摘要
为了快速检测赣南脐橙果树叶片含水率,提出近红外光谱结合最小二乘支持向量机的快速检测方法。采用积分球漫反射方式采集叶片的近红外光谱,通过间隔偏最小二乘法从2 074个光谱变量中优选出345个变量作为建模的输入向量,分别建立最小二乘支持向量机和偏最小二乘校正模型。经比较,以径向基函数为核函数的最小二乘支持向量机模型预测结果最优,预测相关系数为0.942,预测均方根误差为2.7%,模型建立及预测时间为0.176s。实验结果表明近红外光谱结合最小二乘支持向量机的脐橙叶片含水率无损检测方法是可行的。
In order to quickly detect of water content for Gannan naval oranges＂ leaves, a rapid detection method of near infrared （NIR） spectroscopy combined with least squares support vector machine （LSSVM） was proposed. The NIR spectra of leaves were recorded by integrating sphere diffuse reflectance mode. 345 variables were chosen from 2074 spectral variables by interval partial least squares （iPLS） method as the input vector of models. The calibration and prediction models were developed by LSSVM and PLS methods. By comparison, the performance of LSSVM models with kernel function of radial basis function （RBF） was superior, correlation coefficient of prediction was 0.942, root mean square error of prediction was 2.7%, and forecast time of calibration and prediction was 0.176 seconds. The experimental results show that it is feasible to detect water content of navel oranges＂ leaves nondestructively for NIR spectroscopy combined with LSSVM method.
出处
《中国农机化学报》
2015年第2期150-153,168,共5页
Journal of Chinese Agricultural Mechanization
基金
江西省科技支撑计划(20121BBF60054)
江西省教育厅青年基金项目(GJJ12317)
关键词
光谱学
近红外
含水率
最小二乘支持向量机
叶片
spectroscopy
near infrared
water content
least squares support vector machine
leaves
作者简介
孙旭东．男，1978年生，吉林辽源人，讲师；研究方向为农业信息技术。E—mail：sunxudong__18@163．com
