基于Ohta和RGB颜色空间牛胴体背长肌的分割

Segmentation of longissimus dorsi in beef carcass based on Ohta and RGB color systems

提出了一种在Ohta和RGB颜色空间下牛肉胴体眼肌切面中背长肌的分割方法．在合适的光照条件下用CCD摄像头采集以黑色平板为背景的60幅眼肌切面RGB图像，根据R分量的灰度直方图，利用最大方差自动取阈值法（Ostu）去除黑色背景后用中值滤波平滑图像；将预处理后的图像转换到Ohta颜色空间对I3分量用Ostu分割肌肉像素，填充空洞后提取RGB图像对G分量用Ostu进一步去除背长肌和赘肉组织之间的连接，再进行空洞填充、面积阈值分割、数学形态学腐蚀和膨胀运算即可获得背长肌图像，分割的背长肌与人工勾划的背长肌的匹配度为92．36％～99．57％，处理时间在400ms以内．结果表明，结合使用两种颜色空间可以准确、快速地把背长肌从眼肌切面中分割出来．

The segmentation of longissimus dorsi （l.d.） is the foremost step of detection of beef carcass marble grading based on computer vision. A method was introduced to segment the area of l.d. from ribeye imaging based on Ohta and RGB color systems. 60 ribeye images were taken in the suitable illumination conditions with the CCD camera. Preprocessing of images was obtained after removing the background and median filter. Preprocessed images were transformed from RGB color system to Otha color system; muscle pixels were segmented based on I3 variable. The threshold segmenting of G variable was used for removing intermuscular fat on RGB color system, then the area of l.d. was obtained by using the area threshold segmenting, operation of hole-fill, dilation and erosion. Match between the l.d. muscle traced by an expert grader and that segmented by the program was 92.36%～99.57% and computation time was less than 400 ms. The results indicate that this method can better and more quickly segment l.d. from ribeye imaging than other methods.