Off-campus UNL users: To download campus access dissertations, please use the following link to log into our proxy server with your NU ID and password. When you are done browsing please remember to return to this page and log out.
Non-UNL users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Development and evaluation of spectral imaging systems and algorithms for beef tenderness grading
Beef tenderness is an important palatability trait and is related to consumer satisfaction. In this dissertation, hyperspectral imaging (HSI) was implemented and tested for beef tenderness assessment. An acousto-optic tunable filter (AOTF)-based and a spectrograph-based portable HSI system (wavelength range: 400 nm to 1000 nm) were developed. These systems were used to acquire hyperspectral images of fresh beef ribeye muscle (longissimus dorsi) on hanging beef carcasses at 2-day postmortem in multiple commercial beef packing plants. The spectral dimension of the beef images was reduced, image features were extracted, and discriminant models were developed to forecast 14-day aged, cooked beef tenderness. Four different spectral dimensionality reduction methods (sample, chemometric, and mosaic principal component analyses, and partial least squares regression), seven different image feature sets (descriptive statistical features, wavelet features, gray level co-occurrence matrix features, Gabor transform features, Laws features, local binary pattern features, and pooled features), and three different discriminant models (Fisher's linear discriminant analysis, support vector machines, and decision tree) were evaluated. A third-party true validation was conducted to evaluate the performance of the HSI systems. A new evaluation metric, called Accuracy Index (AI), was developed and used to compare beef tenderness prediction models. The AOTF system provided an AI value and accuracies for tender certification, tender classification, and tough classification of 70%, 91.7%, 84.1%, and 56.9%, respectively. The corresponding metrics for the spectrograph system were 66.8%, 86.7%, 65.8%, and 63.6%, respectively. In addition, a multispectral imaging approach was implemented by identifying and analyzing five key wavelengths. This approach provided an AI value and accuracies for tender certification, tender classification, and tough classification of 67.8%, 87.5%, 62%, and 68.2%, respectively, and took only eight seconds to assign a tenderness score for a beef carcass. Both hyperspectral and multispectral imaging show outstanding potential for real-time beef tenderness assessment. The successful adoption of this technology will lead to the development of a value-based system that benefits both consumers and the beef industry. ^ Keywords: Beef tenderness, hyperspectral imaging, multispectral imaging, principal component analysis, and textural features.^
Agriculture, Food Science and Technology|Engineering, Computer|Engineering, Agricultural
Konda Naganathan, Govindarajan, "Development and evaluation of spectral imaging systems and algorithms for beef tenderness grading" (2011). ETD collection for University of Nebraska - Lincoln. AAI3466570.