Improving the accuracy of NIR detection of moldy core in apples using different diameter correction methods

Moldy core in apples is a common disease, with early symptoms not visible on the fruit surface. When affected apples are mixed with healthy ones, overall fruit quality declines, leading to the decay of healthy apples. Therefore, there is an urgent need for a rapid, non-destructive detection method. However, variations in apple diameter significantly affect the intensity of NIR transmission spectra, impacting the accuracy of detecting moldy core in apples. To address this issue, various diameter correction methods including a novel method we proposed were employed in this study to improve the accuracy of near-infrared detection of moldy core in apples, and these methods were also compared. The results indicate that the moldy core classification model is significantly influenced by apple diameter, with the uncorrected model achieving only 83.64 % accuracy in the prediction set. After adopting the diameter information fusion correction method, the performance of model has been slightly improved, with the accuracy of prediction set increasing by 0.91 %. Further improvement is achieved when using spectral normalization based on correlation and spectral correction based on diameter transformation, which has raised the accuracy of prediction set to 86.36 %. And the spectral correction based on polynomial transformation method proposed in this study has significantly improved the model performance, with the calibration and prediction sets achieving sensitivity, specificity, and accuracy of 85.22 %, 95.24 %, 90.00 %, and 85.45 %, 92.7 %, 89.09 %, respectively. Compared to the uncorrected model, the accuracy of the model in prediction set has been improved by 5.45 %. The model also demonstrates a 4.54 % enhancement over the one corrected using the diameter information fusion method. Additionally, when evaluated against the models using spectral normalization based on correlation and spectral correction based on diameter transformation, the accuracy has increased by 2.73 %. Therefore, the method of spectral correction based on polynomial transformation that proposed in this study effectively reduces the impact of apple diameter on transmission spectra, improving the detection accuracy of moldy core in apples and supporting rapid, non-destructive, high-precision detection.