Gerald W. Heitschmidt

Partial least squares regression of hyperspectral images for contaminant detection on poultry carcasses

The US Department of Agriculture has developed multispectral and hyperspectral imaging systems to detect faecal 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 complementary metal oxide semiconductor 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 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) pre-processing. 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 pre-processing with scatter correction, nine-point boxcar smoothing and first derivative pre-processing 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 not be visible during in-plant commercial processing.

Modified Pressure System for Imaging Egg Cracks

One aspect of grading table eggs is shell checks or cracks. Currently, USDA voluntary regulations require that humans grade a representative sample of all eggs processed. However, as processing plants and packing facilities continue to increase their volume and throughput, human graders are having difficulty matching the pace of the machines. Additionally, some plants also have a problem with micro-cracks that the graders often miss because they are very small and hard to see immediately post-processing but grow and become readily apparent before they reach market. An imaging system was developed to help the grader detect these small micro-cracks. The imaging system utilized one image captured at atmospheric pressure and a second at a slight negative pressure to enhance the crack and make detection much easier. A simple image processing algorithm was then applied to the ratio of these two images and the resulting image, containing both cracked and/or intact eggs were color-coded to simplify identification. The imaging system was capable of imaging 15 eggs in about ¾ second and the algorithm processing took about another 10 seconds. These times could easily be reduced with a dedicated, multi-threaded computer program. In analyzing 1000 eggs, the system was 99.6% accurate overall with only 0.3% false positives compared to 94.2% accurate overall for the human graders with 1.2% false positives. An international patent on the system was filed and further automation of the system is needed.

Modified pressure imaging for egg crack detection and resulting egg quality

Cracks in the shell surface compromise the primary barrier for external microbial contamination of the egg. Microcracks are very small cracks in the shell surface that are difficult to detect by human graders. New technology has been developed that uses modified pressure and imaging to detect microcracks in eggs. Research has shown the system to have an accuracy of 99.6% in detecting both cracked and intact eggs. A study was undertaken to determine if quality differences existed between modified pressure imaged and control eggs during extended cold storage. Three replicates were conducted with eggs stored at 4 degrees C for 5 wk with weekly quality testing. The physical quality factors monitored were Haugh units, albumen height, egg weight, shell strength, vitelline membrane strength and elasticity, and whole egg total solids. All measurements were conducted on individual eggs (12/treatments per replicate) each week with the exception of whole egg solids, which were determined from 3 pools (4 eggs each)/treatment per replicate each week. Percentage of whole egg total solids was the only significant difference (P < 0.05) between treatments (23.65% modified pressure imaged and 23.47% control). There was a significant difference (P < 0.05) for egg weight between replicates (60.82, 58.02, and 60.58 g for replicates 1, 2, and 3, respectively). Therefore, imaging eggs in the modified pressure system for microcrack detection did not alter egg quality during extended cold storage. Utilizing the modified pressure crack detection technology would result in fewer cracked eggs reaching the consumer, consequently enhancing food safety without affecting product quality.

Feasibility of detecting aflatoxin B1 on inoculated maize kernels surface using Vis/NIR hyperspectral imaging.

The feasibility of using a visible/near-infrared hyperspectral imaging system with a wavelength range between 400 and 1000 nm to detect and differentiate different levels of aflatoxin B1 (AFB1 ) artificially titrated on maize kernel surface was examined. To reduce the color effects of maize kernels, image analysis was limited to a subset of original spectra (600 to 1000 nm). Residual staining from the AFB1 on the kernels surface was selected as regions of interest for analysis. Principal components analysis (PCA) was applied to reduce the dimensionality of hyperspectral image data, and then a stepwise factorial discriminant analysis (FDA) was performed on latent PCA variables. The results indicated that discriminant factors F2 can be used to separate control samples from all of the other groups of kernels with AFB1 inoculated, whereas the discriminant factors F1 can be used to identify maize kernels with levels of AFB1 as low as 10 ppb. An overall classification accuracy of 98% was achieved. Finally, the peaks of β coefficients of the discrimination factors F1 and F2 were analyzed and several key wavelengths identified for differentiating maize kernels with and without AFB1 , as well as those with differing levels of AFB1 inoculation. Results indicated that Vis/NIR hyperspectral imaging technology combined with the PCA-FDA was a practical method to detect and differentiate different levels of AFB1 artificially inoculated on the maize kernels surface. However, indicated the potential to detect and differentiate naturally occurring toxins in maize kernel.

