文献类型：期刊文献

中文题名：基于二分光反射模型与随机森林的中医舌苔润燥识别研究

英文题名：Research on recognition of tongue-fur in traditional Chinese medicine based on dichroic reflection model and random forest

作者：颜建军[1];曾梦浩[1];郭睿[2];穆伟伟[1];燕海霞[2];周炜[1];王忆勤[2]

机构：[1]华东理工大学机械与动力工程学院,上海200237;[2]上海中医药大学四诊信息综合实验室,上海201203

年份：2022

卷号：37

期号：10

起止页码：5908

中文期刊名：中华中医药杂志

外文期刊名：China Journal of Traditional Chinese Medicine and Pharmacy

收录：CSTPCD;;北大核心:【北大核心2020】；CSCD:【CSCD2021_2022】；

基金：国家自然科学基金面上项目(N o.81673880);上海市健康辨识与评估重点实验室项目(No.21DZ2271000)。

语种：中文

中文关键词：特征分析;二分光反射模型;随机森林;舌苔润燥识别

外文关键词：Feature analysis;Dichroic reflection model;Random forest;Tongue-fur moisturizing recognition

摘要：在以往的中医舌苔润燥识别研究中,只利用单个特征或指标,通过与经验阈值进行比较来进行舌苔润燥的识别。提取的特征不能充分地反映水分亮斑区的特性,同时利用经验阈值作为分类依据,易受人为主观因素的影响。本课题组提出一种基于二分光反射模型与随机森林的舌苔润燥识别方法。在已有研究的基础上,基于二分光反射模型分析亮斑区与本色区的光学特性差异,提取舌图像的水分亮斑区的RGB协方差矩阵特征值、亮度和等特征,并对这些特征进行统计分析;基于随机森林算法建立舌苔润燥识别模型。实验结果表明,该方法分类准确率可达95.9%,与已有的舌苔润燥识别方法相比有所提升。该研究为舌苔润燥识别提供了一种新的思路和方法,对舌诊客观化具有一定的实用价值。
In the past research on the identification of tongue-fur moisturizing in traditional Chinese medicine, only a single feature or index was used to identify tongue-fur moisturizing and dryness by comparing with empirical threshold. The extracted features cannot fully reflect the characteristics of the bright spots of water, and at the same time, the empirical threshold is used as the basis for classification, which is susceptible to subjective factors. This paper proposes a method for identifying tongue coating moisturization based on the bipartite light reflection model and random forest. Based on the existing research,based on the dichroic reflection model to analyze the optical characteristic difference between the bright spot area and the natural color area, extract the RGB covariance matrix eigenvalues, brightness and other characteristics of the moisture bright spot area of the tongue image, and statistical analysis is performed on the features;the tongue fur moisturizing recognition model is established based on the random forest algorithm. The experimental results show that the classification accuracy of this method is improved to 95.9% compared with the existing tongue fur moisturizing recognition method. This research provides a new idea and method for the identification of tongue fur moisturization and has certain practical value for the objectification of tongue diagnosis.