Submitted to: Journal of Food Engineering
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: November 16, 2006
Publication Date: February 2, 2007
Repository URL:http://www,sciencedirect.com Citation: Liu, Y., Chen, Y.R., Kim, M.S., Chan, D.E., Lefcourt, A. M. 2007. Development of simple algorithms for the detection of fecal contaminants on apples from visible/near infrared hyperspectral reflectance imaging. Jrnl. of Food Eng. 81:412-418.
Interpretive Summary: To be assured of healthy and safe apple products to the consumers, the U.S. Food and Drug Administration has issued a policy to minimize the likelihood of bacterial pathogens in fruit juices, and also identified an urgent need to develop methods for the detection of fecal matters on apples. Preventing apples with visible fecal contamination from entering the washer tank is critical for preventing cross-contamination of other apples. Thus, removal of fecal contaminated apples, before entering the washing pool, has been adopted by apple processors as the Pathogen Reduction, Hazard Analysis, and Critical Control Points (HACCP) system. Currently, inspection of fecal contaminants is through visual observation, and such a process is labor intensive and prone to human error. Hence, researchers at the USDA Agricultural Research Services have been developing hyperspectral reflectance imaging system for the use in real-time on-line detection of fecal contaminated apples. To further improve the mathematical model and validate the selection of specific key wavelengths in spectral imaging processing, it is necessary to obtain the characteristic bands for fecal contaminated apples. Here, we first determined the unique bands in visible and NIR region for feces and uncontaminated apples from the spectral intensity variations, then we developed a number of algorithms to process hyperspectral images for the detection of fecal spots on apples. The results revealed that a dual-band ratio (Q725/811) algorithm could be used to identify fecal contaminated skins effectively, and also be a potential to detect the bruised apple skins. It was important as the two bands are away from the absorptions of natural pigments (such as chlorophylls and carotenoids), and hence can reduce the influence from color variations due to different apple cultivars. This result provides apple processors and researchers a new sight in applying both visible/NIR and imaging spectroscopy for safety/quality grading or classifying.
Hyperspectral reflectance images of two cultivars of apples were acquired after fecal treatments at three different concentrations to explore the potential for the detection of fecal contaminants on apple surfaces. Region of interest (ROI) spectral features of fecal contaminated areas showed a reduction in reflectance intensity compared to those of uncontaminated skins. Large spectral differences between uncontaminated and fecal contaminated skins of two types of apples occurred in the 675-950 nm visible/NIR region, which provided the basis for developing universal algorithms in the detection of fecal spots. Comparison of a number of processed images revealed that a dual-band ratio (Q725/811) algorithm could be used to identify fecal contaminated skins effectively. The result was most important as the two bands are away from the absorptions of natural pigments (such as chlorophylls and carotenoids), and hence can reduce the influence from color variations due to different apple cultivars.