The application of remote sensing data as a new source to update GISis increasingly popular in recent years.However

using GIS to improve classifica-tion of images is neglected in most of current approaches.Although attempts havebeen made in using texture and context information

the effects have limitation。The paper proposes a new method to integrate GIS so that the shape information

which is frequently used in visual interpretation

can be easily employed to improvethe performance of classification。The confusion of classification can be significantlyclarified after shape information is involved。