Study of the Intelligent Predition Method of Beef Tenderness Based on Image Texture Teatures and the Marble Characteristics
|School||Nanjing Agricultural College|
|Course||Agricultural Electrification and Automation|
|Keywords||beef tenderness texture gray level co-occurrence matrix multiplelinear regression|
The ribeye image which has abundant texture characteristics is an important position to detect beef tenderness shear force. The conventional evaluate methods of beef tenderness are sensory determination method and Mechanical judge method. The sensory determination method whose error is great and efficiency is low, is susceptible to personal experience, psychology and the environment. The sensory determination method that measures the shear force of meat with the help of instrument is time consuming, complex, and destructive to beef. According to China’s present situation of beef tenderness detection methods, this paper researched the method to predict beef tenderness which provided the theoretical ground of intelligent the beef tenderness method via building a model based on image texture features and marble characteristics.This paper studied the influencing factors of beef tenderness, the beef ribeye marble characteristic extracting method, the means of analysis the image textures and the texture feature extracting, and the predicting method of beef tenderness based on the image texture features and the marble characteristic.In the part of the extracting of the beef ribeye marble characteristic, the hardware platform which is made up with the CCD industrial camera, PC, and the dome light source, the beef samples are studied. The problem of reflection of light on ribeye surface is solved by appropriate light compensation. The captured beef image was pretreated via gray process, denoised by the weighted filtering mask, photographic enhancement by the Laplace mask. The background was removed by the boundary tracking method. Segmentation threshold was calculated with the method of iterative. Then the optimal beef marble binary image emerged after the image binaryzation, corrosion, and expansion. A specific rectangular area where the pixels whose gray value is255were traversed to extract the marble area characteristic was selected from this binary image. And then the image edges whose pixels gray value is0would be traversed too were extracted to extract the marble circumference characteristic.The method of the GLCM features extraction were Studied, such as the normalization of the GLCM matrix and its parameters, and the average method of the GLCM features in4different direction, etc.. The extraction of beef GLCM texture features is realized.The prediction method of the beef tenderness based on image textures were studied, the selection of shear force detective device, and the process of admeasurement of shear force were discussed.The correlation between marble characteristics and beef tenderness and the correlation between GLCM texture features and beef tenderness are detailed studied, then the characteristic which got high correlations was elected. The beef tenderness prediction model based on texture features and the prediction model based on image texture features and marble characteristic were built. After verification, the latter’s prediction effect was better. And based on the latter model linear comparison between the prediction and the true share force was made, and the verification results showed that the prediction accuracy is high. At last, striploin part which is adjacent to ribeye but a different part from the ribeye was used to checking the model. The check results indicated that the model’s prediction effect was good.This project was tested in Hansen Hejinlai Beef Group in Anhui Province and in Qinbao Animal Husbandry Development Company. Through the analysis of the test results and the related data, the prediction method of beef tenderness based on image textures and marble characteristic with high accuracy was proved scientific and reasonable, and is suitable for real application and being used in production. The prediction system has the high research and commercial value.