Research On Water Bodies Extraction Method Of Nine Plateau Lakes In Yunnan Province Based On Multi-Source Remote Sensing Data

Lakes are more sensitive to changes in global ecological environment,and their changes reflect regional ecological environment to certain extent.The nine plateau lakes in Yunnan Province play a significant role in the development of major cities in the province,the implementation of green energy and food brand strategy,the advancement of tourism,the protection of biodiversity,etc.Thus,a timely and accurate grasp of the change information of these lakes can be taken as a scientific reference when formulating relevant policies.The nine plateau lakes in Yunnan are important components of the lake ecosystem in China.Research on the extraction methods of the nine plateau lakes and real-time monitoring of changes in the water bodies of these lakes are of great significance to Yunnan Province's construction of an ecological security barrier facing the southwest,which can also be taken as the technical reference for monitoring similar lakes in other regions.Therefore,how to accurately,timely and automatically extract the water body's change information of the nine plateau lakes in Yunnan Province based on multi-source remote sensing data and visualization technology has far-reaching practical significance on the ecological environmental protection and harmonious economic and social development of the nine plateau lakes.Based on a comprehensive analysis of current research on extracting water body boundaries through remote sensing data,and in order to solve issues and defects such as fuzzy pixels on the lake's water and land boundaries as well as insufficient use of spectral information of neighboring pixels,sensitivity to the “noise points” of remote sensing images and low accuracy of water body extraction when extracting lake water body through traditional fuzzy clustering algorithms,this paper proposes an improved FCM algorithm,which uses the combination of Mahalanobis distance and spectral angle matching to replace the Euclidean distance of the traditional FCM algorithm,so as to fully take into account the spectral information of neighboring pixels,thereby improving the clustering effect and accuracy.In this paper,the nine plateau lake basins in Yunnan Province were taken as the research area,the Sentinel-2 and ZY3 satellites image data during three periods,namely wet periods(June),dry periods(December)and normal periods(March)from 2016 to 2020 were taken as the data source,and traditional FCM clustering algorithm,improved FCM algorithm,K-means clustering algorithm and ISODATA clustering algorithm were employed to extract and study the water information of the nine plateau lakes.Meanwhile,the effect and accuracy of water extraction for the nine plateau lakes were evaluated and analyzed from aspects of water periods,clustering algorithms and research scales.Also,the water body information extraction visualization research was conducted based on the water body's extraction results.Further,by taking the improved FCM algorithm which had the optimal water body extraction accuracy and effect as the core algorithm,an automatic extraction platform of lake water body with general module functions such as data input,parameter setting,data output,result display,etc.was developed.The main research conclusions are as follows:(1)The extraction effect and accuracy of the nine plateau lakes are different depends on water volume periods.Among them,the differences on extraction accuracy of Fuxian Lake during the three water volume periods are relatively small.During the three water volume periods,there are more omission areas in wet periods while less in normal periods.For the commission area,except for Yilong Lake,the difference is not significant.The spatial heterogeneity of water bodies of different lakes in the same area is obvious,and the distribution of omission and commission areas of the same lake in difference water volume periods is different.For the extraction accuracy of difference water volume periods,the kappa coefficient and overall classification accuracy of Chenghai,Erhai and Qilu Lake are quite different.The omission error of Chenghai,Dianchi,Lugu Lake,Erhai,Qilu Lake and Fuxian Lake is generally low.The area with the largest difference in commission error during different water volume periods is located in the Yilong Lake area.The water bodies in the nine lake areas have different reflection characteristics in different seasons,which cause differences in extraction of lake water bodies in different periods.(2)Among the four clustering algorithms,the improved FCM algorithm obtains the optimal extraction effect and accuracy.The four clustering algorithm can roughly extract the lake range,but the improved FCM algorithm is better than the other three clustering algorithms,which can overcome the impact of certain areas of mountain shadows and vegetation on the clustering results.When the improved FCM algorithm is used to extract the water body,the extraction effect of the detailed water body information is cleaner than that of the other three algorithms,and the omission area of the water body is greatly reduced.For extraction accuracy of different clustering algorithms,the kappa coefficient and overall classification accuracy of the improved FCM algorithm can be increased by 6.5 and 4.2 percentage points respectively,and its omission error and commission error can be decreased by 3.3 and 8.1 percentage points respectively.Compared with the other three algorithms,the improved FCM algorithm has significant improvement on water body extraction effect and accuracy in lakes with large water spatial heterogeneity,while the improvement on water body extraction effect in lakes with stable water quality all around the seasons is not significant.(3)The improved FCM algorithm has good scalability and universality under different research scales.The ZY03 satellite based water body extraction of nine plateau lakes is with better extraction accuracy and effect than that of Sentinel-2 image data,the omission area and commission area are greatly reduced and the algorithm has sound scalability.The extraction accuracy of the four clustering algorithms: the improved FCM algorithm extracts water body based on ZY03 satellite data;compared with the other three algorithms,the kappa coefficient and overall classification accuracy of the improved FCM algorithm can be increased by 6 and 3 percentage points respectively and its omission error can be decreased by 3 percentage points.It proves that the improved FCM algorithm proposed in this paper still has proper migration ability under a small research scale.(4)Construction of lake water body extraction platform based on the improved FCM algorithm.Based on the optimal improved FCM algorithm in the extraction results of the lake water bodies as the core algorithm of the visualization platform,the platform is visualized by using the Python language and the third-party library of Py Simple GUI.The platform is with general module functions such as data input,parameter setting,data output,result display,etc.,which can be employed to realize the automatic extraction of the lake water body.