基于小波变换与LS-SVR的柑橘叶片磷含量高光谱监测模型  

Hyperspectrum based models for monitoring phosphorus content of Luogang Orange leaf using Wavelet Denoising and Least Squares Support Vector Regression Analysis

作　　者：黄双萍[1,2,3] 岳学军[1,2,3,4] 洪添胜[1,2,3] 蔡坤[1,2,3] 林诗伦[1] 

机构地区：[1]华南农业大学工程学院,广东广州510642 [2]南方农业机械与装备关键技术省部共建教育部重点实验室,广东广州510642 [3]国家柑橘产业技术体系机械研究室,广东广州510642 [4]南昆士兰大学工程与测绘学院

出　　处：《广东农业科学》2013年第13期37-40,共4页Guangdong Agricultural Sciences

基　　金：国家自然科学基金(30871450);广东省自然科学基金(S2012010009856);省部产学研结合项目(2011B090400359)

摘　　要：快捷、准确、无损地监测柑橘磷(P)含量,对柑橘树磷肥的精准喷施及动态管理有重大意义。高光谱技术的快速发展使柑橘磷含量的快速无损监测成为可能。以117株园栽萝岗橙为试验对象,分别在壮果促梢期和采果期两个不同发育阶段采集234个样本数据,高光谱反射数据构成描述样本的多元矢量,硫酸-双氧水消煮-钼锑抗比色法测得的磷含量值作为样本标签值。在对高光谱反射数据小波去噪的基础上,用LS-SVR算法建立柑橘叶片磷含量监测模型。模型分别在验证集和校正集上进行评估,分别取得模型决定系数0.907和0.953,均方误差0.004和0.002,平均相对误差2.76%和1.77%。结果表明:用高光谱技术进行柑橘叶片磷含量监测是可行的。Citrus phosphorus content plays an important role in citrus physiological and ecological process. The quickness, accuracy and nondestructive determination of citrus phosphorus content provide references for a reasonable phosphate fertilizer, which can improve the yield and quality of citrus. The rapid development of hyperspectral technology makes it possible to measure citrus phosphorus nondestructively. Field experiments were conducted on 117 planted Luogang citrus trees in the Crab Village of Guangzhou and 234 pairs of data sample were collected in two different phenological periods, respectively, bloom period and picking period. Hyperspeetral reflection data was used as high- dimensional vector description, and phosphorus content measured by chemical method as true label to model and to predict the phosphorus content of citrus leaves. A regression analysis algorithm, least squares support vector regression were used to model and predict with the spectral reflectance data denoised by wavelet transform. Validation and calibration sets were used to evaluate the predictive performance of the determination model. The model achieved coefficient of determination of 0.907 and 0.953, the mean square error of 0.004 and 0.002 as well as the mean relative error of 2.76% and 1.77%. The experimental results showed that it is an effective way to predict phosphorus content based on hyperspectral reflection data which is denoised by wavelet transform.

关 键 词：柑橘叶片 磷含量 高光谱 小波去噪 最小二乘支持向量回归 

分 类 号：S24[农业科学—农业电气化与自动化]