基于近红外光谱的番茄农药残留无损检测方法研究  

Nondestructive Evaluation of Pesticide Residual in Tomatoes with Near Infrared Diffuse Reflection Spectrum

作　　者：吴泽鑫[1] 李小昱[1] 王为[1] 展慧[1] 周竹[1] 

机构地区：[1]华中农业大学工程技术学院,武汉430070

出　　处：《湖北农业科学》2010年第4期961-963,共3页Hubei Agricultural Sciences

基　　金：华中农业大学科研项目(52204-02008)

摘　　要：以湖北地区番茄样品为研究对象,对获取的光谱特征信息进行分析,确定了矢量归一化法为最优光谱预处理方法;对各个信息的主成分因子进行了优化,通过主成分分析提取主成分得分向量构成模式识别的输入,利用BP神经网络方法建立番茄有机磷农药残留的无损检测模型。结果表明,当光谱信息主成分因子数为3时,建立的模型最优,预测的识别率达到0.96,训练误差为0.015,相关系数达到0.971。The NIR spectra of tomato samples from Hubei province were studied in this paper. Vector Normalization （VN） as the optimal pre-processing method was used to optimize the principal component factors of all spectra. With the analysis of principal component, the score vectors were obtained and utilized as input of the BP artificial neural network for pattern recognition to establish the nondestructive evaluation model of organophosphorus pesticides in tomato. The results indicated that when the number of principal components was 3, the best model with its correct discrimination rate of 96% , training error of 0.015 and correlation coefficient of 0.971 was obtained.

关 键 词：番茄 近红外光谱 农药残留 神经网络 

分 类 号：O657.33[理学—分析化学] TS207.53[理学—化学]