Development of real-time line-scan hyperspectral imaging system for online agricultural and food product inspection

This paper reports a recent development of a line-scan hyperspectral imaging system for real-time multispectral imaging applications in agricultural and food industries. The hyperspectral imaging system consisted of a spectrograph, an EMCCD camera, and application software. The real-time multispectral imaging with the developed system was possible due to (1) data binning, especially a unique feature of the EMCCD sensor allowing the access to non-contiguous multispectral bands, (2) an image processing algorithm designed for real-time multispectral imaging, and (3) the design and implementation of the real-time application software. The imaging system was developed as a poultry inspection instrument determining the presence of surface feces on poultry carcasses moving at commercial poultry processing line speeds up to 180 birds per minute. The imaging system can be easily modifiable to solve other real-time inspection/sorting problems. Three wavelengths at 517 nm, 565 nm and 802 nm were selected for real-time fecal detection imaging. The fecal detection algorithm was based on dual band ratios of 565nm/517nm and 802nm/517nm followed by thresholding. The software architecture was based on a ping pong memory and a circular buffer for the multitasking of image grabbing and processing. The software was written in Microsoft Visual C++. An image-based internal triggering (i.e. polling) algorithm was developed to determine the start and end positions of birds. Twelve chickens were used for testing the imaging system at two different speeds (140 birds per minute and 180 bird per minute) in a pilot-scale processing line. Four types of fecal materials (duodenum, ceca, colon and ingesta) were used for the evaluation of the detection algorithm. The software grabbed and processed multispectral images of the dimension 118 (line scans) x 512 (height) x 3 (bands) pixels obtained from chicken carcasses moving at the speed up to 180 birds per minute (a frame rate 286 Hz). Intensity calibration, detection algorithm, displaying and saving were performed within the real-time deadlines.

Line-scan hyperspectral imaging for real-time poultry fecal detection

The ARS multispectral imaging system with three-band common aperture camera was able to inspect fecal contaminants in real-time mode during poultry processing. Recent study has demonstrated several image processing methods including binning, cuticle removal filter, median filter, and morphological analysis in real-time mode could remove false positive errors. The ARS research groups and their industry partner are now merging the fecal detection and systemically disease detection systems onto a common platform using line-scan hyperspectral imaging system. This system will aid in commercialization by creating one hyperspectral imaging system with user-defined wavelengths that can be installed in different locations of the processing line to solve significant food safety problems. Therefore, this research demonstrated the feasibility of line-scan hyperspectral imaging system in terms of processing speed and detection accuracy for a real-time, on-line fecal detection at current processing speed (140 birds per minute) of commercial poultry plant. The newly developed line-scan hyperspectral imaging system could improve Food Safety Inspection Service (FSIS)s poultry safety inspection program significantly.

Hyperspectral image reconstruction using RGB color for foodborne pathogen detection on agar plates

This paper reports the latest development of a color vision technique for detecting colonies of foodborne pathogens grown on agar plates with a hyperspectral image classification model that was developed using full hyperspectral data. The hyperspectral classification model depended on reflectance spectra measured in the visible and near-infrared spectral range from 400 and 1,000 nm (473 narrow spectral bands). Multivariate regression methods were used to estimate and predict hyperspectral data from RGB color values. The six representative non-O157 Shiga-toxin producing Eschetichia coli (STEC) serogroups (O26, O45, O103, O111, O121, and O145) were grown on Rainbow agar plates. A line-scan pushbroom hyperspectral image sensor was used to scan 36 agar plates grown with pure STEC colonies at each plate. The 36 hyperspectral images of the agar plates were divided in half to create training and test sets. The mean Rsquared value for hyperspectral image estimation was about 0.98 in the spectral range between 400 and 700 nm for linear, quadratic and cubic polynomial regression models and the detection accuracy of the hyperspectral image classification model with the principal component analysis and k-nearest neighbors for the test set was up to 92% (99% with the original hyperspectral images). Thus, the results of the study suggested that color-based detection may be viable as a multispectral imaging solution without much loss of prediction accuracy compared to hyperspectral imaging.

