Submitted to: Society of Photo-Optical Instrumentation Engineers
Publication Type: Proceedings
Publication Acceptance Date: 12/28/2000
Publication Date: 12/28/2000
Citation: N/A Interpretive Summary: Due to massive production of poultry and the inherent variability and complexity in individual birds, there are great challenges for improvement of the existing organoleptic inspection methods. Development of high speed and reliable inspection systems to ensure safe production of poultry during post-harvest processing has become an important issue, as the public is demanding assurance of better and safer food. This paper reports the results of applying multispectral imaging to inspect chicken skin tumors. Chicken carcass images (60 tumorous and 20 normal) were examined by the system. Statistical analysis of tumorous and normal chickens was performed to select wavelengths for a multispectral imaging system. Knowledge-based classification systems were developed to identify tumorous areas from chicken carcasses. The classifier gave success rates of 91 percent and 86 percent for normal and tumorous skin tissue, respectively. This information nis useful to the Food Safety and Inspection Service (FSIS) and poultry processing plants.
Technical Abstract: Hyperspectral and multispectral imaging techniques were used to detect chicken skin tumors. Hyperspectral images of eight tumorous chickens were taken in the spectral range of 420-850 nm. Principle component analysis was applied to select useful wavelength bands (465, 575, 705 nm) from the tumorous chicken images. A multispectral imaging system capable of simultaneously capturing three registered images was used to image 60 tumorous and 20 normal chickens. Multispectral image analysis was performed to generate ratioed images, which were then divided into regions of interest (ROI's) classified as either tumorous or normal by a veterinarian. Image features for each ROI (coefficient of variation, skewness and kurtosis) were extracted for use as inputs to fuzzy classifiers. The fuzzy classifiers were able to separate normal from tumorous skin with increasing accuracies as more features were used. In particular, use of all three features gave successful detection rates of 9 percent and 86 percent for normal and tumorous tissue, respectively.