Recognition Of Aquaculture Types In Coastal Zone Based On Object-oriented Method

Aquaculture has taken a rapid development in China, especially in coastal areas. The productivity assessment, environmental impact and planning of aquaculture have been one of hot topic of ecology and environment research.There is a long coastal zone in Guangdong province, aquaculture area is widely distributed. However, its conventional survey is time-consuming and cost-expensive, it is difficult to update. The remote sensing image provides the possibility to extract information quickly. This research is focused on the identification and extraction of aquaculture area based on SPOT-5 and TM image by using object-oriented classification method. Using multi-scale segmentation, and the based on the spectral information, shape, texture etc, the process and rules are constructed. Comparing to conventional methods, it takes full advantages of feathers, which improves the accuracy of classification. At the same time, the paper shows that the accuracy of aquaculture area based on TM image is a little better than based on SPOT-5. So it indicates that the choice of image date depends on the study aim and study area.After extraction of aquaculture area, this paper try to recognize different types of aquaculture area such as intensive pond, middle intensive pond and extensive pond. We combine the texture characters of SPOT-5 image with the spectral characters of TM image to use BP neutral net to recognize them automatically. The result shows that this method to extract and investigate the aquaculture area is feasible.