基于卷积神经网络的人体穴位识别研究  

Research on Human Acupuncture Point Recognition Based on Convolutional Neural Network

作　　者：魏雨 马晓阳 高志宇[2] WEI Yu;MA Xiaoyang;GAO Zhiyu(Shangzhen College of Henan University of Chinese Medicine,Zhengzhou 450046,China;School of Information Technology of Henan University of Chinese Medicine,Zhengzhou 450046,China)

机构地区：[1]河南中医药大学尚真书院,河南郑州450046 [2]河南中医药大学信息技术学院,河南郑州450046

出　　处：《中医药信息》2024年第2期39-43,共5页Information on Traditional Chinese Medicine

基　　金：河南中医药大学2022年苗圃工程项目(MP2022-82)。

摘　　要：目的:基于卷积神经网络技术对人体穴位进行识别研究。方法:针对人体穴位识别问题构建FasterRCNN模型,并基于该模型构建实用的微信小程序。结果:综合使用Early Stopping策略和Dropout技术可以有效地避免过拟合。在模型训练过程中,通过设置一个最大迭代次数和一个最小性能提升阈值来触发Early Stopping策略,以提前停止训练。同时,可以在神经网络的各个层中应用Dropout技术,以降低模型的复杂度并增强模型的泛化能力。经过不断的调参训练,穴位模型的测试集map最终达到了92%左右,经过30次的迭代损失函数也达到了收敛该模型充分发挥了卷积神经网络的优势,既保证了识别的准确性,又实现了实时性,经过实验验证具有较高的准确性和稳定性。结论:基于卷积神经网络技术构建的微信小程序,用户可以随时随地获取穴位信息,了解穴位知识,为民众提供便捷的健康服务。Objective:To recognize human acupoints based on convolutional neural network technology.Methods:FasterRCNN model was constructed for the problem of human acupoint recognition,and a practical WeChat applet was built based on the model.Results:The combined use of Early Stopping strategy and Dropout technology could effectively avoid overfitting.In the process of model training,the Early Stopping policy was triggered by setting a maximum number of iterations and a minimum performance improvement threshold to stop the training in advance.At the same time,Dropout technology could be applied in various layers of the neural network to reduce the complexity of the model and enhance the generalization ability of the model.After continuous parameter adjustment training,the test set map of the acupoint model finally reached about 92%,and the loss function reached convergence after 30 iterations.The model gave full play to the advantages of convolutional neural network,which not only ensured the accuracy of recognition,but also realized real-time performance,and proved to have high accuracy and stability through experiments.Conclusion:The WeChat applet based on convolutional neural network technology enables users to obtain acupoint information and understand acupoint knowledge anytime and anywhere,providing convenient health services for the public.

关 键 词：卷积神经网络 FasterRCNN模型 人体穴位识别 微信小程序 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术] R224[医药卫生—中医基础理论]