Abstract:  It is of great significance to detect the moisture content of tobacco leaves during the picking period, which plays a critical role in the tobacco planting industry. In order to rapidly and nondestructively detect the moisture content of tobacco leaves, a method of principal component analysis (PCA) combined with Mahalanobis distance (MD) was proposed to eliminate the abnormal samples. Then partial least squares (PLS) was used to estimate the moisture content of tobacco leaves. Firstly, the hyperspectral data of 141 mature tobacco leaves were obtained by Gaiaskymini2 airborne hyperspectral imager. The original spectra were preprocessed by multiple scattering correction (MSC), standard normal variable exchange (SNV), and SavitzkyGolay convolution smoothing. In addition, the combination of PCA and MD was used to eliminate the abnormal samples in the calibration sample set. Finally, PLS was used to establish the moisture content analysis model of tobacco leaves based on the data set after removing samples. The results showed thatthe PCAMDPLS model preprocesseffectt. The established PLS model has the best predictive ability for tobacco moisture content. The correlation coefficient of the prediction modelis 0.852 7, and the mean square error is 1.376 6.

Key words: water content, tobacco leaf, hyperspectral, principal component analysis, Mahalanobis distance, partial least squares method

摘要： 烟叶含水量的快速检测在烟草种植业中起着关键的作用,检测采摘期烟叶水分含量,对烟草工艺具有重要意义。为了快速、无损地检测采摘期烟叶水分含量,提出一种主成分分析(PCA)结合马氏距离算法(MD)的方法来剔除异常样本,再使用偏最小二乘法(PLS)估测采摘期烟叶水分含量。首先,利用GaiaSky-mini2机载高光谱成像仪获取到141个采摘期烟叶的高光谱数据,采用多元散射校正(MSC)、标准正态变量交换(SNV)和Savitzky-Golay卷积平滑法等对原始光谱进行预处理。然后,应用主成分分析结合马氏距离法对校正集中的异常样品进行剔除。最后,使用偏最小二乘法(PLS)建立采摘期烟叶水分含量分析模型。结果表明:利用SG卷积平滑法预处理的PCA-MD-PLS模型效果最佳,对烟叶含水量预测能力最好,预测模型相关系数为0.852 7,均方差为1.376 6。

关键词: 含水量, 烟叶, 高光谱, 主成分分析, 马氏距离, 偏最小二乘法