Page Banner

United States Department of Agriculture

Agricultural Research Service

Title: Discriminant Analysis of Dual-Wavelength Spectral Images for Classifying Poultry Carcasses

Authors
item Park, Bosoon - UNIVERSITY OF GEORGIA
item Lawrence, Kurt
item Windham, William
item Chen, Yud-Ren -
item Chao, Kevin -

Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: February 10, 2002
Publication Date: March 12, 2002
Citation: PARK, B., LAWRENCE, K.C., WINDHAM, W.R., CHEN, Y., CHAO, K. DISCRIMINANT ANALYSIS OF DUAL-WAVELENGTH SPECTRAL IMAGES FOR CLASSIFYING POULTRY CARCASSES. COMPUTERS AND ELECTRONICS IN AGRICULTURE. 2002.

Interpretive Summary: To ensure a healthy and safe meat supply to consumers, poultry carcasses at poultry processing plants are inspected by Food Safety and Inspection Service (FSIS)inspectors. Inspectors visually inspect approximately 30 to 35 birds per minute for apparent unwholesome carcasses (diseases). These unwholesome carcasses demonstrate a variety of changes in skin and meat color. In order to improve the effectiveness of the formal inspection program, we investigated machine vision with image texture feature analysis as a potential method for discriminating diseased unwholesome carcasses. The results of this research show that a machine vision imaging system can separate the wholesome carcasses from the unwholesome carcasses, allowing human inspectors in poultry slaughter plants to further inspect the condemned carcasses for diseases. The results of this research are useful to the FSIS for improving the federal poultry inspection program and to researchers who are interested in applying machine vision technique for grading or inspection of agricultural and food products.

Technical Abstract: An analysis of texture features, based on co-occurrence matrices (COMs), was conducted to determine the performance of dual-wavelength imaging for discriminating unwholesome poultry carcasses from wholesome carcasses. The variance, sum average, sum variance, and sum entropy of COMs were the most significant texture features (P < 0.005) for identifying unwholesome poultry carcasses. When the COM direction was equal to 0o, the contrast was lower and the inverse difference moment and difference variance were higher (P < 0.01) than any other COM direction in the visible spectral images. The characteristics of variance and sum variance texture features varied with the wavelength of spectral images and with condemnation of poultry carcasses, as well. The sum variance of wholesome carcasses was higher (P < 0.005) than unwholesome carcasses for spectral images at 542 nm. For 700 nm images, a quadratic discriminant model was able to identify unwholesome and wholesome carcasses with a classification accuracy of 83.9 percent and wholesome carcasses with 97.1 percent, respectively. However, a single quadratic discriminant model was not acceptable for identifying three different carcasses (wholesome, septicemic and cadaver), because of a low accuracy for cadaver carcasses. In this case, the accuracies were 91.2 percent for wholesome, 93.3 percent for septicemic and 72.9 percent for cadaver, respectively. A dual-wavelength imaging system with selected optical filters of 542 and 700-nm wavelengths were feasible for detecting unwholesome poultry carcasses with high accuracy.

Last Modified: 10/24/2014