PARTIAL LEAST SQUARES REGRESSION OF HYPERSPECTRAL IMAGES FOR CONTAMINANT DETECTION ON POULTRY CARCASSES

Authors: Lawrence, Kurt; Windham, William; Park, Bosoon; HEITSCHMIDT, G - UGA; Smith, Douglas; Feldner, Peggy

Submitted to: Near Infrared Spectroscopy Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/19/2006
Publication Date: 8/18/2006
Citation: Lawrence, K.C., Windham, W.R., Park, B., Heitschmidt, G.W., Smith, D.P., Feldner, P.W. 2006. Partial least squares regression of hyperspectral images for contaminant detection on poultry carcasses. Near Infrared Spectroscopy Journal.

Interpretive Summary: The U.S. Department of Agriculture has developed imaging technology to detect fecal contaminants. These imaging systems are known as multispectral and hyperspectral imaging systems. Multispectral imaging systems can only image at a few discrete wavelengths which limits their ability to fully detect all fecal contaminants on poultry carcasses. Yet, hyperspectral imaging systems can collect images at hundreds of wavelengths but they typically are too slow for real-time applications like fecal detection on poultry carcasses. Now, hyperspectral imaging system can be used with a camera based on CMOS technology that will allow hyperspectral imaging systems to work much faster by collecting only the wavelength information that is important. Therefore, this paper reports on the development of fecal detection models based on discrete wavelength across a broad spectral range. These models use broad-spectrum multivariate statistical analysis technique to detecting contaminants with hyperspectral imaging. Results indicate that using the full-spectral range models were very good at predicting contaminants. Validating the models on large hypercube images resulted in the best validation with about 95% of the over 400 contaminants detected with only 26 false positives (errors of commission). About one-third of the false positives were from bruised wingtips which would no be visible during in-plant commercial processing. This technique shows promise for use in detecting contaminants in poultry processing plants.

Technical Abstract: The U.S. Department of Agriculture has developed multispectral and hyperspectral imaging systems to detect fecal contaminants. Until recently, the hyperspectral imaging system has been used as a research tool to detect a few optimum wavelengths for use in a multispectral imaging system. However, with the development of CMOS cameras, discrete wavelengths or subsets of the full detector range can be used to greatly increase the speed of hyperspectral imaging systems. This paper reports on the use of a broad-spectrum multivariate statistical analysis technique for detecting contaminants with hyperspectral imaging. Partial least squares regression (PLSR) was used for model development. Calibration models from spatially averaged region of interest data were developed with and without smoothing, with and without scatter correction, and with and without first derivative (difference) preprocessing. Results indicate that using the full spectral range with scatter correction was needed for good model development. Furthermore, validation of the various calibration models indicated that preprocessing with scatter correction, nine-point boxcar smoothing, and first derivative preprocessing resulted in the best validation with about 95% of the over 400 contaminants detected with only 26 false positives (errors of commission). About one-third of the false positives were from bruised wingtips which would no be visible during in-plant commercial processing.