Detection of early collision and compression bruises for pears based on hyperspectral imaging technology

Abstract Early detection of bruising is one of the major challenges in postharvest quality sorting processes for pears. In this study, visible/near infrared (VIS/NIR) hyperspectral imaging technology (400–1000 nm) was used to rapidly detect the type of damage and the time period (1, 12, and 24 h) for damage to pears. Spectral images of nonbruised pears and pears subject to mechanical collision and compression bruises were acquired for three different time periods (1, 12, and 24 h), and the average spectrum was calculated for modeling. After analyzing and processing the spectral data obtained for the samples, principal component analysis (PCA) and uninformative variable elimination (UVE) were used to select optimum wavelengths, and an extreme learning machine (ELM) and support vector machine (SVM) were used to build the classification model. Then, the classification results were compared with the genetic algorithm-sooty tern optimization algorithm-support vector machine (STOA-GA-SVM). The accuracy of the PCA-ELM, UVE-ELM, PCA-SVM and UVE-SVM calibration and validation sets is determined to be 98.99%, 89.29%, 98.98%, 87.97%, 96.94%, and 88.78% and 99.23% and 88.78%, respectively, with varying degrees of overfitting. The STOA-GA-SVM model shows the best performance, and the accuracy of the calibration set and validation set is determined to be 95.92% and 91.84%, respectively. This study shows that the use of the VIS/NIR hyperspectral imaging technique combined with the STOA-GA-SVM algorithm is feasible for the rapid and nondestructive identification of the damage type and time for pears.