|Ariana, Diwan - FAS|
|Guyer, Daniel - MICHIGAN ST UNIVERSITY|
Submitted to: Proceedings of SPIE
Publication Type: Proceedings
Publication Acceptance Date: October 19, 2005
Publication Date: December 15, 2005
Citation: Ariana, D., Lu, R., Guyer, D. 2005. Detection of mechanical injury on pickling cucumbers using near-infrared hyperspectral imaging. Proceedings of SPIE. 5996:59960P. Interpretive Summary: Mechanical injuries to cucumber fruit occur frequently during mechanical harvesting and postharvest transport and handling. These injuries often cause hidden internal physical damage to cucumber fruit that is invisible or difficult to detect by human inspectors. The injured cucumbers that passed inspection may lead to increased bloating during brining, causing economic losses to the pickle processor. This research was to develop a hyperspectral imaging system in the near-infrared region which is invisible to human eye for detecting mechanical injury on pickling cucumbers. Hyperspectral imaging system captures both spatial and spectral information of the reflected light from an object, which provides us with a much greater amount of information than does conventional imaging technique. Freshly harvested pickling cucumbers were mechanically injured by means of dropping and rolling which simulated typical loads encountered in mechanical harvesting. The injured cucumbers were then monitored by the hyperspectral spectral imaging system over a six-day time period. The near-infrared hyperspectral imaging system achieved 100% accuracy in detecting all mechanically injured cucumbers within two hours after bruising. The detection accuracy was lower when the cucumbers had been bruised for more than one day because of the self-healing of the bruised tissue. Hyperspectral imaging is effective for detecting mechanical injury and possibly other defects on pickling cucumbers and for identifying selected wavelengths that are useful for developing a practical imaging system for online inspection and sorting of defective cucumbers. Automated detection through hyperspectral or multispectral imaging of defective pickling cucumbers will help the industry provide high quality pickle products and reduce economic losses.
Technical Abstract: Automated detection of defects on freshly harvested pickling cucumbers will help the pickle industry provide high quality pickle products and reduce potential economic losses. Research was conducted on using a hyperspectral imaging system for detecting defects on pickling cucumbers caused by mechanical stress. A near-infrared hyperspectral imaging system was used to capture both spatial and spectral information from 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. Cucumber samples were subjected to two forms of mechanical loading, dropping and rolling, to simulate stress caused by mechanical harvesting. Hyperspectral images were acquired from the cucumbers over time periods of 0, 1, 2, 3, and 6 days after mechanical stress. Hyperspectral image processing methods, including principal component analysis and wavelength selection, were developed to separate normal and mechanically injured cucumbers. Results showed that reflectance from normal or non-bruised cucumbers was consistently higher than that from bruised cucumbers. The spectral region between 950 and 1350 nm was found to be most effective for bruise detection. The hyperspectral imaging system detected all mechanically injured cucumbers immediately after they were bruised. The overall detection accuracy was 97% within two hours of bruising and it was lower as time progressed. Lower detection accuracies for the prolonged times after bruising were attributed to the self-healing of the bruised tissue after mechanical injury. This research demonstrated that hyperspectral imaging is useful for detecting mechanical injury on pickling cucumbers.