我国煤矿数量众多, 分布广泛, 大量堆积的煤矸石对矿区环境造成严重影响, 其中部分煤矸石处理不当可能引发自燃和爆炸, 对矿区安全构成直接威胁。 根据煤矸石的燃烧状态可以分为燃烧矸石和未燃烧矸石两类, 其存在的安全隐患和对环境的危害性有所不同, 同时其综合利用的途径亦不相同。 因此, 对煤矸石进行燃烧矸石和未燃烧矸石的分类识别与监测就显得尤为重要。 目前的监测方法主要为实地勘查调研, 其效率低、 成本高, 难以满足煤矸石监测的实际需求。 选择辽宁省铁法矿区作为研究区, 首先从矿区矸石山现场采集典型的煤矸石样本106个; 然后, 利用SVC HR1024光谱仪测试其可见光-近红外光谱, 分析燃烧和未燃烧矸石的光谱特征, 并基于可见光波段构建光谱指数NDGI, 用于识别燃烧矸石和未燃烧矸石。 选择实验室测试的光谱数据和实际卫星遥感数据对该指数进行了验证, 并与随机森林法进行对比。 结果显示: 在350～760 nm燃烧矸石光谱曲线斜率整体较高, 在550～630 nm反射率存在陡升现象, 而未燃烧矸石在整个可见光波段光谱曲线斜率较低; 以0.25作为NDGI指数阈值, 可以很好地将燃烧矸石和未燃烧矸石区分开来, 实验室样本验证结果显示, NDGI指数的分类精度可达99.1%, 高于随机森林分类法的95.2%; 现场的验证结果表明, 使用铁法矿区的landsat8 OLI数据, 并基于NDGI指数对矿区内的矸石山进行燃烧和未燃烧区域识别划分, 所提取的燃烧和未燃烧矸石在形态和大小上与Google Earth具有很好地一致性, 表明该指数对于矸石的燃烧状态具有很好识别效果。 在上述研究基础上, 分别取燃烧和未燃烧矸石进行矿物鉴定, 通过对比矸石燃烧前后矿物种类的变化, 分析造成燃烧和未燃烧矸石的光谱特征差异的原因。 结果表明: 燃烧使矸石中的Fe2+被氧化为Fe3+。 Fe3+的大量增加造成光谱曲线在550nm处形成明显的波谷特征, 在整个燃烧过程中生成的玻璃质在750nm处形成高反射率, 二者综合造成燃烧和未燃烧矸石的NDGI指数差异。 研究结果为煤矿区燃烧和未燃烧矸石的区分识别提供了一种快速、 高效、 较为准确的实用方法。

A large number of coal mines are widely distributed over China. Bulk coal gangue deposits seriously affect the mining area environment, and some mishandling of coal gangue may cause spontaneous combustion and explosion, which poses a direct threat to mine safety. The comprehensive utilization of coal gangue can effectively alleviate this problem, and it is of significance to the ecological safety and sustainable development of mine. Depending on the burning state, coal gangue is divided into two types - burned and unburned gangue, whose hidden dangers of security and harm to the environment are different, as well as ways of comprehensive utilization. Therefore, it is very important to do the classification, recognition and monitoring of the coal gangue. The current monitoring methods are mainly the field investigation with low efficiency and high cost, almost impossible for meeting the actual demand of coal gangue monitoring.Tiefa mine in Liaoning Province was chosen as the study area. Firstly, a total of 106 typical coal gangue samples were collected from waste dump in mining areas. Then, SVC HR1024 spectrometer was used to test the visible and near infrared spectrum of samples, and a differential spectral index NDGI was constructed to identify the burned and unburned gangue based on the difference of spectral characteristics of the burned and unburned gangue. Finally, the laboratory spectral data and the corresponding satellite remote sensing images were utilized for verifying the index. The random forest classification method was used as a contrast to the results of the laboratory spectrum treatment. The results showed that the slope of the spectral curves of burned gangue samples was higher ranging from 350 to 750 nm, and the reflectance within range of 550~630 nm increased sharply, while the slope of the unburned gangue in the whole visible bands of spectrum was lower. The threshold of the NDGI index was set as 0.25 to distinguish the burned and unburned gangue. The laboratory spectral data showed that the classification accuracy of the NDGI index is up to 99.1%, higher than that of 95.2% of the random forest classification method. The Field results showed burned and unburned areas of waste dump were distinguished and classified in Landsat8 OLI images based on the NDGI index, and the burned and unburned coal gangue areas were in good agreement with the Google Earth on the morphology and size. The overall results showed that the index can effectively distinguish the combustion states of gangue. In addition, burned and unburned gangue samples were taken for mineral identification respectively. By comparing the changes of mineral species before and after combustion, the cause of spectral difference was analyzed between the burned and unburned gangue. The results showed the oxidation from Fe2+ to Fe3+ of gangue in the process of combustion. A large increase in Fe3+ caused the formation of an obvious spectral valley characteristic at 550 nm band, and a highly reflectance appeared at 750 nm band due to the glass quality generated during combustion. The above conditions cause differences in NDGI index between the burned and unburned gangue. In this paper, the results provide a fast, efficient and accurate model and method for burned and unburned gangue distinguishment in coal mine.