Chun-Chieh Yang

Hyperspectral-Multispectral Line-Scan Imaging System for Automated Poultry Carcass Inspection Applications for Food Safety

A hyperspectral-multispectral line-scan imaging system was developed for differentiation of wholesome and systemically diseased chickens. In-plant testing was conducted for chickens on a commercial evisceration line moving at a speed of 70 birds per minute. Hyperspectral image data was acquired for a calibration data set of 543 wholesome and 64 systemically diseased birds and for a testing data set of 381 wholesome and 100 systemically diseased birds. The calibration data set was used to develop the parameters of the imaging system for conducting multispectral inspection based on fuzzy logic detection algorithms using selected key wavelengths. Using a threshold of 0.4 for fuzzy output decision values, multispectral classification was able to achieve 90.6% accuracy for wholesome birds and 93.8% accuracy for systemically diseased birds in the calibration data set and 97.6% accuracy for wholesome birds and 96.0% accuracy for systemically diseased birds in the testing data set. By adjusting the classification threshold, 100% accuracy was achieved for systemically diseased birds with a decrease in accuracy for wholesome birds to 88.7%. This adjustment shows that the system can be feasibly adapted as needed for implementation for specific purposes, such as paw harvesting operations or prescreening for food safety inspection. This line-scan imaging system is ideal for directly implementing multispectral classification methods developed from hyperspectral image analysis.

Machine vision system for on-line wholesomeness inspection of poultry carcasses.

A line-scan machine vision system and multispectral inspection algorithm was developed and evaluated for differentiation of wholesome and systemically diseased chickens on a high-speed processing line. The inspection system acquires line-scan images of chicken carcasses on a 140 bird/min processing line and is able to automatically detect individual birds entering and exiting the field of view of the camera, locate a specified region of interest for spectral image analysis, and produce a decision output for each bird. The same spectral line-scan imaging system was used for hyperspectral data acquisition-analysis to develop the multispectral detection and differentiation algorithm and for multispectral implementation of the algorithm for real-time on-line inspection on the processing line. Results showed that effective multispectral inspection could be achieved by analysis of a selected region of interest across the breast area from images at the 580- and 620-nm wavebands. Overall system performance was evaluated during two 8-h shifts in which the system inspected over 100,000 chickens, with system results compared with Food Safety and Inspection Service inspector tallies of wholesome and systemically diseased birds for that same time period. During system verification, the system accurately classified wholesome and systemically diseased chickens that were observed by a veterinarian posted beside the system to perform real-time identifications of the same birds. The high accuracy of the results demonstrated that the spectra line-scan imaging system and multispectral detection and differentiation algorithm can be effectively used for on-line high-speed presorting applications for young broiler chickens.

Classification of fecal contamination on leafy greens by hyperspectral imaging

This paper reported the development of hyperspectral fluorescence imaging system using ultraviolet-A excitation (320-400 nm) for detection of bovine fecal contaminants on the abaxial and adaxial surfaces of romaine lettuce and baby spinach leaves. Six spots of fecal contamination were applied to each of 40 lettuce and 40 spinach leaves. In this study, the wavebands at 666 nm and 680 nm were selected by the correlation analysis. The two-band ratio, 666 nm / 680 nm, of fluorescence intensity was used to differentiate the contaminated spots from uncontaminated leaf area. The proposed method could accurately detect all of the contaminated spots.

Application of Multispectral Imaging for Identification of Systemically Diseased Chicken

A multispectral imaging system for automated inspection of wholesome and systemically diseased chickens was developed and demonstrated. The disease of septicemia was selected as the detection target because it is the most common chicken disease in US. From visible/near-infrared reflectance spectra of poultry carcasses, average CIELAB L*(lightness), a*(redness), and b*(yellowness) values were analyzed. The difference of lightness between wholesome and septicemic chickens was significant. The multispectral imaging system consisted of a backilluminated CCD camera and a spectrometer with four narrow-band interference filters for 488, 540, 580, and 610 nm wavelengths, respectively. The 16-bit multispectral images of chicken carcasses were collected for image processing and analysis. Image processing algorithms, including image registration, flat-field correction, image segmentation, region of interest identification, feature measurement, and symptom recognition, were developed to differentiate septicemic chickens from wholesome ones. The image from 610-nm wavelength was used to create a mask to extract chicken images from background. The average reflectance intensities at 488, 540, 580, and 610 nm from different parts of the carcass in the front side were calculated. Moreover, four normalization methods and four normalized differentiation methods between two wavelengths were also calculated for comparison. Decision tree was applied to generate thresholds for differentiation between wholesome and septicemic chickens. Images were collected at three time frames. The images from the first time frame were used to generate first thresholds that were tested by the images from the second time frames. Then, the images from the first and second time frames were used together to generate second thresholds. The first and second thresholds were tested by the images from the third time frame, respectively. The results showed that using average intensity at 580 nm from the region of interest, 98.6% of septicemic chickens and 96.3% of wholesome chickens could be differentiated from each other. More training data could help to generate more appropriate thresholds used at different time frames.

The development of the line-scan image recognition algorithm for the detection of frass on mature tomatoes

In this research, a multispectral algorithm derived from hyperspectral line-scan fluorescence imaging under violet LED excitation was developed for the detection of frass contamination on mature tomatoes. The algorithm utilized the fluorescence intensities at two wavebands, 664 nm and 690 nm, for computation of the simple ratio function for effective detection of frass contamination. The contamination spots were created on the tomato surfaces using four concentrations of aqueous frass dilutions. The algorithms could detect more than 99% of the 0.2 g/ml and 0.1 g/ml frass contamination spots and successfully differentiated these spots from clean tomato surfaces. The results demonstrated that the simple multispectral fluorescence imaging algorithms based on violet LED excitation can be appropriate to detect frass on tomatoes in high-speed post-harvest processing lines.

