A study on monitoring land use/cover change of mining area based on ticket-voting SVM classification

Based on the development of classification algorithm applied in monitoring spatio-temporal dynamic changes of coal-- mining areas, several improvements were made on feature space and classification model in this paper. There were two innovations in our study: 1) During building the feature spaces, a new index for extracting information about mining area was created, which can classify mining area and settlements efficiently; 2) a special ticket-voting SVM algorithm with wavelet kernel function was proposed, which provides higher classification accuracy than other traditional classifiers via the secondary classification. Here we took the northeast plain of Pei county in Xuzhou city as a studying region, applying the proposed method to implement the classification by using the image of multi-temporal TM/ETM from the year of 1987 to 2013. How to carry on deep analysis combined with various non-spatial data is much more significant. Then we studied the rules of dynamic changes of land use/cover and further analyzed their driving factors by combining RS interpretation with GIS spatial analysis techniques. In this study, image recognition technology was applied to the problems of environmental change in coal mining area. These explanations provide some valuable supports for human to recognize and deal with the conflicts between economic development and environmental protection in coal mining areas.