Abstract: The Sanjjiang Plain is one of the biggest marsh areas and the most concentric distribution area of marsh in our country.The area of marsh makes up 10.3% of the total region.Based on the principle that marsh types can be distinguished on the Landsat image we make marsh plant community as a basis dividing marsh types.There are 6 marsh types:Carex lasiocarpa marsh,Carex pseudoca raica marsh,Phragmites communis Deyeuxia angustifolia marsh,and Glyceria sp-Carex sp marsh,Carex sp marsh and Carex marsh with taro in the whole region. After analysing spectral curve characteristics of marsh waterbody,marsh soil, marsh plant and difference among them and non-marsh waterbody,soil,plant we think that spectral curves of marsh plant represent spectral curve characteristics of marsh to a great extent.The marsh plant grows on water and soil of marsh.Various types of marsh plant reflect different environmental conditions,different soil and water sources and marsh waterbody existence form.The extractive experiments of Carex lasiocarpa marsh, Carex pseudocuraica,peat and its quality were done with the method of optical image processing.Applying Landsat information of multitemporal images,through optical enhancement,enhancement of digital images based on band,image classification processing and false colour composite technology image of various marsh types have been extracted satisfactorily.The marsh types,the range of distribution and the area of various marsh types were determined by field investigation and verification.Thereby,it is obvious that remote sensing is an effective method in marsh survey of vast extent into which human being go difficultly.