Skip to main content
ARS Home » Research » Publications at this Location » Publication #149109
Title: ANALYSIS OF HYPERSPECTRAL FLUORESCENCE IMAGES FOR POULTRY SKIN TUMOR INSPECTION

Author
item KONG, SEONG - U.TN, KNOXVILLE, TN
item Chen, Yud
item KIM, INTAEK - U.MYONGJI, S.KOREA
item Kim, Moon

Submitted to: Applied Optics
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/14/2003
Publication Date: 2/1/2004
Citation: Kong, S.G., Chen, Y.R., Kim, I., Kim, M.S. 2004. Analysis of hyperspectral fluorescence images for poultry skin tumor inspection. Applied Optics. 43(2):1-10.

Interpretive Summary: Animal carcasses with pathological problems must be detected and removed from food processing lines. FSIS employs approximately 2200 federal inspectors to inspect about 9 billion birds per year. Each federal inspector is often required to examine 30-35 poultry samples per minute for eight hours a day. The inspection is a highly repetitive and tedious task. Such working conditions can lead to repetitive motion injuries, attention and fatigue problems. One of many tasks performed by the federal inspectors is to detect poultry skin tumors and to decide if the carcass is unfit for human consumption and to be condemned. Tumors are ulcerous lesions that are surrounded by a rim of thickened skin and dermis. They are not easily detected by conventional vision systems operating in the visual spectrum. This paper presents an analysis of hyperspectral fluorescence images for detecting skin tumors on poultry carcasses. Analysis shows that the hyperspectral fluorescence imaging system developed by the Instrument and Sensing Laboratory, together with fuzzy inference scheme, can be effective for inspection of skin tumors on poultry carcasses. This information is of interest to researchers working on the detection of surface skin tumors, and to FSIS which is interested in implementing on-line poultry inspection system.

Technical Abstract: This paper presents a hyperspectral fluorescence imaging system with fuzzy inference scheme for detecting skin tumors on poultry carcasses. Hyperspectral images reveal spatial and spectral information useful for finding pathological legions or contaminants on agricultural products. A skin tumor is not visually obvious since its signature appears as shape distortion rather than discoloration. Fluorescence imaging visualizes poultry skin tumors more easily than reflectance. Hyperspectral image samples for poultry tumor inspection are obtained with 65 spectral bands of fluorescence in the visible region of the spectrum at wavelengths ranging from 425[nm] to 711[nm]. A large amount of hyperspectral image data is compressed using the discrete wavelet transform in the spatial domain. Principal component analysis finds an effective representation of the spectral signals of each pixel for compression in the spectral domain. Features are extracted from the two major spectral peaks of relative fluorescence intensity that are identified as meaningful spectral bands for detecting tumors. A fuzzy inference scheme, with a small number of fuzzy rules and Gaussian membership functions, successfully detects skin tumors on poultry carcasses. Spatial filtering techniques significantly reduce false positives.