An object-oriented approach to map wetland vegetation:a case study of yellow river delta

Remote sensing is an important and effective tool for mapping and monitoring wetland vegetation condition and change. Because providing more spatial characteristics except spectral information, high spatial resolution images have become the main source for resources and environmental management and application. However, the conventional pixel-oriented image analysis techniques have much difficult in extracting information from high resolution images. The paper presents a case study of detecting wetland vegetation information in the Yellow River Delta based on the SPOT-5 image. Going far beyond the methodical limits of per-pixel and manual interpretation approaches, an object-oriented image analysis technique is used for extracting information and classifying the vegetation types. In the process, the SPOT-5 image is segmented into highly homogeneous image objects at multiscale and a network of image objects is generated. Not only spectral information but also spatial and contextual characteristics of image objects are used for classification, which is conducted by fuzzy logic, and image objects are evaluated using membership function classifiers. Membership functions are used to produce class description, which consists of a set of fuzzy expressions from appropriate object features. The results of vegetation information extraction are promising and the classification accuracy is significantly higher than supervised classification.