Research On Automatic Inspections Of Grading Features Of Beef Carcasses And Beef Grading System

Effective beef grading systems play a positive role on the production, circulation, pricing, consumption of the beef, and also significant in cattle species breeding. Improved standards of the grading will promote the beef industry development reasonably and scientifically.Currently the mainly method for beef grading is combination manual measurement with sensorial evaluation in china, include measuring beef eye muscle area and back-fat thickness manually, evaluating carcass structure and percentage of fat coverage organolepticly, in addition investigating the grading of beef marbling and physiological maturity degree according to standard of beef grading. Since the method is subjective and qualitative,it is inefficient, inequitable and poor precision in practical applications. Consequently, research of automatic beef grading techniques with machine vision and image processing is of great value and significant to future.In this paper, foundational technique for automatic beef grading has been studied by applying machine vision and image processing. Automatic inspections of grading features of beef carcass is developed, meanwhile beef quality evaluation assessment model specific to Simmental beef grading and beef grading system are established, realizing the purpose of beef grading for Simmental cattle automatically. Works focus on following:1. The scheme of automatic inspection system of beef carcass grading was designed, using an industrial video camera as image gathering device, with the trigger mode to detect beef carcass, at the same time designed trigger circuit to control industrial video camera and lighting system to acquire images. The platform of automatic inspection of grading features was built, which can acquire high-quality image automatically.2. Extraction the grading features such as area of carcass region, carcass length, carcass width and percentage of fat coverage from cattle carcass image by image processing was carried out. Firstly, light compensation and image enhancement were used as pretreatment work. Then removed background with Background Color Model and segmented beef carcass region by edge detection and contour tracking processing precisely. Finally, the central point of beef carcass region was extracted; and based that, extraction grading features of carcass length, carcass width and area of carcass region were carried out automatically, and the extraction for grading features of fat coverage percentage by image segmentation methods of iteration threshold can process. Compare grading features obtained from image processing with manual measurement, we can acquire the result that image processing can extract grading features precisely, meanwhile the time for that only 1.6 second.3. Factors of carcass weight, length, width and thickness of back fat were chosen applying Stepwise regress method, which were used to set up the prediction model for Simmental cattle. After examination, the accurate rate of the model is 80%.4. Factors of carcass length, width, and region area, percentage of fat coverage and thickness of back fat were also chosen to build a prediction model for Simmental cattle on carcass assigned based on GRNN Neural Network using 0.05 as smooth factor. After examination, accurate rate of the model is 85%.5. Based on the accomplished work, our team analyze the distribution of the cutability of beef carcass(CBP) for Simmental cattle. The grading method of Simmental cattle is established as following: A(excellent): CBP≥86% B(good): 84%≤CBP<86% C(average): 82%≤CBP<84% D(poor): CBP<82% According to cutability of beef carcass prediction model,we can evaluate beef yield grading.6. Based on Neural Network, we choose area and amount of fat particles, together with percentage of fat coverage to build the beef quality grading model. After examination, accurate rate of the model is 87.5%.7. Based on VC++6.0 software development platform, we developed the automatic inspection system of grading features, using application interface library supported by microvision corporation, which consists of three modules: image acquisition, image processing and grading features saving unit. The whole system can control the industrial video camera to acquire image, at the same time extract and save the grading features automatically.8. Based on VC++6.0 software development platform, we also developed the beef grading system using yield and quality grading model of Simmental cattle. This system can administer beef grading features in database, evaluate beef quality and yield grade for Simmental cattle.