Application of DebNet combined with fluorescence spectroscopy for rapid multi-pesticide residue classification

The illegal use of pesticides has led to severe residual pollution, posing serious threats to both human health and the environment. This situation underscores the urgent need for rapid and highly accurate classification methods for multi-pesticide residue detection. Although fluorescence spectroscopy remains a mainstream technique in this field, its classification performance is often limited by spectral overlap and background noise. To address these challenges, this study proposes DebNet, a deep learning model based on one-dimensional fluorescence spectral data. DebNet integrates one-dimensional convolutional neural networks (1D-CNN), long short-term memory (LSTM) networks, and self-attention mechanisms to collaboratively mitigate spectral interference. Experimental results demonstrate that DebNet achieves a classification accuracy of 99.83 % on preprocessed data, with a training time of approximately 5 min. It enables fast and accurate classification of four high-risk pesticides, including cyromazine, captan, metolachlor and thiamethoxam. Overall, the proposed method offers a lightweight and effective solution for real-time monitoring of pesticide residues in agricultural environments. Its robustness under spectral overlap conditions makes it particularly suitable for on-site applications requiring rapid and accurate classification.