基于AdaBoost级联框架的舌色分类

Tongue color classification based on AdaBoost cascade framework

目的基于图像处理的舌质颜色分析是中医舌诊现代化的重要内容,提高舌色的正确识别率是其中的关键问题。本文使用集成学习的分类方法来探讨舌色分类,以达到客观、准确地识别中医(traditional Chinese medicine,TCM)舌色。方法首先通过AdaBoost算法对舌图像进行初步分类,再将该算法与级联框架进行结合;然后通过"一对其余"的方法将AdaBoost从二分类扩展到多类来完成舌质颜色的分析;最后通过实验进行验证,并与其他方法所得出的结果进行对比。结果针对各类舌质颜色分类问题,使用随机森林与传统的AdaBoost分类器进行分类的正确率分别在78.0%~90.2%与89.4%~95.5%之间,而基于AdaBoost级联框架的分类器的各类舌质分类正确率在93.0%~98.7%之间。结论基于AdaBoost级联框架的舌质颜色分类方法与其他经典方法相比,具有较高的正确分类率,为基于图像处理的中医舌诊辅助诊断奠定了一定的基础。

Objective The color analysis of tongue based on image processing is an important part of the automatic analysis of tongue images in Chinese medicine.It is the key to improve the correct recognition rate of tongue color.This paper uses the classification method of integrated learning to explore the problem of tongue color classification to achieve objective and accurate identification of traditional Chinese medicine(TCM)tongue color.MethodsFirstly,the tongue image is preliminarily classified by AdaBoost algorithm,and then the algorithm is combined with the cascade framework.AdaBoost can be extended from two categories to multiple classes by a"one pair of rest"approach.In this way,the analysis of the color of the tongue can be achieved.Finally,the experiment is carried out and the results obtained by other methods are compared.ResultsFor various tongue color classification problems,the accuracy of classification using random forest and traditional AdaBoost classifiers is in the range of 78.0%~90.2%and 89.4%~95.5%,respectively,and the types of classifiers based on the AdaBoost cascade framework the accuracy of tongue classification is between 93.0%~98.7%.Conclusions Compared with other classical methods,the tongue color classification method based on AdaBoost cascading framework has a higher correct classification rate,which lays a foundation for thediagnosis of TCM tongue diagnosis based on image processing.