|Ariana, Diwan - MICHIGAN ST UNIVERSITY|
Submitted to: Experiment Station Bulletins
Publication Type: Experiment Station
Publication Acceptance Date: January 6, 2006
Publication Date: January 20, 2006
Citation: Ariana, D., Lu, R. 2006. Detection of bruises on pickling cucumbers using near-infrared hyperspectral imaging. Experiment Station Bulletins. Technical Abstract: Near-infrared (NIR) hyperspectral imaging technique was investigated for the detection of bruises on pickling cucumbers caused by mechanical stress. An NIR hyperspectral imaging system was developed to acquire both spatial and spectral information from pickling cucumbers in the spectral region of 900 – 1700 nm. Experiments were conducted to acquire hyperspectral images from cucumber samples during time periods of six days after they were subjected to dropping or rolling under load which simulated damage caused by mechanical harvesting and handling systems. Results showed that reflectance spectra from bruised tissue were consistently lower than those of normal tissue. Bruise spectra changed over time from lower reflectance to higher reflectance. A large spectral difference between good and bruised tissue occurred in the 950 – 1350 nm region. The image processing algorithm based on principal component analysis correctly detected 95% of the bruised cucumbers within one day of bruising. Bruise detection accuracies were progressively lower as the time prolonged after bruising, which were attributed to the self-healing of the bruised tissue after mechanical injury. Band ratio and difference of two wavelengths achieved as high as 93% bruise detection accuracy. This research demonstrated that NIR hyperspectral imaging is useful for detecting bruises on pickling cucumbers. The key wavelengths found from the study can be used for further development of a machine vision system for rapid, real time inspection of bruises or other defects on pickling cucumbers.