关键词: 牛肉大理石花纹标准, 彩色梯度, 局部二值模式, 灰度共生矩阵, 线性回归

Abstract:

【Objective】Image processing has become one of the primary means of automatic detection of beef quality. This paper is a study on how to describe BMS (beef marbling standards) based on image texture features. 【Method】 Based on Japanese, American and Australian BMS grading images, linear regression was used to analyze the internal relationship between texture features and BMS.【Result】Four image texture features including contrast, correlation, energy and consistency, could be extracted after color gradient processing and LBP processing, and can be used to describe three different national BMS. One of the features, energy feature, was not sensitive to the beef grade image difference, so it can be used as the common feature among the three BMS.【Conclusion】The linear regression prediction model, based on LBP texture features, can be used as a reasonable basis of evaluation of BMS.

Key words: BMS, color grads, LBP (Local Binary Patterns), GLCM, linear regression