基于高光谱成像技术的脐橙叶片的叶绿素含量及其分布测量

为实现脐橙叶片叶绿素含量无损检测及其分布可视化表征,采用高光谱成像技术,结合自适应重加权算法(CARS)和连续投影算法(SPA),筛选特征光谱变量,进行脐橙叶片叶绿素含量及可视化分布研究。选取叶绿素测量位置的7×7矩形感兴趣区域,提取并计算脐橙叶片平均光谱。基于Kennard-ston方法,将148个脐橙叶片样品划分成建模集和预测集(111∶37)。采用CARS和SPA算法分别筛选出了32个和6个叶绿素特征光谱变量,用于建立偏最小二乘(PLS)回归模型。采用37个未参与建模的脐橙叶片样品评价模型的预测能力,经比较,CARS-PLS和SPA-PLS模型均优于变量筛选前的PLS模型,且CARS-PLS和SPA-PLS模型的预测能力几乎相同,其预测集相关系数分别为0.90和0.91,均方根误差分别为1.53和1.60。SPA-PLS模型计算脐橙叶片每个像素点的叶绿素含量,经伪彩色变换,绘制了脐橙叶片叶绿素含量可视化分布图。实验结果表明:变量筛选方法结合高光谱成像技术,能够实现脐橙叶片叶绿素含量无损检测及叶绿素分布可视化表达,并简化了数学模型。

Measurement of Chlorophyll Distribution in Navel Orange Leaves Based on Hyper-spectral Imaging Technique

The chlorophyll content and distribution in the Gannan navel orange leaves were non-destructively measured by competitive adaptive reweighted algorithm (CARS) and successive projections algorithm (SPA) combined with hyperspectral imaging technology. 32 and 6 characteristic wavelengths were extracted by CARS and SPA, and then partial least squares (PLS) was used for modeling quantitative analysis. The results show that SPA-PLS and CARS-PLS model can obtain better results than PLS model through the analysis of prediction of 37 samples. The prediction set correlation coefficients were 0.90 and 0.91, the root mean square error is 1.53 and 1.60 respectively. The chlorophyll content of each pixel was calculated with SPA-PLS model, then the chlorophyll distribution map of navel orange leaves was visualized using imaging processing technology. Overall results sufficiently demonstrate that the variable selection method combined with hyperspectral imaging technology can be used to measure the chlorophyll content and distribution in navel orange leaves.