Near-infrared hyperspectral reflectance imaging for detection of bruises on pickling cucumbers

Mechanical injury often causes hidden internal damage to pickling cucumbers, which lowers the quality of pickled products and can incur economic losses to the processor. A near-infrared hyperspectral imaging system was developed to capture hyperspectral images from pickling cucumbers in the spectral region of 900–1700 nm. The system consisted of an imaging spectrograph attached to an InGaAs camera with line-light fiber bundles as an illumination source. Hyperspectral images were taken from the pickling cucumbers at 0–3, and 6 days after they were subjected to dropping or rolling under load which simulated damage caused by mechanical harvesting and handling systems. Principal component analysis (PCA), band ratio, and band difference were applied in the image processing to segregate bruised cucumbers from normal cucumbers. Bruised tissue had consistently lower reflectance than normal tissue and the former increased over time. Best detection accuracies from the PCA were achieved when a bandwidth of 8.8 nm and the spectral region of 950–1350 nm were selected. The detection accuracies from the PCA decreased from 95 to 75% over the period of 6 days after bruising, which was attributed to the self-healing of the bruised tissue after mechanical injury. The best band ratio of 988 and 1085 nm had detection accuracies between 93 and 82%, whereas the best band difference of 1346 and 1425 nm had accuracies between 89 and 84%. The general classification performance analysis suggested that the band ratio and difference methods had similar performance, but they were better than the PCA.