煤岩界面的高光谱识别原理

Study on the principle of hyperspectral recognition of coal-rock interface

为研究以物性成分为辨识依据的煤岩高光谱识别技术,对来自我国不同煤矿生产线的煤岩试样在350~2 500 nm波段范围进行了反射光谱的采集。通过分析代表性样品的光谱反射率曲线,得出了煤岩的主要吸收谱带,发现了岩在2 200 nm附近表现为强吸收,而煤在此波长点附近吸收不明显,原因为岩中含Al-OH振动结构的矿物含量较高,而煤中此类矿物含量较低。以此2 200 nm附近煤岩吸收差异性为煤岩识别的基本原理,通过初步预处理和包络线去除预处理的方法,在全波段和2 150~2 250 nm吸收谷特征谱带,采用了4种识别算法模型,对训练集光谱数据进行训练,预测测试集光谱类型。测试集试样类型总体识别精度达到90%左右,且具有较好的一致性,识别速度达到毫秒级,实时性好,这些原理和识别方法为实际工程应用提供了参考。

In order to study the technology of hyperspectral recognition of coal and rock based on their components,the reflectance spectra of coal and rock samples from production lines of different coal mines in China were acquired in the 350-2 500 nm range.By analyzing the spectral reflectance curves of the representative samples,the main absorption bands of coal and rock were obtained.It was found that the spectral reflectance curves of rocks are strongly absorbed near 2 200 nm while the absorption of coals near this wavelength point is not obvious.The reason is that Al-OH vibration structure-contained minerals are highly contained in rock while coal contains less this type of minerals.The difference of absorption between coal and rock near 2 200 nm was regarded as the basic principle of coal-rock recognition.After preliminary and continuum removal pretreatment,the spectral data of the training set were trained and the spectral types of the test set were predicted using four recognition models in the 350-2 500 nm range and the absorption valley of 2 150-2 250 nm.The overall recognition accuracy of sample types of the test set is about 90%,and the consistency is high.The recognition speed is in milliseconds and the real-time is better.These principles and recognition methods provide a reference for practical engineering applications.