Rapid detection of residual chlorpyrifos and pyrimethanil on fruit surface by surface- enhanced Raman spectroscopy integrated with deep learning approach

Abstract

Chlorpyrifos and pyrimethanil are widely used insecticides/fungicides in agriculture. The residual pesticides/fungicides remaining in fruits and vegetables may do harm to human health if they are taken without notice by the customers. Therefore, it is important to develop methods and tools for the rapid detection of pesticides/fungicides in fruits and vegetables, which are highly demanded in the current markets. Surface-enhanced Raman spectroscopy (SERS) can achieve trace chemical detection, while it is still a challenge to apply SERS for the detection and identification of mixed pesticides/fungicides. In this work, we tried to combine SERS technique and deep learning spectral analysis for the determination of mixed chlorpyrifos and pyrimethanil on the surface of fruits including apples and strawberries. Especially, the multi-channel convolutional neural networks-gate recurrent unit (MC-CNN-GRU) classification model was used to extract sequence and spatial information in the spectra, so that the accuracy of the optimized classification model could reach 99% even when the mixture ratio of pesticide/fungicide varied considerably. This work therefore demonstrates an effective application of using SERS combined deep learning approach to rapid detection and identification of different mixed pesticides in agricultural products.