The method study of extraction of tidal marsh vegetation based on object-oriented in Chongming Dongtan national nature reserve, Shanghai

Chongming Dongtan national nature reserve is an internationally important wetland. Accurate survey of invasive alien vegetation and indigenous vegetation is the foundation of the study on the ecological impact of invasive vegetation, wetland protection and management. The high spatial resolution images have more spatial details and texture characteristics, and offer the possibility to extract information of tidal marsh vegetation. In this study, an object-oriented technology was applied to extract the information of tidal marsh vegetation from QuickBird image. Firstly, the image was segmented at two scales and objects were formed; secondly, the feature space was built using the characteristics of spectrum, shape and expert knowledge; and finally, a combining of near neighboring and fuzzy expert system was used to identify vegetation species. The results show that the classification accuracy for wetland reached 87.11%, higher than maximum likelihood classification by 18.75%.