Multi-source Remote Sensing Water Depth Surveying In South China Sea Coral Reef Periphery

Water depth surrounding the coral reef is important basic marine data. For study ing the ability of detecting the water depth surrounding the coral reef with the satellite imagery, WorldView-2 and ETM+ multi-spectral satellite images was used as data sources, DongDao coral reef area as the study area. The main results of research on water depth remote sensing detection were as follows:(1)Worldview-2 water depth detection ability analysis experiment. Experiment selected blue, green and blue edge band data of Worldview-2 images, using band combination method, ratio method and BP artificial neural network to detect water depth information. The results show that: The band of WorldView-2 image sorted according to the ability of water depth detection followed by green, blue bands and blue edge band. The responding ARE value is 27%,29% and 36%. WorldView-2 especially sets coast band(blue edge band) to support based on chlorophyll and water seepage characteristic of deep sea exploration research, but water depth remote sensing detection precision is low in the south China sea waters; All kinds of model ARE of the best inversion precision is band combination method 21%,ratio method 28% and BP artificial neural network method 40%.Inversion model with better effect of band group legal, ratio method and BP artificial neural network; In band combination method, Single band model depth detection accuracy is lowest, double-band and three-band combination model depth detection accuracy is higher. Double-band combination model ability of water depth retrieval, blue and green band>green and blue edge band>blue edge and blue band.(2)In order to compare the ability of detecting the water depth with the Worldview-2 multispectral image, we download the ETM+ multi-spectral image data, and choosen blue, green, red band which is sensitive to water depth information, using the same three methods as above to carry out experiments. Results showed the depth values of the three methods, the ARE values ranged from 30% to 67%,which showed a large difference compared with the WorldView-2 results. So the quality of the image would affect the precision of depth detection,WorldView-2 images had a good performance in detecting the water depth arounding the south China sea areas.(3)The inconformity of seabed sediment type, leading to the different contribution to the visible light attenuation coefficient, and cause an deviation of the depth detection results. The article adopt image incision method to divide the seafloor sediments into the reef and reef sand two categories, Then using band combination model to carry out water depth detection research. Results show that the accuracy of detection with the seafloor sediments classification method is improved obviously. Double-band combination model had a better accuracy relative to three-band combination model, The accuracy with blue and green double-band combination model improve the accuracy from the original 26% to 9%,the overall accuracy is 11%.That is because three-band combination model itself has the function of the weaking and eliminating the influence of the seabed sediment inequality. Therefore, the accuracy of ascension is not obvious. The sediment classification method in combination with blue and green band combination model for water depth detection is the best.Due to the traditional measurement methods can not meet the needs of modern large-scale water depth detection, developing water depth remote sensing detection technology has important significance. The method with multispectral satellite images to detect the coral sea area water depth can play a positive role to promote the sea water depth mapping.