Submitted to: Transactions of the ASABE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: April 21, 2007
Publication Date: August 30, 2007
Citation: Yoon, S.C., Lawrence, K.C., Park, B., Windham, W.R. 2007. STATISTICAL MODEL-BASED THRESHOLDING OF MULTISPECTRAL IMAGES FOR CONTAMINATE DETECTION ON POULTRY CARCASSES. Transactions of the ASABE. 50(4):1433-1422. 2007. Interpretive Summary: The Food Safety and Inspection Service in USDA established a zero tolerance policy on fecal contaminants visible on processed poultry carcasses. While the industry still relies on human inspectors to detect surface fecal contaminants, the Agricultural Research Service (ARS) has been developing a high-tech camera system in order to find fecal contaminants on poultry carcasses. The ARS imaging system consists of a hardware module including light sources, a camera, and a computer as well as a software module to process the image and to make a decision of fecal contamination. The data collected by the ARS imaging system provides both spatial and spectral information which is crucial for the detection and identification of feces. The objective of this study was to design an image processing algorithm for reducing false positive rates while achieving high detection accuracy. Data distributions of feces and clean carcasses were modeled by statistical estimation methods. A mathematical function was designed to make a decision about the presence of any feces at each bird. When fully implemented into the ARS imaging system, this technique can improve the system performance.
Technical Abstract: Developing an algorithm to decide the presence or absence of fecal contamination on the surface of poultry carcasses is critical to food safety. The global threshold strategy for a band-ratio algorithm has been known to be limited to pixel-basis detection. In an attempt to develop a statistical decision rule for carcass-basis detection from multispectral images, probability density functions of both contaminated and uncontaminated materials were estimated by parametric and nonparametric methods. We found that uncontaminated poultry carcasses could be modeled by a Gaussian distribution, whereas contaminated materials were non-Gaussian. A kernel density estimator was used to analyze the non-Gaussian characteristic of the fecal material on a transformed projection axis. A linear mixture of the density functions was introduced to model the observations made on the projection axis. A new detection algorithm was designed using the mixture model and tested for 112 birds (56 dirty and 56 clean birds). A test on the sample birds revealed that the algorithm needed at least 12 contaminated pixels to reach the perfect detection results. The test also showed a false positive rate of less than 5%.