Classification and quantification of volatile compounds in cherry tomatoes under different pesticide treatments using an electronic nose

With the growing severity of pesticide residue issues, it is crucial to develop a rapid and reliable detection method for assessing pesticide residues. This study proposed an approach for both qualitative and quantitative detection of pesticide residues using an electronic nose. A self-developed electronic nose system equipped with 32 metal oxide semiconductor (MOS) sensors was employed to detect various types and concentrations of pesticide residues in tomatoes. Support Vector Machine (SVM), Artificial Neural Network (ANN) and Random Forest (RF) algorithms were used to establish the qualitative recognition model, and SVM, Partial Least Squares Regression (PLSR) and RF algorithms were used to perform the quantitative prediction. The results demonstrated that the detection accuracy of the qualitative RF model was 98.3%, coefficient of detemination(R2) of the quantitative RF model of four pesticides: chlorothalonil, procymidone, iprodione and Beta-cypermethrin were 0.94, 0.99, 0.97, 0.99, and the root-mean-square-error (RMSE) was reduced to 3.76, 1.21, 1.92 and 0.54 mg/L respectively. The RF model showed advantage in both qualitative and quantitative detection. These results revealed great potential of the electronic nose technology in detection of pesticide residues on farm plants.