data Fusion of GIS and RS by Neural Network

Geographic Information System (GIS) is a kernel technology of the earth observation. Remote Sensing (RS) is an important external information source of GIS and a useful tool for renewing data. On the other hand, GIS can assist in analyzing RS data. Integration of RS and GIS is important for collecting, managing and analyzing spatial information. For this purpose now scholars are studying the basal data structure for integrating the two types of data. The paper provides a method of data fusion of GIS and RS using neural network with unchanging data memory structure based on users´ aim. The paper constructs a 5 layered neural network. Inputs of the network are the feature values of RS and GIS. Adjust the distance between input vector and center vectors and at the same time regulate the value of the center vectors. The weights of the network are determined by both ways. Outputs of the network are changes of attribute values. At last update the database using the changed attribute value. The paper verifies the validity of this method by supervising the land use change with TM image and the vector map to update the attribute database.