ASSESSMENT OF HYPERSPECTRAL IMAGING SYSTEM FOR POULTRY SAFETY INSPECTION

Authors: Park, Bosoon; Lawrence, Kurt; Windham, William; Smith, Douglas

Submitted to: International Society for Optical Engineering
Publication Type: Proceedings
Publication Acceptance Date: 5/16/2002
Publication Date: 6/10/2002
Citation: Park, B., Lawrence, K.C., Windham, W.R., Smith, D.P. 2002. Assessment of hyperspectral imaging system for poultry safety inspection. International Society for Optical Engineering. 289:269-279.

Interpretive Summary: Identification and separation of poultry carcasses contaminated by feces and/or ingesta is a very important step necessary to protect consumers from a potential source of food poisoning when pathogens enter the food chain. Development of accurate and reliable science-based inspection systems to ensure safe production of poultry processing is an important issue for Hazard Analysis, Critical Control Point (HACCP) compliance. We developed a hyperspectral imaging system for the detection of fecal and ingesta contaminants on the surface of poultry carcasses. The system provides the basic engineering design suitable for use in poultry processing plants. This paper reports on the assessment of a hyperspectral imaging system including several different models to identify fecal and ingesta contaminants. With high accuracy, the system will allow contaminated carcasses to be separated from non-contaminated carcasses prior to treatment to remove the feces and ingesta.

Technical Abstract: A hyperspectral imaging system demonstrated potential to detect surface fecal and ingesta contaminants on poultry carcasses. Hyperspectral data were analyzed with four pre-processing methods considering two parameters: calibration and 20-nm spectral smoothing. A band-ratio image-processing algorithm, using band equation including 2-wavelengths (565 nm / 517 nm) and 3-wavelengths (576 nm - 616 nm)/(529 nm - 616 nm) equations, was then applied to each pre-processed method that included applying a background mask to the ratio of images, and finally applying a fecal threshold. Based on a high accuracy of 96.2% for predicting surface contaminants and significantly less false positives on the 64 birds measured, the calibrated smooth method was considered the best pre-processing method for contaminant detection. In conjunction with an appropriate image-processing algorithm, the hyperspectral imaging system is an effective technique for the identification of fecal and ingesta contaminants on poultry carcasses. Specifically, band ratio with 2-wavelength equation (565/517) performed very well with 96.4% accuracy and 147 false positives for detecting both feces (duodenum, ceca, colon) and ingesta contaminants.