基于高光谱成像技术的不同时期番茄植株的快速无损判别研究  

Rapid and Nondestructive Identification of Tomato Plants in Different Periods Based on Hyperspectral Imaging Technique

作　　者：马玲[1] 马倩[1] 李亚娇 张祎洋 王静[1] 马思艳 马燕 杜明华 吴龙国[1,2] MA Ling;MA Qian;LI Ya-jiao;ZHANG Yi-yang;WANG Jing;MA Si-yan;MA Yan;DU Ming-hua;WU Long-guo(School of Agriculture,Ningxia University,Yinchuan 750001,China;Ningxia Modern Facility Horticulture Engineering Technology Research Center,Yinchuan 750001,China)

机构地区：[1]宁夏大学农学院,宁夏银川750001 [2]宁夏现代设施园艺工程技术研究中心,宁夏银川750001

出　　处：《分析测试学报》2023年第2期210-215,共6页Journal of Instrumental Analysis

基　　金：宁夏重点研发计划项目(2021BBF02024,2021BBF02019,2021YCZX0016,2021BEB04077,2022BBF02024-2,2022BBF03010,2022WZYQ0001);国家重点研发计划子课题专项(2021YFD1600302-3);第四批“宁夏青年科技人才托举工程”(TJGC2019065)。

摘　　要：利用近红外高光谱成像技术对不同浓度盐胁迫下的番茄叶片进行了定性判别。采集192个叶片样本的平均光谱反射率数据,并对原始光谱数据分别进行多元散射校正(MSC)、标准正态化(SNV)、正交信号校正(OSC)、相关优化偏移(COW)4种预处理,建立了偏最小二乘回归(PLSR)模型。建模结果显示:OSC预处理光谱的建模效果最优。分别采用间隔变量迭代空间收缩法(iVISSA)、间隔随机蛙跳法(IRF)、遗传偏最小二乘算法(GAPLS)、竞争性自适应加权算法(CARS)、变量组合集群分析(VCPA)等方法提取特征波长,建立PLSR模型。结果表明:VCPA提取特征波长所建立的模型最优。将VCPA法提取的11个特征波长(945、975、990、1 002、1 005、1 067、1 204、1 326、1 595、1 642、1 660 nm)用于建立番茄叶片定性判别预测模型,最优预测模型的决定系数(r^(2)P)与预测均方根误差(RMSEP)分别为0.917、0.456。该研究为在线监测植物长势提供了技术支撑。A near infrared hyperspectral imaging(NIR) technique was adopted for qualitative identification on the tomato leaves under different salt stress in order to quickly monitor the growth of tomato plants.The average spectral reflectance data of 192 leaf samples were collected.The original spectral data were preprocessed by multiple scattering correction(MSC), standardized normal variate(SNV),orthogonal signal correction(OSC) and correlation optimized warping(COW) to establish a partial least squares regression(PLSR) model.The modeling results showed that the modeling effect of OSC preprocessed spectrum was the best.Interval variable iterative space shrinking analysis(iVISSA),interval random frog(IRF),genetic algorithm and partial least squares(GAPLS),competitive adaptive weighted sampling(CARS) and variable combination population analysis(VCPA) were used to extract the feature wavelengths,thus the PLSR model was established.The results showed that the model established by VCPA extracting characteristic wavelengths was optimal.The VCPA method was used to extract 11 characteristic wavelengths(945, 975, 990, 1 002, 1 005, 1 067,1 204,1 326,1 595,1 642,1 660 nm),which were used to establish the qualitative discriminant prediction model for tomato leaves,while the determination coefficient(r^(2)P) of the optimal prediction model and the root mean square error(RMSEP) were 0. 917 and 0. 456,respectively.This study provided a technical support for online monitoring of plant growth in the future.

关 键 词：高光谱成像技术 番茄叶片 盐胁迫 无损检测 

分 类 号：O657.3[理学—分析化学] S641.2[理学—化学]