|
EfficientNet在阴虚证眼象识别中的应用研究 |
Application Research of EfficientNet in Eye Recognition of Yin Deficiency SyndromeTitle |
投稿时间:2019-10-21 录用日期:2019-10-28 |
DOI: |
中文关键词: 卷积神经网络 阴虚证 目诊 图像识别 |
英文关键词: Convolutional neural network Yin deficiency syndrome Eye diagnosis Image recognitio |
基金项目:国家重点研计划“中医药现代化研究”专项 |
|
摘要点击次数: 36 |
全文下载次数: 0 |
中文摘要: |
目的 本文基于EfficientNet建立可靠的阴虚证眼象识别模型,为阴虚证共性诊断模型的研究提供基础,推动中医目诊客观化的研究。方法 构建以阴虚证为主的常见中医证候眼象数据库,通过轮廓检测切割眼象图片,基于retinex理论增强图像,使用反射、对比度变换、噪声扰动等方法扩大数据集作为模型训练材料。对图片进行批归一化预处理后通过卷积神经网络构建眼象特征提取模型以及分类模型,判断最终归属的证候。结果 本文的阴虚证识别模型对验证组数据的识别准确率达90.01%,对于阴虚证的诊断有一定的辅助价值。并以此模型为基础构建了阴虚证健康管理平台,将研究成果投入实际应用同时进一步收集眼象图片数据。 |
英文摘要: |
Objective: To establish a reliable eye image recognition model for yin deficiency syndrome, provide a basis for the study of the common diagnosis model of Yin deficiency syndrome, and promote the objective study of TCM clinic. Methods: A common TCM syndrome eye image database based on Yin deficiency syndrome was constructed, cut the image by contour detection and enhance the image based on retinex theory, and the data set was expanded as a model training material by means of reflection, contrast transformation and noise disturbance. The image is normalized and preprocessed, and the eye feature extraction model is constructed by convolutional neural network. Results: Based on the triplet loss yin deficiency syndrome eye recognition model, the accuracy rate is up to 90.01%, which has great reference value for the diagnosis of yin deficiency syndrome. Based on this model, the health management platform of Yin deficiency syndrome was constructed, and the research results were put into practical application and the eye image data was further collected. |
View Fulltext
查看/发表评论 下载PDF阅读器 |
关闭 |