大面积高光谱遥感监测是稀土矿区环境监管的重要手段, 复垦植被在矿区环境胁迫下的特征变异分析, 可为准确实现矿区生态恢复动态监测提供必要基础。 通过实地采集稀土矿区六种典型复垦植被及其对应正常环境植被叶片原始光谱, 对照分析其光谱变异。 将原始光谱进行常用的导数变换之外, 还应用信号处理中的分形维数计算、 离散小波变换分析技术和短时傅里叶变换处理放大植被叶片光谱的细部信息, 探究复垦植被在稀土矿区环境胁迫下的光谱特征。 结果表明: （1）在一阶导数光谱中, 除湿地松外, 其他植被均出现“红边位置”的蓝移现象, 表明了复垦植被在矿区受到不同程度环境胁迫等外界因子的影响。 （2）通过计算矿区植被光谱曲线的分形维数, 得到同种复垦植被分形维数高于正常植被的规律, 说明矿区环境胁迫多条件因素的影响致使复垦植被光谱曲线的波形变复杂。 （3）植被叶片光谱经过离散小波变换, 其中原始光谱离散小波变换最佳细节系数为d5, 一阶导数光谱离散小波变换最佳细节系数为d6; 并且一阶导数光谱离散小波变换在更小的尺度下放大了光谱特征细节差异, 取得更好的效果。 （4）光谱通过短时傅里叶变换在空频图上实现局域化, 原始光谱空频特征出现在“红边”与中红外第一个“波谷”处 , 而一阶导数在更小的尺度上, 更多的波段放大并增加了光谱曲线空频特征。 总体而言, 将信号处理方法应用于光谱处理, 较导数变换能获取更多光谱特征, 其中短时傅里叶变换以获得光谱空频特征的特点又优于分形维数计算和离散小波变换分析技术。 该研究为稀土矿区复垦植被生理参数反演和复垦效果监测提供技术支持, 有助于稀土复垦矿区的生态重建。

Large-scale hyperspectral remote sensing monitoring is an important means of environmental supervision in rare earth mining areas, and the analysis of characteristic variation of reclaimed vegetation under environmental stress in mining areas provides a necessary basis for accurate dynamic monitoring of ecological restoration in mining areas. The original spectra of six typical reclaimed vegetation and their corresponding normal environment vegetation leaves in rare earth mining areas were collected on the spot, and their spectral variations were compared and analyzed. In addition to subjecting the original spectra to the usual derivative transform (DT), fractal dimension (FD) calculations in signal processing, discrete wavelet transform (DWT) analysis techniques, and short-time Fourier transform (STFT) processing are applied to amplify the detailed information of vegetation leaf spectra to investigate the spectral characteristics of reclaimed vegetation under environmental stress in the rare earth mining area. The results show that: (1) In the first-order derivative spectra, all vegetation except wetland pine show a blue shift in the “red edge position”, indicating that the reclaimed vegetation is affected by external factors such as environmental stress to varying degrees in the mining area. (2) By calculating the FD of vegetation spectral curves in mining areas, the FD of the same species of reclaimed vegetation is higher than that of normal vegetation, indicating that the influence of multiple conditions of environmental stress in mining areas will cause the waveforms of reclaimed vegetation spectral curves to become complex. (3) The vegetation leaf spectra are discrete wavelet transformed, where the best detail coefficient of the original spectral DWT is d5, the best detail coefficient of the first-order derivative spectral DWT is d6, and the first-order derivative spectral DWT amplifies the difference in spectral feature details at a smaller scale, achieving better results. (4) The spectra are localized in the null-frequency diagram by the STFT, with the original spectral null-frequency features appearing at the “red edge” and the first “trough” in the mid-infrared, while the first-order derivatives amplify and increase the spectral curve null-frequency features at smaller scales and in more bands. In general, applying signal processing methods to spectral processing can obtain more spectral features than the DT, where the STFT is superior to FD calculation and DWT analysis techniques in terms of spectral mapping into null-frequency features. The study results provide technical support for the inversion of physiological parameters of the reclaimed vegetation in rare earth mining areas and the monitoring of reclamation effects, which will help the ecological reconstruction of reclaimed mining areas.