基于Roujean和Ross-Li模型算法的不同户外光照南疆冬枣BRDF特性研究

不同光照下,如何消除或减小反演数据差异,提高检测精度,是目前南疆冬枣户外检测中遇到的一大难题,因此通过用高光谱相机获得的南疆冬枣二向反射分布函数(BRDF)测量值,采用最小二乘法拟合Roujean模型和Ross-Li模型的参数,最后对比Roujean模型和Ross-Li模型反演的结果,提出何种天气何段波长用何种模型反演效果最好的建议,实验结果表明:(1)多云天气,反演南疆冬枣线偏振度(Dolp)时,Ross-Li模型的R^( 2)是0.9748,Roujean模型的R^( 2)是0.9699;反演南疆冬枣强度分量时,Ross-Li模型的R^( 2)是0.9723,Roujean模型的R^( 2)是0.9749。阴天反演南疆冬枣Dolp时,Ross-Li模型的R^( 2)是0.9651,Roujean模型的R^( 2)是0.9778;反演南疆冬枣强度分量时,Ross-Li模型的R^( 2)是0.9420,Roujean模型的R^( 2)是0.9688。晴天反演南疆冬枣Dolp时,Ross-Li模型的R^( 2)是0.9655,Roujean模型的R^( 2)是0.9262;反演南疆冬枣强度分量时,Ross-Li模型的R^( 2)是0.9285,Roujean模型的R^( 2)是0.8331。整体反演的最佳方案是多云天气下的南疆冬枣Dolp用Ross-Li模型反演,强度分量用Roujean模型反演;晴天南疆冬枣Dolp和强度分量均用Ross-Li模型反演;阴天南疆冬枣Dolp和强度分量均用Roujean模型反演。(2)多波段反演的最佳方案是:多云天气下,反演南疆冬枣强度分量时,波长为1000~1100 nm范围,需用Ross-Li模型,波长为1450~1600 nm范围,需用Roujean模型,其余波段处两种模型均可;反演南疆冬枣Dolp时,在波长为1300 nm附近,需用Ross-Li模型,其余波段处两种模型均可。阴天反演南疆冬枣强度分量时,在1000~1350 nm范围,需用Roujean模型,在1600 nm附近,需用Ross-Li模型,其余波段处两种模型均可;反演南疆冬枣Dolp时,在1000~1350 nm范围,需用Roujean模型,在1600 nm附近,需用Ross-Li模型。晴天反演南疆冬枣强度分量时,波长为1000~1350和1600 nm附近,需用Ross-Li模型,其余波段处则无特殊要求;反演南疆冬枣Dolp时,在1000 nm附近,需用Roujean模型,在1600 nm附近,需用Ross-Li模型。探索出消除或减小反演数据差异的方法,为南疆冬枣户外检测提高精度奠定基础。

Comparative Study on Polarization Spectral Parameters of Fruit in Southern Xinjiang Based on Roujean Model

Based on the study of phase angle and wavelength in pBRDF (Polarized bidirectional reflectance distribution function), roujean model was proposed to describe Orient (Polarization phase angle) quantitatively. The Roujean model was used to quantitatively describe different fruits intensity components (F00) and polarization phase angle (Orient), and the simulation results were analyzed and compared using statistical analysis and comparison methods to realize the prediction from the regular model to the outdoor fruit tree canopy to the canopy of outdoor fruit tree canopy random distribution. The experimental results showed that: 1) when the phase angle of jujube was 52.19°, 66.51° and 88.26°, the R2 and average errors of F00 parameters described by Roujean model are 0.9982, 0.9963, 0.9912 and 3.80%, 4.17%, 6.40%, respectively; and the R2 and average error of Orient parameters described by Roujean model are 0.9056, 0.9223, 0.9260 and 6.23%, 3.32%, 8.05%, respectively;It can be seen that roujean model can quantitatively describe the Orient parameter of jujube;2) When the phase angle of apricot was 70.99°, 71.28° and 67.91°, the R2 and average errors of F00 parameters described by Roujean model is 0.9862, 0.9823, 0.9792 and 3.40%, 4.82%, 5.19%, respectively; And the R2 and average error of Orient parameters described by Roujean model are 0.9382, 0.8947, 0.8849 and 7.19%, 9.28%, 9.47%, respectively. Roujean model can also quantitatively describe the Orient parameter of white apricot. In summary, the Roujean model can provide a good quantitative description of f00 and a good quantitative description of Orient, which in turn can predict the pBRDF parameter for more fruits with different incidence and detection directions. It can correct the influence of angle factor in the nondestructive testing of outdoor fruits.

TM影像地面二向性的分析及校正方法研究

在影响辐射失真的各种因素中,二向反射特性越来越被人们所重视。本文针对地面二向反射特性造成的辐射失真问题进行深入的研究,从辐射失真的根本角度上补偿色调的差异,使得尽量恢复图像灰度与实际地表辐射的真实对应关系。并采用适合TM影像的Roujean模型进行实验处理,得到了较好的减弱辐射差异的效果。最后采用影像拼接等方法对实验结果进行了分析和评价。

方位和品质对南疆果品空间特性光谱影响及反演方法研究

高光谱无损检测技术在果品定量无损检测中应用广泛,以冬枣、红提、香梨三种果品空间特性光谱为研究目标,探索空间特性光谱的影响因素和反演方法,为提高户外果品无损检测精度提供了一种新思路。分别提取三种果品的光谱库并计算空间特性光谱,依次使用马氏距离、浓度残差等预处理方法以及竞争性自适应权重取样算法选取特征波长,将处理后的三种果品空间特性光谱分别与品质(糖度、水分)和方位(探测角、方位角、相位角)建模,建模结果如下:三种果品(按照冬枣、红提、香梨的顺序)与糖分模型的相关系数r分别为:0.8533,0.8227和0.9133;水分模型的相关系数r分别为:0.7413,0.7847和0.8913;探测角模型相关系数r分别为:0.9856,0.9927和0.9747;方位角模型相关系数r分别为:0.9418,0.9105和0.9369;相位角模型相关系数r分别为:0.9609,0.9570和0.9563。可以看出,不同果品方位模型相关性都明显高于品质模型相关性,因此方位因素是影响空间特性光谱的主要原因。使用Roujean模型和Walthall模型分别对不同方位的空间特性光谱进行反演,反演结果如下:使用Roujean模型反演三种果品(按照冬枣、红提、香梨的顺序)空间特性光谱时R^(2)分别为0.9344,0.9281和0.8306;r分别为0.9902,0.9839和0.9691;RMSEP分别为0.0309,0.0487和0.0627;平均模型误差分别为7.27%,11.02%和8.61%。使用Walthall模型描述不同果品空间特性光谱时R^(2)分别为0.9433,0.8597和0.8390;r分别为0.9918,0.9718和0.9702;RMSEP分别为0.0366,0.0661和0.0687;平均模型误差分别为6.19%,15.40%和7.84%。可以看出,Roujean模型可以很好的描述冬枣和红提的空间特性光谱,也可以较好的描述香梨空间特性光谱;Walthall模型可以很好的描述冬枣空间特性光谱,也可以较好的描述红提和香梨空间特性光谱。综上所述,在今后试验中可以使用Roujean模型反演红提和香梨的空间特性光谱,使用Walthall模型反演冬枣的空间特性光谱,进而提高户外果品户外果品无损检测精度。