Coral Reef Benthos Information Extraction Based On High Resolution Remote Sensing Images

As one of the most important marine ecosystems,coral reefs play an important role in regulating marine climate and ecosystem stability.With the influence of global greenhouse effect and human activities in recent years,coral reef ecosystem has been greatly damaged.As the most critical factor to maintain the ecological health of coral reef,coral reef benthos also reflect the health status of coral reef ecosystem.Therefore,in order to better monitor the health status of coral reef ecosystem,it is necessary to monitor the distribution of coral reef benthic information.Based on Sentinel-2 data and GF-2 data,the information extraction of coral reef benthos in the coastal area of Sanya is studied.The spectral characteristics and extraction methods of coral reef benthos are used to realize the identification and extraction of coral reef benthos,and provide data support for better monitoring of coral reef ecological health.The main results are as followsBy simulating the spectral response function of visible light range of Sentinel-2image,the benthos such as coral reef,albino coral reef,seaweed,seaweed and sand are recognized,and the detection ability of Sentinel-2 image for coral reef benthos is verified.SVM and Maximum likelihood Classification method are used to realize the simple extraction of coral reef benthos,and the coral reef,albino coral reef and sand are extracted The maximum likelihood method and SVM method were used to extract83.39% and 88.89% of the three simple benthos types,respectively.Based on GF-2images,the optimal segmentation scale of coral reef region was studied by using artificial experience method,Esp scale model method and mean variance method.Finally,the optimal segmentation scale of large-scale and small-scale coral reef regions was 74 and 3,respectively Six benthic types,including coral reef,albino coral reef,algae,algal ridge,sand and rock,were extracted by nearest neighbor method and SVM method.The extraction accuracy of nearest neighbor method and SVM method was 85.39% and89.39% respectively.Through the comparative analysis of the extraction ability of the two images,it is found that GF-2 can achieve more precise benthos extraction on the basis of Sentinel-2.Based on the results of coral reef benthos extracted by GF-2 image SVM method,the coral reef benthos in Sanya coastal area were analyzed from the spatial and quantitative distribution characteristics.The results show that the coral reefs in East hawksbill island are the best,the albinism in Luhuitou and Yulin Bay is more serious,and the threat of other benthos in West hawksbill island is more serious.Figure 33;Table 9;Reference 60...