Differentiation of Beef and Fish Meals in Animal Feeds Using Chemometric Analytic Models

Purpose: The research presented in this paper applied the chemometric analysis to the near-infrared spectral data fromline-scanned hyperspectral images of beef and fish meals in animal feeds. The chemometric statistical models weredeveloped to distinguish beef meals from fish ones. Methods: The meal samples of 40 fish meals and 15 beef meals wereline-scanned to obtain hyperspectral images. The spectral data were retrieved from each of 3600 pixels in the Region ofInterest (ROI) of every sample image. The wavebands spanning 969 nm to 1551 nm (across 176 spectral bands) wereselected for chemometric analysis. The partial least squares regression (PLSR) and the principal component analysis (PCA)methods of the chemometric analysis were applied to the model development. The purpose of the models was to correctlyclassify as many beef pixels as possible while misclassified fish pixels in an acceptable amount. Results: The results showedthat the success classification rates were 97.9% for beef samples and 99.4% for fish samples by the PLSR model, and 85.1%for beef samples and 88.2% for fish samples by the PCA model. Conclusion: The chemometric analysis-based PLSR and PCAmodels for the hyperspectral image analysis could differentiate beef meals from fish ones in animal feeds.

Inspection of Fecal Contamination on Strawberries Using Line-Scan LED-Induced Fluorescence Imaging Techniques

In the United States, fecal contamination of produce is a food safety issue associated with pathogens such as Escherichia coli and Salmonella that can easily pollute agricultural products via animal and human fecal matters. Outbreaks of foodborne illnesses associated with consuming raw fruits and vegetables in the US have occurred more frequently in recent years. This problem not only threatens public health, but also results in a huge amount of unnecessary economic losses every year. Therefore, developing optical sensing technologies for the detection of contaminants on fresh produce is urgent and essential. Among fruits, strawberry is one high-potential vector of fecal contamination and foodborne illnesses since the fruit is often consumed raw and with minimal processing. In the present study, line-scan LED-induced fluorescence imaging techniques were applied for inspection of fecal material on strawberries, and the spectral characteristics and specific wavebands of strawberries were determined by detection algorithms. The results indicated that the combination of two-waveband intensity ratios, such as 680 nm / 688 nm and 680 nm / 704 nm, can successfully distinguish fecal contamination, uncontaminated surfaces, and leaves. The binary images showed that the algorithm could successfully detect all of the fecal contamination spots that were applied to the strawberry surfaces. The results of this study would improve the safety and quality of produce consumed by the public.

Fast and accurate image recognition algorithms for fresh produce food safety sensing

This research developed and evaluated the multispectral algorithms derived from hyperspectral line-scan fluorescence imaging under violet LED excitation for detection of fecal contamination on Golden Delicious apples. The algorithms utilized the fluorescence intensities at four wavebands, 680 nm, 684 nm, 720 nm, and 780 nm, for computation of simple functions for effective detection of contamination spots created on the apple surfaces using four concentrations of aqueous fecal dilutions. The algorithms detected more than 99% of the fecal spots. The effective detection of feces showed that a simple multispectral fluorescence imaging algorithm based on violet LED excitation may be appropriate to detect fecal contamination on fast-speed apple processing lines.

Physical and mechanical properties of spinach for whole-surface online imaging inspection

The physical and mechanical properties of baby spinach were investigated, including density, Youngs modulus, fracture strength, and friction coefficient. The average apparent density of baby spinach leaves was 0.5666 g/mm3. The tensile tests were performed using parallel, perpendicular, and diagonal directions with respect to the midrib of each leaf. The test results showed that the mechanical properties of spinach are anisotropic. For the parallel, diagonal, and perpendicular test directions, the average values for the Youngs modulus values were found to be 2.137MPa, 1.0841 MPa, and 0.3914 MPa, respectively, and the average fracture strength values were 0.2429 MPa, 0.1396 MPa, and 0.1113 MPa, respectively. The static and kinetic friction coefficient between the baby spinach and conveyor belt were researched, whose test results showed that the average coefficients of kinetic and maximum static friction between the adaxial (front side) spinach leaf surface and conveyor belt were 1.2737 and 1.3635, respectively, and between the abaxial (back side) spinach leaf surface and conveyor belt were 1.1780 and 1.2451 respectively. These works provide the basis for future development of a whole-surface online imaging inspection system that can be used by the commercial vegetable processing industry to reduce food safety risks.

Poultry carcass inspection by a fast line-scan imaging system: results from in-plant testing

During in-plant testing of a hyperspectral line-scan imaging system, images were acquired of wholesome and systemically diseased chickens on a commercial processing line moving at a speed 70 birds per minute. A fuzzy logic based algorithm using four key wavelengths, 468 nm, 501 nm, 582 nm, 629 nm, was developed using image data from the validation set of images of 543 wholesome and 66 systemically diseased chickens. A classification method using the fuzzy logic based algorithm was then tested on the testing set of images of 457 wholesome and 37 systemically diseased chickens, as well as 80 systemically diseased chickens that were imaged off-shift during breaks between normal processing shifts of the chicken plant. The classification method correctly identified 89.7% of wholesome chicken images and 98.5% of systemically diseased chicken images in the validation set. For the testing data set, the method correctly classified 96.7 % of 457 wholesome chicken images and 100% of 37 systemically diseased chicken images. The 80 images acquired off-shift were also 100% correctly identified.