藏北高原地表砾石粒径高光谱遥感定量反演及空间分异格局研究

The northern Tibetan Plateau grassland degradation trend obviously, forefathers use growth of grassland and soil quality as the research object of remote sensing monitoring, This study uses the reverse thinking mode,takes surface gravel size and spatial pattern as the research object. The gravel is playing a key role in the northern Tibet grassland ecological restoration and Determines the degree of difficulty in the transformation and utilization of grassland animal husbandry. Based on hyperspectral remote sensing technology, we obtained the background data of the gravel surface spectral database and gravel size. These data are used to analyze the variation of the spectral and spectral values due to the position deviation of the wave crest or trough. The best spectral bands and spectral features are selected by the inversion of the surface spectral parameters and the surface gravel size. Based on mathematical morphology and field data of mixed pixel decomposition technique, extract the abundance images of different size endmembers. According to the size of gravel, the hyperspectral remote sensing image and the inversion model of surface gravel particle size are established. Thus the northern Tibet Grassland different gravel size spatial distribution were obtained. Hierarchical Bayesian model of fractal theory is used to calculate the gravel clearance degree and spatial heterogeneity, in-depth analysis of effect of gravel differentiation pattern.

藏北高原草地退化趋势明显，前人多以草地长势和土壤质量作为遥感监测对象，本研究采用反向思维模式，以地面砾石粒径大小和空间格局为研究对象，砾石对于藏北草原生态修复的利用中起到关键性作用，决定了草原牧业改造利用的难易程度。基于高光谱遥感技术，并获取砾石地表光谱库、砾石粒径等本底数据，分析波峰或波谷的位置偏差而造成谱形及光谱值的变化；将地面光谱特征参数与地表砾石粒径反演建模，筛选出显著性相关的最佳波段与光谱特征参数；结合数学形态学和数据场的混合像元分解技术，提取不同粒径端元的丰度影像，将其与砾石粒径相对应，建立高光谱遥感影像与地表砾石粒径反演模型，获得了藏北草原不同砾石粒径的空间分布位置，并利用分形学的多水平贝叶斯模型，计算砾石间隙度和空间异质性，深入分析砾石分异空间格局的效应关系。

藏北高原草地退化趋势明显，前人多以草地长势和土壤质量作为遥感监测对象，本研究采用反向思维模式，以地面砾石粒径大小和空间格局为研究对象，砾石对于藏北草原生态修复的利用中起到关键性作用，决定了草原牧业改造利用的难易程度。基于高光谱遥感技术，并获取砾石地表光谱库、砾石粒径等本底数据，分析波峰或波谷的位置偏差而造成谱形及光谱值的变化；将地面光谱特征参数与地表砾石粒径反演建模，筛选出显著性相关的最佳波段与光谱特征参数；结合改进的PPI、SMACC、N-FINDR端元提取算法和全约束、部分约束、以及无约束最小二乘法解混算法，构建混合像元分解技术体系，提取不同粒径端元的丰度影像，将其与砾石粒径相对应，建立高光谱遥感影像与地表砾石粒径反演模型，获得了藏北草原不同砾石粒径的空间分布位置，并利用机器学习的图像分割、空间自相关、空间变异函数等方法，计算砾石空间异质性与空间分异规律，深入分析砾石分异空间格局的效应关系。