Research On Land Surface Water Extraction Method Based On Medium Resolution Data Of Fengyun Satellite

With the development and popularization of remote sensing,accuratly and rapidly extracting land surface water from remote sensing images is of great significance for the monitoring of land surface water changes.Fengyun satellite image has outstanding advantages such as high time resolution and multi-spectral bands.It is an important image data suitable for high-frequency land surface water monitoring.In this paper,a detailed study on the land surface water extraction method based on the data of Fengyun 3(FY-3)star is carried out.The main contents are as follows:(1)Research on the extraction method of land surface water.The single-band threshold method,water body index method,knowledge decision tree classification method,supervised classification method,unsupervised classification method,spectral matching based on the discrete particle swarm optimization(SMDPSO)and improved spectral matching based on the discrete particle swarm optimization and linear feature enhancement(SMDPSO+LFE)extract the land surface water of Poyang Lake,and cheak samples on the Landsat image with similar time to the FY-3 image,and calculate the classification accuracy by the confusion matrix.The results showed that the overall classification accuracy(OA)of SMDPSO+LFE is 97.64%,and the Kappa coefficient is 0.95.The land surface water extraction results of SMDPSO+LFE are better than the other land surface water extraction methods,especially for the small tributaries and discretes water.(2)Stability and sensitivity analysis of the SMDPSO+LFE method.In order to analyze the stability of SMDPSO+LFE for land surface water extraction in different regions,this paper selectd 8 test sites with different land surface water types,landscapes and topography to extract land surface water.By analyzing the results of the land surface waters of the eight test points,the OA of the eight locations is higher than 94.5%,the Kappa coefficient is above 0.88,and the OA of some regions can reach 99%.The results showed that the method can effectively distinguish different land surface water and background,and have good stability.In order to analyze the influence of parameters in SMDPSO+LFE on the classification accuracy of this method,the sensitivity of reference water spectrum,block size and iteration number in this method is analyzed.The results show that the reference water spectrum is from Lake Superior,sampling block size is 4,and the number of iterations is 2500 times for better classification and efficiency.(3)Application of SMDPSO+LFE method based on FY-3/MERSI data.In this paper,SMDPSO+LFE is applied to the land surface water extraction and change detection in the Poyang Lake area in 2016.The results of land surface water extraction are compared with the remaining GF-1 image surface water body extraction results of the remaining scholars in 2016.The results of the land surface water extraction of FY-3/MERSI and GF-1 have the same trend.The results showed that the method can extract the land surface water of the Poyang Lake research area in a timely manner,which can effectively detect the land surface water changes of the Poyang Lake research area.