Abstract:In view of the fact that the current remote sensing supervision cannot meet the regulatory requirements of high precision, all-round and short period, especially in the field of rapid supervision of impaired water activities, there are still obvious shortcomings. Satellite remote sensing technology is used to obtain high-resolution remote sensing image water dody data set, based on deep convolutional neural network, water body is extracted from high-resolution remote sensing images and intelligent monitoring of impaired water body is carried out. An Improved Water Detection Network ( IWDNet ) model based on the proposed U-Net. Firstly, based on the U-Net structure, skip multi-scale feature fusion is introduced, and different Multi-scale Feature Fusion Attention Mechanisms ( MFFAM ) modules are generated by combining SE, ECA and CBAM attention mechanisms for comparison. The dilated convolution is introduced to expand the network receptive field, and finally the recognition and detection of black and odorous water bodies are realized. Experiments show that the MFFCBAM-IWNet model based on skip multi-scale fusion and CBAM attention mechanism effectively improve the recognition accuracy, and performs best on the high-resolution remote sensing image water body data set. The overall accuracy is 98.56 %, and the Kappa coefficient is 0.9784. A total of 30 suspected damaged water points were detected in Guangdong Province and other places, including 27 suspected impaired rivers and 3 suspected impaired lakes. The research results can provide data supports for intelligent supervision of ecological environment remote sensing.
