摘要
目前,我国草莓种植的主要方式为设施栽培,由于土壤连作、温湿度调控不当等问题极易造成病虫害发生,种植户用药明显增多,草莓中农药残留问题较为突出。为此,设计了一种快速检测草莓上残留农药的电子鼻系统。针对残留农药挥发气味浓度低且易受到草莓本身气味的影响,电子鼻识别效果差的问题,对电子鼻系统进行改进。结构上,借鉴人体鼻腔结构特点,设计了一款仿生气室。采用计算流体力学(computational fluid dynamics,CFD)模拟对仿生气室的结构进行了优化,保证采集端信号的质量。算法上,建立了基于陷阱规避算子(trap-avoidance operator,TAO)改进的牛顿-拉弗森优化算法(Newton-Raphson based optimizer,NRSR)优化反向传播(BP)神经网络的分类模型(NRBO-BP),提高分类算法对低浓度信号识别效果。采用电子鼻对含有多菌灵和吡虫啉及其混合农药的草莓进行检测实验。结果表明,基于仿生气室电子鼻的NRBO-BP分类模型的准确率为93.44%,召回率为94.16%,准确度总体高于粒子群算法(PSO)优化BP模型的88.33%和BP神经网络的83.33%,能够准确检测草莓上残留的农药,可以作为草莓质量安全的快速评价方法。
At present,the main way of strawberry planting in our country is facility cultivation.Due to soil continuous cropping,improper temperature and humidity control and other problems,it is easy to cause diseases and pests,and the use of drugs by farmers has increased significantly,and the problem of pesticide residues in strawberries is more prominent.In this paper,an electronic nose system for detecting pesticide residues on strawberries was designed.The electronic nose system was improved to address the problem of poor recognition of the electronic nose due to the low concentration of the volatile odor of the residual pesticide and its susceptibility to the odor of the strawberry itself.Structurally,a bionic air chamber is designed by drawing on the structural characteristics of the human nasal cavity.The structure of the bionic gas chamber was optimized using computational fluid dynamics(CFD)simulations to ensure the quality of the signal at the acquisition end.Algorithmically,a classification model based on the trap-avoidance operator(TAO)improved Newton-Raphson based optimizer(NRBO)optimizing back-propagation(BP)neural network was established to improve the classification algorithm′s effect on the recognition of low-concentration signals.Strawberry containing carbendazim and imidacloprid and their mixed pesticides were detected by electronic nose.The results showed that the NRBO-BP classification model based on the bionic air chamber electronic nose had an accuracy of 93.44%and a recall of 94.16%.NRBO-BP classification model was generally higher than the 88.33%of the PSO-BP model and the 83.33%of the BP neural network,and was able to accurately detect pesticide residues on strawberries.It can be used as a rapid method for the evaluation of strawberry quality and safety.
作者
杜文斌
王伟
Du Wenbin;Wang Wei(School of Information and Communication Engineering,North University of China,Taiyuan 030051,China)
出处
《国外电子测量技术》
2024年第8期189-197,共9页
Foreign Electronic Measurement Technology
基金
山西省自然科学基金面上项目(202203021221117)
山西省高等学校教学改革创新项目(J20230779)
山西省研究生科研创新项目(2023KY607)资助。
关键词
农药残留
电子鼻
计算流体力学
仿生气室
pesticide residues
electronic nose
computational fluid dynamics
bionic air chamber
