Remote Sensing Monitoring Of LUCC In Mining Area Based On Pixel Unmixing

Coal mining plays an important role in urban development,industrial development and the process of modernization in China.Along with the rapid development of the scale of coal mining fields,serious damage to an increasing serious environmental disaster,such as mining subsidence,soil pollution etc..Therefore,it possesses important practical meanings to monitoring the land use and land cover change(LUCC)in coal mining areas.As an advanced information collecting method,remote sensing has obvious advantages in LUCC researches and has become one of the mainly methods on monitoring of the LUCC in coal mining areas.However,there exist a lot of mixed pixels in remote sensing images of mining fields due to the limited the spatial resolution of sensors and the complicated ecological environment in mining area.The precision of result often does not meet practical requirements when the mixed pixel corresponds to one land cover type simply.Thus,it is necessary to enhance the accuracy of LUCC detection and analysis with the help of mixed pixel decomposition technology.Endmember extraction and depixel process are the key factors in pixel unmixing,especially the former one.Endmember extraction has always been the difficulty and the focus in mixed pixel interpretation as the quality of extracted endmembers may affect the accuracy of LUCC monitoring result directly.In this study,the existing problems in monitoring LUCC taking advantage of remote sensing image in mining area as well as difficulties in pixel unmixing were analyzed.Improved GSA is introduced to optimize endmember extraction.The area calculation and change analysis of land covers is quickly and accurately realized with the utilization of TM imagery.Main researches of this paper are the following.1 Theoretical study and algorithm realization is achieved in aspects of unmixing model,endmember number detection,endmember extraction algorithm,abundance estimation,and accuracy evaluation methods.The problems existing in available endmember algorithm are also analyzed.2 Repulsive force is introduced to GSA after deep analyze of GSA and other intelligent algorithm.Repulsive force based GSA(RFGSA)is achieved by introducing fuzzy rules to control repulsive force self-adaptively and RFGSA is applied in optimizing endmember extraction method.3 Area statistic and change analysis of different land covers in the study area are achieved by abundance image taking advantage of abundance estimated by the extracted endmembers.