Determination of dichlorvos contamination on navel orange surface based on least squares support vector machines

Spectral technique can provide a rapid, nondestructive means to assess quality and safety of agricultural commodities for human consumption. A procedure for determination of dichlorvos contamination has been developed with Vis-NIR spectroscopy. Four spectral preprocessing methods, including multiplicative scatter corrections (MSC), standard normal variate (SNV), first derivative (FD) and second derivative (SD) were used to reduce or eliminate scatter effects. Based on the spectral preprocessing methods, least squares support vector machines (LS-SVM) was performed. A total of 160 navel oranges were separated into two sets, one was calibration set (including 110 samples), and the other was prediction set (including 50 samples). It was found that LS-SVM model with the application of the preprocessing method of FD could predict dichlorvos residue. In the prediction set, the root mean squared error of prediction samples (RMSEP) was 6.2598 and the correlation coefficient Rpre was 0.8174.