Submitted to: Meeting Proceedings
Publication Type: Proceedings
Publication Acceptance Date: April 6, 2010
Publication Date: April 26, 2010
Citation: Yang, C., Jun, W., Kim, M.S., Chao, K., Kang, S., Chan, D.E., Lefcourt, A.M. Classification of Fecal Contamination on Leafy Greens by Hyperspectral Imaging. Meeting Proceedings. http://dx.doi.org/10/1117/12.851069.
Interpretive Summary: This study was to search the optimal algorithm in the hyperspectral imaging system for detection of fecal contamination spots on the leaf surface of romaine lettuce and baby spinach. In the algorithm, the background was firstly removed by a mask; the object image at the 666-nm waveband was then divided by the same image at the 680-nm waveband to obtain the ratio image. The threshold of 0.45 was applied to the image to extract the contaminated spots as a binary image. The 3X3 filter was applied to the binary image to remove the false positive pixels. The final images showed that all of the contaminated spots were successfully identified while none of uncontaminated pixels were incorrectly marked as contaminated ones. Thus, the algorithm with the hyperspectral imaging system could be used for detection of fecal contamination on leafy green vegetables
A hyperspectral fluorescence imaging system was developed and used to obtain several two-waveband spectral ratios on leafy green vegetables, represented by romaine lettuce and baby spinach in this study. The ratios were analyzed to determine the proper one for detecting bovine fecal contamination on romaine lettuce and baby spinach. Two wavebands corresponding to fluorescence emission peaks for fecal matter and chlorophyll a were considered useful to detect fecal contamination on samples of relatively high chlorophyll content such as leafy green vegetables. The emission peak from bovine fecal matter was in 666 nm and the emission peak from green plant leaves was found in 680 nm. The results indicated that a two-waveband spectral ratio, the spectral reflectance in 666nm over one in 680nm, could be applied to the multispectral algorithm for contamination detection. The threshold and erosion analyses were performed to reduce false positive response from leaf vein and inter-veinal regions of leafy green vegetables. The results showed that the fluorescence-based hyperspectral/multispectral imaging system can successfully detect fecal contamination on leafy green vegetables.