基于光谱技术识别不同农药污染脐橙的研究  

Recognition of Different Pesticide Contamination in Navel Oranges Based on Spectra Technology

作　　者：黎静[1] 薛龙[2] 刘木华[1] 王晓[1] 罗春生[1] 

机构地区：[1]江西农业大学工学院,江西南昌330045 [2]华东交通大学机电学院,江西南昌330013

出　　处：《江西农业大学学报》2010年第4期723-728,共6页Acta Agriculturae Universitatis Jiangxiensis

基　　金：国家自然科学基金资助项目(30760101);江西省教育厅科学技术研究项目(GJJ08513)

摘　　要：用遗传算法(Genetic Algorithm,GA)搜寻可识别被不同农药污染脐橙的可见/近红外光谱的最佳特征光谱区间及波长,并建立了支持向量机(Support Vector Machines,SVM)定性分析模型。实验供试农药为灭多威、氰戊菊酯和氧乐果3种。通过GA来搜寻整个波段范围(460～1800nm),将得到的9个最佳特征光谱区间所包含的波长(共318个)作为SVM建模的输入变量,对识别被3种农药污染脐橙的准确率为100%。并继续应用GA优化,得到71个特征波长,此时建立的SVM模型的识别准确率为99.57%。虽然识别的准确率有所下降,但是模型的复杂程度得到了很大的优化,其输入变量减少到71个。实验结果表明利用可见/近红外光谱技术结合SVM方法可以有效识别被不同农药污染的脐橙。Genetic algorithm （GA） was used to search for the best characteristic spectral ranges and wavelengths of visible/near-infrared spectra （Vis/NIRs） ,a qualitative analysis model of support vector machine （SVM） was set up to recognize navel oranges contaminated with different pesticides.The pesticides in the experiment were Methomyl,fenvalerate and omethoate.Using GA to search the entire band range （460 ～ 1 800 nm） ,the 9 best characteristic spectral ranges （318 wavelengths） were used as the input variables of SVM model and the accuracy of the prediction set classification was 100% .Then GA method was used continually and 71 wavelengths were extracted,the corresponding SVM model was built with 99.57% accuracy.Although the classification accuracy rate declined,the complexity of the model was greatly optimized by reducing the input variables to 71.The experiment results showed that the application of Vis/NIRs combined with SVM can effectively detect the navel oranges contaminated with different pesticides.

关 键 词：农药污染 脐橙 可见/近红外光谱 无损检测 支持向量机 遗传算法 

分 类 号：O436[机械工程—光学工程] S123[理学—光学]