The coastline change is one of the research bases of coastal evolution and global climate change. Automatic extraction of different coastline types from the remote sensing images quickly and accurately is always difficult,especially for the shoreline of muddy coast. In this paper, the object-oriented method was adopted to build the rule sets including the spectrum transform information and texture feature information, to classify the images in high definition, and then to extract the coastline accurately based on the TM remote sensing images of muddy coast zone in northern Jiangsu in 2006. The results showed that the method can excavate the image information sufficiently with a high degree of automation and fast extraction, and also test the results from two indexes of accuracy and stability independently of the reference coastline.