This paper used Chongming Dongtan Nature Reserve as the research object for salt marsh vegetation classification based

on Landsat TM image. According to such image preprocessing measures as image geometric correction and subset image and on the

basis of analyses of Landsat TM remotely sensed images integrated with field survey and other studies of spatio-temporal dynamics

of Chongming Dongtan Nature Reserve, this paper confirmed the species of the vegetation in this area. The authors used knowledge

engineer to classify the vegetation, built knowledge base on the basis of vegetation spectral information and presented a

vegetation classification method based on the spectral information. The overall precision of the vegetation classification method

based on knowledge engineer is 92.35%, and the kappa coefficient is 0.907 2. The precision is higher than the overall precision

of the vegetation classification based on unsupervised classification and supervised classification (maximum likelihood): the

overall precisions of unsupervised classification and supervised classification are respectively 86.92% and 90.10%. The result

shows that the vegetation classification method can classify and discriminate vegetation effectively and the precision is higher

than that of other methods. The vegetation classification method provides a theoretical foundation and effective method for

automatic extraction of vegetation.