|
Water body is one of the most important natural elements in coastal zone. Water bodies in coast are subdivided into offshore sea, aquaculture ponds, inland water bodies, river and so on. Remote sensing is an effective tool to obtain coastal typical objects with high spatial resolution imageries. This paper aims at existing problems of object-based image analysis application to monitor resources and environment in coastal area. For object-based recognition for water body types, relevant works have been carried out by adding spatial semantic features to the extraction process. Through analyzing the spectral, spatial and texture features of water body, the rule set for extracting water body type is established based on the topological and contextual relationship between segments. The recognition method of water body types proposed in this paper gets rid of the traditional object-based classifications based on statistical law. Using prior knowledge to construct knowledge rules with spatial semantic information makes spatial distribution characteristics in coastal zone effective in improving the accuracy of type identification.
|
