Study On Spectral Analysis Method And Remote Sensing Application Of Coal And Coal Gangue

There are a large number of coal mining areas in China,in which the coal and gangue is the main solid deposits.The accumulation of coal gangue could lead to not only the land occupation but also the environmental problems and disasters,such as self-ignition,landslide,and collapse,etc.Therefore,it is an important and practical issue for the effectively and rapidly identification and dynamic monitoring of coal and gangue by remote sensing technology.In this paper,both the visible,near-infrared and thermal infrared spectral of coal and coal gangue samples were tested and analyzed firstly.And then the identification,classification and inversion models of coal and gangue were established,after this the spatial distribution of coal and gangue was extracted by satellite remote sensing in the coal mining area,and finally the identification and monitoring of coal and coal gangue were realized.The contents and contributions to this paper are concluded as follows:(1)The visible,near-infrared spectral characteristics of the main coal types in China were measured and analyzed,and the advantages of different coal classification methods were compared.The visible,near-infrared spectra of the coal samples acquired from 14 coal mines in China were measured.And three types of typical coal were classified by four classification methods,including the MAO model,random forest algorithm,BP neural network algorithm,and ELM algorithm.The advantages of the four classification methods are compared by the classification accuracy and time-consuming performance,in order to formulate the principle of optimization for different remote sensing applications.(2)The coal and gangue classification method based on the visible,near-infrared and thermal infrared spectra was proposedThe visible,near-infrared spectra of the coal and gangue were measured and the spectral features were analyzed.The results illustrate that the spectral features of most gangues are quite different from that of the coals,whereas some of them are still difficult to be distinguished from it.Therefore,the method combined near-infrared and thermal infrared spectra was proposed to classify the coal and gangue.The classification accuracy of the combined method is 99.2%,which is greatly improved compared with the classification accuracy of 92.9%using visible,near-infrared spectra.(3)The classification of the burned and unburned gangue based on visible,nearinfrared spectraThe visible,near-infrared spectra of the burned and unburned gangue were measured and the spectral features were analyzed.There are significant differences between the spectra in visible bands.The slope of the spectral curve of the burned gangue is overall higher in the 350?750 nm range,and surges obviously to the 550?630 nm range,while there are no such characteristics on the spectra of the unburned gangue.Based on the spectral feature differences and Landsat 8 OLI data,the NDGI index was constructed to distinguish the burned and unburned gangues.The results showed that the identification accuracy of the burned and unburned coal gangue samples is 99.1%based on the NDGI index,which is higher than the accuracy of the random forest algorithm with the value of 95.2%.The burned and unburned areas in coal gangue can be distinguished by the NDGI index calculated from the satellite image.(4)Quantitative inverting of fixed carbon content in coal gangue by the thermal infrared spectraThe experimental results showed that the thermal infrared spectroscopy characteristics of carbonaceous gangues are significantly different from those of coal samples,and the fixed carbon contents are firmly correlated to the trough of the spectra.The linear model of the fixed carbon content and the spectral difference indexes was established and used for the inversion of the fixed carbon content.The results show that the linear correlation between the DI models is 0.867.Compare with the prediction results of the fixed carbon content of coal gangue calculated from the absorption depth,spectral absorption area,random forest,and support vector machine algorithm,the DI-based inversion model possesses the best prediction accuracy.The average error are 5.00%and the root mean square error is 6.70%,which indicates the effectiveness of this method of the prediction of the fixed carbon content of coal gangue.