基于无人机偏振遥感的水稻冠层氮素含量反演模型

受水稻冠层几何结构的影响，传统的无人机高光谱获取到的反射光谱信息中包含与水稻内部组成物质无关的镜面反射信息，从而影响水稻氮素含量的反演精度，因此在利用无人机获取水稻冠层反射光谱信息时，有必要考虑通过偏振测量技术去除反射光谱中的镜面反射分量，进而实现提升水稻氮素含量反演精度的目的。基于无人机偏振遥感测量得到的水稻分蘖期多角度偏振光谱数据和与之对应的氮素含量数据，采用植被指数方法分析二者之间的相关性，得到了水稻冠层偏振光谱数据与其对应氮素含量相关性最高时对应的角度，选取该观测角度下的偏振光谱数据，利用连续投影法(Successive projections algorithm, SPA)提取特征波段，在此基础上，基于数学变换的方法，提出了构建植被指数的新思路，构建了由2个波段组成的偏振光谱植被指数(Polarisation spectrum vegetation index, PSVI),并利用线性回归方法建立水稻冠层氮素含量的反演模型。结果表明，通过对不同观测天顶角下水稻冠层偏振光谱数据与氮素含量相关性分析，得到最佳观测角度为-15°(后向观测15°);利用连续投影法提取得到该角度下偏振...

Inversion Model of Nitrogen Content of Rice Canopy Based on UAV Polarimetric Remote Sensing

Due to the geometry of the rice canopy, the reflectance spectral information obtained by conventional UAV hyperspectroscopy contains specular reflection information which is not related to the internal composition of rice, thus affecting the inversion accuracy of the nitrogen content of rice. The inversion accuracy of rice nitrogen content was improved by removing the specular reflection component from the reflectance spectra. Based on the multi-angle polarimetric spectral data of rice tillering stage and the corresponding nitrogen content data obtained from UAV polarimetric remote sensing measurements, the correlation between them was analysed by the vegetation index method, and the angle with the highest correlation between the polarimetric spectral data of the rice canopy and its corresponding nitrogen content was obtained. The polarisation spectrum vegetation index (PSVI) was constructed based on a mathematical transformation method. The inverse model of the nitrogen content of the rice canopy was developed by using a linear regression method. The results were as follows: the best observation angle of -15° (15° for backward observation) was obtained by analyzing the correlation between the polarisation spectral data and the nitrogen content of the rice canopy at different observation zenith angles; the six characteristic bands of the polarisation spectral information at this angle were extracted by the continuous projection method, specifically 500nm, 566nm, 663nm, 691nm, 736nm and 763nm; the mathematical transformation idea was applied to the polarization spectral vegetation index (PSVI), consisting of 500nm and 566nm was constructed; the PSVI was used as the model input, and the linear regression method was used to establish the inversion model of nitrogen content in the rice canopy. The inversion results were better than the inverse models of nitrogen content constructed by difference vegetation index (DSI), ratio vegetation index (RVI) and other common vegetation indices. In conclusion, based on the polarization spectral data of rice tillering stage acquired by UAV and using PSVI vegetation index as model input, the accuracy of inversion of nitrogen content in rice canopy can be improved.