Automatic grading of pork color based on self-organizing map neural network

The pork color was graded with Self-Organizing Map(SOM)neural network after processing the pork image.First,images of porcine longissimus dorsi muscle profile were captured by a camera,then the images were filtered with average weights template and the muscle pixels were separated by removing the background and fat pixels,at last the feature vector of the pork color was expressed by H,S,L values.The H value was a new value rather than the hue of the pork color,and the S,L values were saturation and luminance values respectively.By entering the H,S and L values of muscle into SOM neural network,73.2%,80.1%,90.5%,65.8%,54.7%,33.1% grading accuracy were obtained for each grade respectively.These results showed the grading results were credible in some certain extent,the computer intelligent automatic grading system can be used in the processing effectually,but accuracy rate need to be improved further.