南疆红枣多尺度高光谱偏振定量遥感模型研究

南疆红枣产业规模化，信息化发展，迫切需要发展定量遥感技术。辐射定标精度是定量遥感的核心问题。不同结构特征对定标精度有较大影响，光谱偏振态表征精细结构差异。本项目通过仿形辐射定标和多角度光谱偏振方法提高定量遥感精度。首先，依据枣树冠层树形及红枣生长特点。通过标准板多方位近似，对枣园空间网格化、枣树结构仿形等方法，进行立体化辐射定标。获得枣树仿形结构辐射定标曲线。其次，通过多尺度高光谱偏振实验，特征波段偏振参量dolp和orient,进行高光谱图像偏振特征彩色增强，分别建立不同距离，不同角度的室内红枣偏振背景谱。获得不同尺度结构特征与红枣品质光谱偏振态对应经验公式。结合前期红枣品质光谱模型，通过光谱重构和模型传递，建立多尺度红枣品质定量遥感模型。对枣园红枣品质整体分布情况进行立体差异化表达。持续增加不同时期枣园偏振实验，对模型预测结果修正提高精度。该成果在大范围遥感红枣品质方面有广泛应用前景。

The quantitative remote sensing technology is urgently needed for the scale and information development of the red date industry in southern xinjiang. Radiometric calibration accuracy is the core problem of quantitative remote sensing. Different structure characteristics have great influence on the precision of calibration. In this project, the accuracy of quantitative remote sensing is improved by copying radiometric calibration and multi-angle spectral polarization. First, according to the canopy shape of jujube and jujube growth characteristics. In this paper, the space grid of jujube garden and the structure of jujube tree were measured by standard plate multi-azimuth approximation. The radiative calibration curve of the copying structure of jujube tree was obtained. Secondly, the multi-scale hyperspectral polarization experiment was conducted, and the polarization parameters dolp and Orient in the characteristic band were used to enhance the polarization characteristic color of the hyperspectral image, and the indoor red date polarization background spectra with different distances and angles were established respectively. The corresponding empirical formulas of different scale structure characteristics and polarization states of quality spectra of jujube were obtained. A multi-scale quantitative remote sensing model of quality of jujube was established by spectral reconstruction and model transfer. The overall quality distribution of jujube in jujube garden was expressed in a stereoscopic and differentiated way. The polarization experiments of zaoyuan in different periods were continuously added to improve the accuracy of model prediction. The results have wide application prospects in the quality of jujube.