In-plant Test of In-line Multispectral Imaging System For Fecal Detection during Poultry Processing

Abstract

Food Safety Inspection Service established new requirements for the zero tolerance for fecal contaminated broiler carcasses entering the chiller. The hyperspectral imaging research accomplished at ARS has demonstrated the real-time multispectral imaging system was able to effectively detect fecal contaminants during poultry processing. The imaging system with two-band (517 and 565 nm) was able to inspect every bird in real-time, in-line mode at commercial processing speed. With several real-time image processing methods including binning, cuticle removal filter, median filter, and morphological analysis, the system performance was improved by maximizing detection accuracy and minimizing false positive errors. Most false positives were found at wing bone, thigh, vent, and boundary. Since the number of true positives confirmed organoleptically by human inspector was only 14 out of 29,821 carcasses (0.05%), more samples need to be tested to confirm the performance of detection accuracy. Overall, the ARS multispectral imaging system was able to detect contaminants with high detection accuracy (approximately 91%) and 3.3% false positive errors at the processing speed of 150 birds/min from the two-day in-plant trials. However, the performance of ARS real-time in-line imaging system for fecal contaminant detection on broiler carcasses can be improved by adopting a better camera which has higher signal-to-noise ratio such as line-scan real-time hyperspectral imaging system currently being investigated for its applicability for real-time processing.