|
|
|
| Detection of multiple pesticide residues on the surface of broccoli based on hyperspectral imaging |
| GUI Jiangsheng, GU Min, WU Zixian, BAO Xiao’an* |
(School of Information Science and Technology, Zhejiang Sci-Tech University, Hangzhou 310018, China)
|
|
|
|
Abstract A method for the detection of pesticide residues on broccoli was proposed based on hyperspectral image technology. Four groups of broccoli samples were used as experimental samples, which contained imidacloprid, abamectin and propineb as the first third groups respectively, and the last group was sprayed with water. A total of 180 broccoli samples were scanned by hyperspectral image system in the range of 383.70-1 032.70 nm. The average spectral information of region of interest (ROI) was extracted. Then, piecewise multiplicative scatter correction (PMSC) was adopted to eliminate light scattering of the average spectral information. To increase efficiency of the model and reduce the redundancy of the hyperspectral image, using the principal component analysis (PCA) algorithm and successive projection algorithm (SPA) for feature extraction. Mahalanobis distance (MD), least square support vector machine (LSSVM), artificial neural networks (ANN) and extreme learning machine (ELM) models were created to predict the pesticide residues from full spectra and characteristic wavelengths. The results showed that the optimal model is the SPA-ELM model, and the accuracy of training set is 98.33%, and the correct rate of test set is 96.67%, suggesting that it is feasible to use the principal component analysis algorithm and the artificial neural network algorithm to identify the pesticide residues on the surface of broccoli. In sum, this study develops a new method for rapid and nondestructive detection of pesticide residues on the surface of broccoli.
|
|
Published: 30 October 2018
|
|
|
基于高光谱图像的西兰花表面多种农药残留检测
针对西兰花农药残留问题，提出一种基于高光谱图像技术的西兰花农药残留定性检测新方法。首先，采集4组（共180 颗)分别喷洒了清水和吡虫啉、阿维菌素、丙森锌3 种农药的西兰花的高光谱(383.70～1 032.70 nm)图像，根据其图像信息提取感兴趣区域的平均反射光谱值，并采用分段多元散射校正对原始光谱数据进行预处理。为了提高模型效率，减少高光谱数据冗余，分别使用主成分分析和连续投影算法选择特征光谱。最后，使用马氏距离、最小二乘支持向量机、人工神经网络和极限学习机4 种分类算法建立基于全波段和特征波段信息的农药残留检测模型。结果表明：基于连续投影算法的极限学习机模型的识别效果最好，训练集和测试集的正确率分别为8.33%和96.67%。说明利用高光谱图像技术鉴别西兰花表面农药残留种类是可行的，为西兰花表面的农药残留无损检测提供了一种新的方法。
关键词：
高光谱图像,
西兰花,
农药残留,
人工神经网络,
极限学习机
|
Viewed |
|
|
|
Full text
|
|
|
|
|
Abstract
|
|
|
|
|
Cited |
|
|
|
|
| |
Shared |
|
|
|
|
| |
Discussed |
|
|
|
|
