铁尾矿高-多光谱遥感特征分析与信息识别

Feature analysis and information identification of the iron tailings by high-multispectral remote sensing

研究选取河北省不同粒径、不同干湿状态及不同矿区的铁尾矿样本进行高光谱测量,对比分析了不同状态下的铁尾矿高光谱特征,通过拟合分析确定了铁尾矿高光谱识别的有效窗口.结果表明:不同粒径和不同矿区的铁尾矿样本,光谱值在500~600 nm范围内差异较小,不同干湿状态尾矿样本在500~600 nm反射率数值差别明显;若将尾矿分干湿两类分别提取时,光谱区间500~600 nm是一个铁尾矿光谱特征受粒径、湿度和矿床类型等因素影响较小区间,适合铁尾矿多光谱遥感信息提取的主要识别窗口.通过多光谱遥感影像在DT-tailings(DT:Decision Tree)和DT-tailings-B2两个决策树判别规则信息提取对比试验中,DT-tailingsB2规则下提取结果在湿尾矿、干尾矿、露天矿和山体的提取精度上均有提高,分别提高了5.00%、22.04%、4.49%和19.59%.验证了铁尾矿高-多光谱遥感信息提取有效性.

The iron tailing samples of different particle sizes,from different dry and wet conditions and different mining areas in Hebei Province were selected for high-spectral measurement in the study.The high-spectral features of iron tailings from different conditions were compared.Then the effective window of high-spectral recognition of iron tailings was determined by fitting analysis.The results show that:firstly,as for the iron tailing samples of different sizes from different mining areas,no significant spectral value difference is found within the range of 500—600 nm;while significant reflectance value difference within the range of500—600 nm is observed concerning iron tailing samples from different dry and wet conditions;secondly,if the iron tailings are divided into dry and wet types and analyzed accordingly,the spectral range of 500—600 nm is proved to be an area where the spectral characteristics of iron tailings are only slightly affected by factors such as particle sizes,humidity and deposit types,which is a primary recognition window,suitable for multispectral remote sensing information extraction of iron tailings.After the comparative tests of discriminating rule information extraction through multi-spectral remote sensing imagery in the DT-tailings and DT-tailings-B2,the resulting extraction accuracy by DT-tailings-B2 rule is found to have increased in wet tailings,dry tailings,and ores from open pit mines and mountains;the extraction accuracy increases by 5.00%,22.04%,4.49%and 19.59%respectively.The results verify the validity of high-multispectral remote sensing information extraction from iron tailings.