Page Banner

United States Department of Agriculture

Agricultural Research Service

Title: Using Fuzzy Inference Systems for Assessment of Color Imaging on Chicken Livers

Authors
item Chao, Kuanglin
item Chen, Yud
item Park, Bosoon
item Early, Howard - FSIS, WASHINGTON, DC

Submitted to: ASAE Annual International Meeting
Publication Type: Proceedings
Publication Acceptance Date: July 17, 1998
Publication Date: N/A

Interpretive Summary: Poultry and poultry products have increased in popularity with U.S. consumers in recent years. With the increased demand for poultry products, food safety becomes an increasingly important issue. Automated (i.e., machine vision based) poultry inspection systems are currently being developed at the Instrumentation and Sensing Lab, ARS, USDA, to improve the reliability and productivity in poultry processing. This paper reports the results of applying color imaging technique to identify individual conditions of condemned poultry. A knowledge-based neuro-fuzzy classifier was developed to identify the condemnation conditions for individual poultry. Statistical inference extracted from color viscera images and information from field experts were utilized to construct the neuro-fuzzy classifier (NFC). The NFC can perform heuristic learning from data. The classifier performed with 93.8% accuracy for the separation of normal livers from the airsacculitis livers. This information is useful to the Food Safety and Inspection Service (FSIS), poultry processing plants, and researchers who are interested in food safety inspection or agricultural products classification based on the machine vision techniques.

Technical Abstract: A fuzzy inference system (FIS) for poultry postmortem classifications using color viscera images is presented. The viscera images consist of color features in the red, green, and blue (RGB) format and are processed by a FIS, which integrates knowledge-based techniques to handle both linguistic and numerical descriptions of feature images. The FIS uses data gathered from color images to generate a set of fuzzy if-then rules. Knowledge extracted from statistic analysis was used to refine the fuzzy rule base. A neuro-fuzzy classifier was derived from the generic model of a three-layer fuzzy perceptron. The classifier performed with 93.8% accuracy for the separation of normal livers from the airsacculitis livers. This paper presents the background for development and methods for incorporation of new fuzzy logic relationships employed to assess the color viscera images for the automated poultry inspection systems.

Last Modified: 7/24/2014