Feature extraction and classification of coal mine overburden dumps based on data merging of multi-source remote sensing

Feature extraction of coal mine using hyperspectral data remote sensing (RS) images was help for decision-making of revegetation and could provide technology reference for further researches. Object spectral characteristics of coal mine overburden dumps were quantitatively analyzed and classified based on multi-source remote sensing data. The hyper-spectral data were collected and the surface soil and vegetation were measured in Fuxin opencast coal overburden dumps. According to the analysis of remote sensing images, different bands images of SPOT-5 were used to classify the surface objects. The results indicated that the classification accuracy was not improved obviously using bands combination and merging. However, the evaluation accuracy can be improved effectively by combining four SPOT5 multi-spectral images, four vegetation index (VI) images and the hyper-spectral characteristics curves of the ground objectives, which achieved the total accuracy 85.48% and Kappa coefficient 0.8197. The results satisfied the needs of the objects investigation of overburden dump and provided the references for gradation of revegetaion.