Smartphone Based Portable Spectrometer System for Pesticide Residue Detection

Abstract

: To address the global challenge of pesticide residue contamination in agricultural products, this study developed an integrated smartphone-based portable spectrometer system that combines a multi-model compatible optical structure with a Segmented Modeling Regression (SMR) algorithm. The spectrometer achieves high stability (MSE < 0.01) across diverse mobile devices, while the SMR algorithm leverages XGBoost for concentration-tier classification (98.1% accuracy) followed by tier-specific regression modeling, enabling precise detection across a broad concentration range (0.1 –100 mg/kg). The system’s efficacy was validated through multi-scenario applications, including dimethoate residue detection in water (relative error < 20%), soil nutrient analysis (relative error < 30%), and surface pesticide screening on fruits and vegetables. This work provides a cost-effective, field-deployable solution for rapid on-site detection, demonstrating significant potential to advance intelligent food safety monitoring and sustainable agricultural practices. Future work will focus on expanding the training dataset, advancing embedded hardware for real-time analysis, and ultimately evolving the technology into an integrated decision-support platform for agricultural intelligence.