Salmonella contamination in shell eggs exposed to modified-pressure imaging for microcrack detection

Microcracks in egg shells are a food safety risk and are difficult for professional human graders to detect. Modified-pressure imaging technology with 99.6% accuracy has been developed to detect microcracks. This study was conducted to determine whether the microcrack detection system would increase penetration of Salmonella into egg contents or lead to cross-contamination within the system. Thirty dozen grade A large white retail eggs were used for each of 3 replicates. Cracked eggs were removed and 72 eggs/replicate were dip inoculated in buffered peptone water containing 10(5) cfu/mL of nalidixic acid-resistant Salmonella Typhimurium (ST), whereas 144 eggs were dipped in sterile buffered peptone water. All eggs were incubated overnight at 25°C before imaging. Forty-five eggs of each treatment were imaged in the following order: control, inoculated, control. Imaged and nonimaged eggs from each treatment were used for cultural analysis of a shell rinse, shell emulsion, and contents sample for each egg. The ST levels were monitored on brilliant green sulfa agar with 200 mg/L of nalidixic acid. Egg contents were also enriched to determine the prevalence of ST in low levels. Salmonella Typhimurium was not detected on or in any of the control eggs, including the eggs imaged after the inoculated eggs. The highest level of ST was detected in inoculated shell emulsions (4.79 log cfu/mL). No differences in ST levels were found for any sample location between imaged and nonimaged inoculated eggs. Therefore, the modified-pressure imaging system for microcrack detection did not result in microbial cross-contamination or increase the level of microbial penetration in inoculated eggs. The imaging system can be used to assess eggs for cracks without negative food safety implications.

Hyperspectral imaging using a color camera and its application for pathogen detection

This paper reports the results of a feasibility study for the development of a hyperspectral image recovery (reconstruction) technique using a RGB color camera and regression analysis in order to detect and classify colonies of foodborne pathogens. The target bacterial pathogens were the six representative non-O157 Shiga-toxin producing Escherichia coli (STEC) serogroups (O26, O45, O103, O111, O121, and O145) grown in Petri dishes of Rainbow agar. The purpose of the feasibility study was to evaluate whether a DSLR camera (Nikon D700) could be used to predict hyperspectral images in the wavelength range from 400 to 1,000 nm and even to predict the types of pathogens using a hyperspectral STEC classification algorithm that was previously developed. Unlike many other studies using color charts with known and noise-free spectra for training reconstruction models, this work used hyperspectral and color images, separately measured by a hyperspectral imaging spectrometer and the DSLR color camera. The color images were calibrated (i.e. normalized) to relative reflectance, subsampled and spatially registered to match with counterpart pixels in hyperspectral images that were also calibrated to relative reflectance. Polynomial multivariate least-squares regression (PMLR) was previously developed with simulated color images. In this study, partial least squares regression (PLSR) was also evaluated as a spectral recovery technique to minimize multicollinearity and overfitting. The two spectral recovery models (PMLR and PLSR) and their parameters were evaluated by cross-validation. The QR decomposition was used to find a numerically more stable solution of the regression equation. The preliminary results showed that PLSR was more effective especially with higher order polynomial regressions than PMLR. The best classification accuracy measured with an independent test set was about 90%. The results suggest the potential of cost-effective color imaging using hyperspectral image classification algorithms for rapidly differentiating pathogens in agar plates.

Improved egg crack detection algorithm for modified pressure imaging system

Shell eggs with microcracks are often undetected during egg grading processes. In the past, a modified pressure imaging system was developed to detect eggs with microcracks without adversely affecting the quality of normal intact eggs. The basic idea of the modified pressure imaging system was to apply a short burst of vacuum within a transparent chamber in order to cause a momentary and forced opening in the egg shell with a crack and thus to utilize the changes in image intensities during this process. The intensity changes from dark to bright in the shell surface were recorded by a highresolution digital camera and processed by an image ratio technique. The performance of the imaging system, however, was sometimes compromised by false readings due to motion of intact eggs relative to the camera. The uneven movement of the lid hinged on the chamber was considered as the main cause of motion errors. In this paper, a machine vision technique to compensate the motion errors was developed to reduce the false detection readings caused by motion of intact eggs. The developed motion compensation algorithm is based on motion estimation of individual eggs.