A method for discrimination of processed ginger based on image color feature and a support vector machine model
Abstract
The discrimination of Chinese herbal medicine uses many appearance properties. Of which, color is one of the most important attributes. However, color descriptions in pharmacopoeia can be vague. It's difficult to unify cognition by artificial and subjective discrimination methods. In this regard, digital image processing and machine learning technology were introduced to help solve these problems. By extracting numerical variables of quantified color from images of processed ginger, experiments demonstrated that the HSV colour space was consistent with the true color and could effectively discriminate three processed ginger. After describing colors in terms of HSV, a support vector machine (SVM) model for discrimination of the processed ginger was constructed. An accuracy rate of 98.0277% was acquired to identify the unknown samples. Therefore, the proposed discrimination method based on the image color feature and SVM model is very able to quantify the color of processed ginger, and to evaluate the quality of appearance characters of Chinese herbal components objectively.