青菜叶面药液浓度光谱相关特性研究.docVIP

 青菜叶面药液浓度光谱相关特性研究# 殷磊，邱白晶** （江苏大学农业工程研究院，江苏 镇江 212013） 5 10 15 20 25 摘要：本文依据青菜叶面药液浓度的光谱相关特性曲线，寻找到叶面药液光谱的最佳建模波 段，最佳波段范围为 350～1900nm。在该波段下进行偏最小二乘法回归，建立了青菜叶面药 液浓度大小与样本光谱反射率之间的定量数学模型。PLS 主成分数为 4，校正集相关系数为 0.984，校正均方根误差为 0.009；预测集相关系数为 0.929，预测均方根误差为 0.075，预 测平均相对误差为 11.469%。试验结果表明，所建模型的预测精度较高，运用相关谱图优选 建模波段，以该波段建模预测青菜叶面药液浓度是可行的。 关键词：近红外光谱；青菜叶面；偏最小二乘；敏感波段 中图分类号：S127 Research of related character spectra of pesticide concentration on the green vegetables leaf YIN Lei, QIU Baijing (Agricultural Engineering Research School,Jiangsu University, JiangSu ZhenJiang 212013) Abstract: This paper based on spectra related characteristic curve of pesticide concentrtion on the green vegetables leaf. finding the best spectra band of establishing model, the best band interval is 350～1900nm. making partial least squares regression in this band,and developing the quantitative analysis model for pesticide concentration with diffuse reflectance spectroscopy. Principal component count of Partial least squares is four,The correlation coefficient between the prediction values and the truth values in the calibration set (Rc) was 0. 984, the root mean square error of cross-validation of the calibration set(RMSEC) was0.009.The correlation coefficient between the prediction values and the truth values in the prediction set (Rp) was 0. 929, the root mean square error of cross-validation of the prediction set(RMSEP) was 0.075. Predict average relative error is 11.469%. Experiment results show that model prediction accuracy is higher, it is feasible that near infrared spectroscopy test pesticide concentration on the green vegetables leaf. Keywords: Near infrared spectroscopy; Green vegetables leaf; Partial least squares; Sensitive band 30 0 引言 我国是农业大国，施药技术落后于发达国家[1]，大量农药的使用，给周围生存环境带来 压力，也引起了人们的高度重视[2]。提高农药的利用率，科学合理的使用农药，最大限度的 减少农药对周围环境的污染，降低防治成本是我们迫切解决的农业工程问题。目前，国内外 35 40 研究者正在进行一种新的研究工作，不是调整药液喷量，而是调整药液浓度[3] [4]，以减少在 施药过程中农药的大量浪费，实现田间病虫害的控制的同时减轻对环境的污染。在这种喷施 过程中，需要测量农药喷施后沉降在叶子表面上药液浓度值，可以实现叶面药液浓