Simple algorithms for the classification of visible/near-infrared and hyperspectral imaging spectra of chicken skins, feces, and fecal contaminated skins

As part of a research aimed at developing multispectral and hyperspectal imaging systems for use in real-time on-line detection of fecal contaminated chicken carcasses, the ingesta, doudenum, ceca, and colon feces, as well as uncontaminated skin of male broilers were scanned on a visible/NIRSystem. Reflectance spectra were recorded over the 400-2498 nm wavelength range at 2-nm intervals and 32 scans. From this spectral range, a principle component analysis (PCA) 2-class model was developed. Spectral and spatial images of the poultry carcasses were collected using a hyperspectral imaging system. Using two algorithms, subtraction and ratio, classification analysis between skins and feces classes was carried out with great success.