|Park, Bosoon - UNIVERSITY OF MARYLAND|
Submitted to: Applied Engineering in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: September 5, 2000
Publication Date: January 1, 2001
Interpretive Summary: Inspection of poultry viscera is one of the important tasks currently performed by human inspectors at poultry slaughter plants. Full automation of poultry inspection awaits the development of techniques that can effectively identify individually contaminated conditions of poultry viscera. This study was conducted to determine the feasibility of a novel multi-spectral system for detection of diseases from images of individual chicken hearts. Optimum wavelengths were selected by stepwise discriminate analysis of visible/NIR spectral data. Gray level image intensities at the selected wavelengths were used for disease detection. The single and difference-wavelength spectral images of chicken hearts were compared to identify individual chicken diseases. Classification accuracy above 90% was achieved for separating normal, airsacculitis, ascites, and septicemia. The multi-spectral imaging system shows promise for real-time identification of poultry condemnation categories by automated image analysis of chicken hearts. This information is also useful to researchers and FSIS scientists who are interested in an automated process for poultry inspection.
Technical Abstract: Optical spectral reflectance and multi-spectral image analysis techniques were investigated to characterize chicken hearts for real-time disease detection. Spectral signatures of five categories of chicken hearts (airsacculitis, ascites, normal, cadaver, and septicemia) were obtained from optical reflectance measurements taken with a visible/near-infrared spectroscopic system in the range 473-974 nm. Multivariate statistical analysis was applied to select the most significant wavelengths from the chicken heart reflectance spectra. By optimizing the selection of wavelengths of interest for different poultry diseases, four wavelengths at 495, 535, 585, and 605 nm were selected. The multi-spectral image system utilizes four narrow-band filters to provide four spectrally discrete images on a single CCD focal-plane. Using the filters at the wavelengths selected from the reflectance spectra, it was possible to easily implement multi-spectral arithmetic operations for disease detection. Based on analysis (t-test) of spectral image data, multi-spectral imaging method had potential to differentiate individual diseases in chicken hearts. All conditions except cadaver were shown to be separable (92 to 100%) by discriminant algorithms involving differences of average image intensities. The results show that multi-spectral imaging may provide a robust method to identify poultry diseases with high potential for real-time application.