Research on method of detecting beef fat content based on maximum entropy segmentation

Ke Xiao, Guan-dong Gao and Li-

Abstract

In this paper, a method of detecting beef fat content based on maximum entropy image segmentation is introduced for the application of estimating beef quality in computer vision field. Beef images were captured by DC and transferred to the computer at first. Then, Otsu algorithm was performed to isolate the pork image background, and R layer in RGB color space was used to isolate accurately. For isolating beef muscle and fat, maximum entropy segmentation method, hue segmentation method in HSV color space, minimal cross-entropy method and Otsu method were tested and compared. The fat content results that estimated by the first two methods were also tested in different light conditions, and compared with the results that tested by Soxhlet extractor method. Then, a detective formula was proposed to estimate the real fat content by the results of maximum entropy. Experimental results show that in natural light condition, maximum entropy method is accurate for isolating fat and then estimating fat content. The method in this paper can detect beef fat content accurately, rapidly and non-destructively, and it laid the foundation for the practical application.