基于光谱角算法的鲜茶叶表面农药残留荧光高光谱图像无损检测研究

Nondestructive Detection of Pesticide Residues on Fresh Tea Leave using Fluoresce Hyperspectral Imaging Combined with Spectral Angle Algorithm

本文尝试利用荧光高光谱图像技术和光谱角算法快速无损检测鲜茶叶表面的农药残留状况。分别将多菌灵商用农药稀释成1∶200,1∶500和1∶1000的梯度农药溶液,滴在鲜茶叶表面,自然凉干后采集其高光谱图像。采用主成分分析方法优选得到768.74 nm波长下的特征图像,用光谱角算法识别叶面上的农药点与非农药点。结果表明,喷有稀释浓度为1∶200的鲜茶叶,其农药点与叶面信息能清楚地区分开,喷有稀释浓度为1∶500的样本,叶面上的农药点和非农药点能很好的区分开,但叶柄的部分像素误判为农药点,质量浓度比为1∶1000的样本,叶面上的叶茎部分也误判为是农药点。研究表明,利用荧光高光谱图像技术能较好地快速无损检测叶面上的高浓度农药残留。

Fluoresce hyperspectral image technology combined with spectral angle algorithm was used to fast and detect the situation of pesticide residues on fresh tea in this study. Distilling water was used to dilute the commercial carbendazim pesticide and the gradient pesticide dilution of 1 ∶ 200,1 ∶ 500 and 1 ∶ 1000 were gained. Then the dilutions were respectively dropped on the surface of fresh tea leaves. Hyperspectral images were acquired after dried naturally. Principal component analysis method was used to optimize the feature wavelength and the characteristic images under 768. 74 nm were extracted. Spectral angle mappe（ SAM） method was used to identify the pesticide information on the leaves. The results showed that the pixels of pesticides and the pixels of leaves（ 1 ∶ 200） could be clearly distinguished. But for the pixels of pesticides and the the pixels of leaves（ 1 ∶ 500 and 1 ∶ 1000）,part of the pixels on the petiole and leaf stems were mistaken for the pixels of pesticides. The study indicated that fluoresce hyperspectral image technology could be used to nondestructively and effectively detect high concentration of pesticide residues on fresh leaves.