锰矿胁迫植被群落光谱变异机理及高光谱遥感识别模型研究

Remote sensing of vegetation coverage area prospecting prospecting breakthrough is difficult, with high Spectroscopy stress canopy spectral variability and thus indirectly seeking superficial concealed ore, must require images with a high enough spectral resolution and spatial resolution, and meet the high signal to noise ratio, which is often very difficult to meet! These techniques predict bottlenecks increased the uncertainty of remote sensing, and also restricting the application of hyperspectral remote sensing prospecting results. The project another way, from the ground measured spectrum to stress vegetation communities as the starting point, subject to heavy metal stress by constructing a fine time spectrum of vegetation databases and leveraging their dynamic spectral feature extraction, analysis of the spectral vegetation manganese stress variation mechanism, and explore spectral variation caused by manganese trace element geochemistry of vegetation index is established based on spectral characteristics of high manganese stress vegetation target recognition model, which effectively identify for manganese are indicative of coercion and delineation of target vegetation, abundant vegetation coverage area of hyperspectral remote sensing geochemical prospecting theoretical system.

植被覆盖区的遥感找矿是找矿突破的难点，借助高光谱技术研究胁迫植被冠层的光谱变异从而间接寻找浅表隐伏锰矿，必须要求影像具备足够高的光谱分辨率和空间分辨率、且满足较高信噪比，这往往又很难满足！这些技术瓶颈增加了遥感预测的不确定性，也制约了高光谱遥感找矿的应用效果。本项目另辟蹊径，从地面测谱上，以胁迫植被群落为切入点，通过构建受重金属胁迫植被的精细时间谱数据库，并借助其动态光谱特征的提取，分析锰矿胁迫植被群落的光谱变异机理，探索引起锰矿植被光谱变异的微量元素地球化学指标，建立基于锰矿胁迫植被高光谱特征的目标识别模型，从而有效地识别对锰矿有指示作用的胁迫植被群落而圈定靶区，丰富植被覆盖区高光谱遥感地球化学找矿的理论体系。

为了研究锰矿胁迫植物毒化后引起的光谱变异特征及变化规律，优选出代表性植物青葙(沉积型锰矿分布广泛的伴生超富集喜锰植物)，进行不同胁迫地球化学环境下的生长培育实验并测谱观测其物候性，通过监测胁迫环境下Mn含量梯度变化引起植物生物叶片参量的改变进而导致光谱特征变异，以寻找锰胁迫植物的光谱变异浓度效应和时间谱效应。 . 浓度效应上在多阶段大量试验后，优选设置锰胁迫浓度梯度至15个(前5个浓度以矿区与背景土壤混合配比；后10个浓度加入不同胁迫剂量MnCl4配制)。以精细光谱仪监测其整个生长周期内地物波谱变化特征及异常表现，对监测数据进行光滑、去包络线、谱微分等方法提取波谱变异特征参数；胁迫土壤和青葙样品加工后测定37个微量和2个常量金属元素；植物叶片测定非蛋白巯基、还原性谷胱甘肽以及植物螯合肽等3种生物参量。探索SPA-PLSR、多元偏回归等算法对锰矿区胁迫植被群落高光谱弱信息异常提取，建立青葙胁迫光谱特征与地球化学因子间的反演模型。从反射谱曲线、导数谱及吸收参量中提取出11种光谱特征参数，与经过因子分析的23种重金属元素组成地化因子进行模型重建,得到出现斑点和变黄等外在特征时的饱和浓度阈值。. 浓度效应研究旨在为矿区外围普查时寻找敏感光谱波段或组合。. 时间谱效应上提取反射波谱22种特征参量和25种植被指数，经机器深度学习后优选出能有效表达青葙体内含锰量变化的绿峰振幅波谱特征值和R680/R440植被胁迫指数并建立了时间序列上的估测模型，运用连续投影算法(SPA)提取其特征向量，将1603个波段降维成8个敏感波段，取其导数谱并通过偏最小二乘回归(PLSR)建立敏感波段特征参量与土壤锰浓度间的映射模型，表明绿峰面积和振幅参数都随时间发生了先增后减的规律性变化，随着胁迫浓度增加青葙的波谱反射率逐渐降低、红谷变浅、绿峰与红谷间坡度越来越缓；时间谱还表现为胁迫浓度一定，序列随青葙生长出现先增后减的周期性波动，序列拟合且呈递减趋势。. 时间谱效应旨在利用胁迫植物找矿时探索最佳监测的时间窗口。