基于改进YOLOv5的舌面特征检测  

Tongue Feature Detection Based on Improved YOLOv5 ZHANG Delong,JIN

作　　者：张德龙 金春阳 张志东[1] 曹溪源 薛晨阳[1] Chunyang;ZHANG Zhidong;CAO Xiyuan;XUE Chenyang(Key Laboratory of Instrumentation Science&Dynamic Measurement of Ministry of Education,North University of China,Taiyuan 030051,China)

机构地区：[1]中北大学仪器科学与动态测试教育部重点实验室,太原030051

出　　处：《计算机测量与控制》2025年第4期89-94,146,共7页Computer Measurement &Control

基　　金：山西省重点研发计划(202102130501011);中北大学研究生科技立项(20231940)。

摘　　要：针对传统中医舌诊视觉诊断存在主观性强且耗费精力的问题,提出一种基于改进YOLOv5的舌面齿痕和裂纹特征自动检测模型:该模型在YOLOv5模型的骨干网络中引入SimAM-CSP模块以增强网络的特征提取能力;在瓶颈层和预测部分之间加入瓶颈注意力模块,进一步聚焦关键信息;通过调整YOLOv5的特征融合结构,增加对图像细节的感知能力,提高网络性能;将定位损失函数GIoU替换为EIoU,提升模型的训练收敛速度和预测回归精度;利用K-Means聚类算法对YOLOv5的初始锚框进行调整,使算法更加契合舌面齿痕和裂纹特征检测;将改进后的YOLOv5模型在自制舌象数据集中进行训练,得到的平均检测精度(mAP)为79.5%,较原算法提升了6.3个百分点。实验结果表明改进YOLOv5模型能够有效提高舌面齿痕和裂纹特征检测精度,有助于辅助医生诊断。Visual diagnosis has the shortages of subjectivity and energy-consuming in traditional Chinese medicine tongue diagnosis,a detection model for tongue tooth mark and fissure features based on improved YOLOv5 is proposed.In the backbone of YOLOv5,this model introduces the SimAM-CSP module to enhance the feature extraction capability of the network.The Bottleneck Attention Module is added between Neck layer and Head layer to further focus on critical information.The feature fusion structure of the YOLOv5 is adjusted to enhance the perception ability of image details and improve the performance of the network.The localization loss function GIoU is replaced with the EIoU to improve training convergence speed and predict regression accuracy.The initial anchor frames of the YOLOv5 are adjusted by using the K-Means algorithm to make the model more suitable for tongue tooth mark and fissure detection.The improved YOLOv5 model is trained on the self-built tongue image dataset,the obtained mean average precision(mAP)is 79.5%,which is 6.3%higher than that of the original algorithm.Experimental results show that the improved YOLOv5 model can effectively improve the accuracy of tongue tooth mark and fissure detection,which assists doctors in diagnosis.

关 键 词：YOLOv5 深度学习 目标检测 异常舌 舌诊 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]