Method Comparison of Extraction of Gangue Yard Based on Remote Sensing

Gangue,as one of the industry-specific solid waste produced from coal mining and coal washing,effects on around environment significantly.Therefore,monitoring of coal yard is essential for the management and protection of ecological environment.Before gangue yard detailed investigation,obtaining preliminary data of the location and area of gangue yard by remote sensing image is needed and good for subsequent investigations.This article took the image of Landsat5 TM,received in September,2011 as data source and did the radiometric and geometric correction to the images.According to the composition and formation characteristics of gangue,we extracted the gangue yard with following two steps: firstly,got land classification information that is confused with gangue through spectral analysis and unsupervised classification;secondly,combined spectral information and terrain,temperature and other ancillary information of the study region,and used four methods,i.e.,unsupervised classification,supervised classification,spectrum-photometric method and hierarchical classification respectively to extract the gangue yard of the study area.By comparison of the above methods,we found that the unsupervised classification and supervised classification methods had a faster data extraction but with low extraction accuracy.The accuracy of spectrum-photometric method is a little higher than the former two methods.The hierarchical classification method has the highest accuracy in preliminary data extraction,and the identification accuracy of the gangue yard is up to 78% after post-processing.The result basically meets the requirement on dynamic supervision of gangue yard.Thus,these methods are also useful,as guidance,to continuing extract data of the area and location of the gangue yard under high resolution remote sensing images.Especially,the hierarchical classification method is more suitable for gangue yard information extraction.