This study introduces a novel method for swift detection of AFB1 in peanuts using millimetre wave technology. The research team devised a portable millimetre-wave detection device employing a double-external-difference mixing structure. The device measured millimetre-wave transmission coefficients in the 20 GHz - 40 GHz frequency range for peanut samples. Results showed that the PCA-KNN model excelled in qualitative AFB1 detection, achieving 100 % accuracy in the prediction set. In quantitative analysis, by condensing the feature variables into a 16-dimensional space, the BOSS-PSO-SVR model enhanced performance. Compared to the full transmission coefficient SVR model, the BOSS-PSO-SVR model exhibited improved coefficients of determination (RP2), reducing root mean square error of prediction (RMSEP) from 36.49 μg∙kg−1 to 19.08 μg∙kg−1, and enhancing relative prediction deviation (RPD) from 3.17 to 6.06. This study concludes that the integration of a custom miniaturized millimetre-wave device with appropriate chemometric methods facilitates rapid and accurate detection of peanut AFB1 levels.
Read full abstract