柑橘植株营养状况的遥感监测是实现果树轻简高效管理和优质丰产的重要手段,但迄今有关基于低空遥感信息的果树营养诊断研究鲜见报道。本文采用具有490 nm、550 nm、570 nm、671 nm、680 nm、700 nm、720 nm、800 nm、840 nm、900 nm、950 nm等11个波段光谱的八旋翼飞行器(UAV)载多光谱遥感系统,获取距地面100 m高度的哈姆林甜橙植株春季冠层近地遥感信息,对比分析基于多元散射校正(MSC)和标准正态变量(SNV)两种预处理光谱和原始光谱(OS)的偏最小二乘(PLS)、多元线性回归(MLR)、主成分回归(PCR)及最小二乘支持向量机(LS-SVM)等4种模型对冠层叶片氮素、叶绿素a、叶绿素b和类胡萝卜素含量预测精度的影响。结果显示,距地面100 m高度的多光谱信息,通过SNV光谱预处理和MLR建模对冠层叶片氮素、叶绿素a和叶绿素b含量的预测效果均较好,预测集相关系数(R_p)值分别达0.8036、0.8065和0.8107,预测均方根误差(RMSEP)值分别为0.1363、0.0427和0.0243;而在SNV光谱预处理基础上的LS-SVM建模对冠层类胡萝卜素含量预测效果更优,R_p值达到了0.8535,RMSEP值为0.0117。表明利用机载多光谱图像信息可实现对柑橘植株冠层全氮及叶绿素a、叶绿素b和类胡萝卜素含量的较好估算,为大规模柑橘园植株冠层营养状况的精准和高效监测提供了一条新途径。

Remote measurement and diagnosis of the plants nutritional status is an important means for efficient easily and simple management system, and high-yield and high quality cultivation. So far, there is not yet much research on the nutrition diagnostic of fruit trees through low-altitude remote sensing data. We carried out the following experiments in order to provide a theoretical basis and technical support for the research and development of nutritional diagnosis technology of fruit trees based on low-altitude remote sensing data.In this work, the multi-spectral image information of 'Hamlin' orange plant canopies were obtained by a multi-spectral camera array mounted on the eight rotor Unmanned Aerial Vehicle(UAV) at an altitude of 100 m above the canopy at 11:00-13:00 on a sunny day in spring. Then, the multi-spectral images were pre-processed by PixelWrench 2 of tetracam, average spectral reflectance of the whole canopy were individually extracted based on ENVI 4.7. Twenty leaves from the mature spring shoots were collected from around crown of every tree. Total nitrogen, chlorophyll a, chlorophyll b and carotenoids contents of each plant were measured in the laboratory. The characteristic wavelengths were extracted by means of the correlation analysis of the average spectra of the plants with the nutrition content. A total amount of 88 citrus trees were collected and randomly grouped into two sets of samples:66 plants for the calibration set and 22 plants for the prediction set. The two kinds of spectral pre-processing methods(Multiplicative Scatter Correction(MSC) and Standard Normal Variable(SNV)) were adopted and four kinds of modeling methods(Partial Least Squares(PLS), Multiple Linear Regression(MLR), Principal Component Regression(PCR) and Least Squares Support Vector Machine(LS-SVM)) were employed to estimate total nitrogen, chlorophyll a, chlorophyll b, and carotenoids content in canopy leaves.The results showed that the prediction accuracy of the MLR model based on SNV spectral pre-processing methods was the best for the prediction of total nitrogen, chlorophyll a and chlorophyll b content, correlation coefficients of prediction(R_p) were 0.8036, 0.8065, 0.8107, and Root Mean Square Error of Prediction(RMSEP) were 0.1363, 0.0427 and 0.0243, respectively. The LS-SVM model based on SNV spectral pre-processing methods for the carotenoids content of crown was the best, which is with R_p=0.8535, RMSEP=0.0117.The results demonstrated that the airborne multi-spectral image information of citrus plants canopy could be used to estimate total nitrogen, chlorophyll a, chlorophyll b and carotenoids content in canopy leaves. This research results would provide a new way for accurate, efficient prediction of plants nutrition status of large-scale citrus orchards.