基于改进YOLOv4的石窟壁画病害检测分析  

Detection and Analysis of Grotto Mural Diseases Based on Improved YOLOv4

作　　者：申晋祥[1] 鲍美英[1] SHEN Jin-xiang;BAO Mei-ying(School of Computer and Network Engineering,Shanxi Datong University,Datong Shanxi,037009)

机构地区：[1]山西大同大学计算机与网络工程学院,山西大同037009

出　　处：《山西大同大学学报（自然科学版）》2023年第2期15-17,22,共4页Journal of Shanxi Datong University(Natural Science Edition)

基　　金：山西大同大学科研专项项目[2020YGZX016];山西大同大学科研项目[2020K10];山西大同大学教学改革创新项目[XJG2021249]。

摘　　要：针对石窟壁画病害检测中存在的漏检、误检导致检测精度不高的问题,提出一种基于改进的YOLOv4目标检测算法。算法在YOLOv4网络结构中引入注意力机制,增强神经网络特征信息的权重,减少无效信息的权重,使网络更注重有效特征信息部分,降低漏检、误检情况。实验结果表明,通过不同算法进行对比,改进的YOLOv4算法对石窟壁画病害检测具有更好效果,能够有效提高检测精度。Aiming at the problem of low detection accuracy caused by missing and false detection in the detection of grotto mural diseases,a target detection algorithm based on improved YOLOv4 is proposed.The algorithm introduces the attention mechanism in YOLOv4 network structure,enhances the weight of neural network feature information,reduces the weight of invalid information,makes the network pay more attention to the effective feature information,and reduces the missing and false detection.The experimental results show that by comparing different algorithms,the improved YOLOv4 algorithm has a better effect on the detection of grotto mural diseases and can effectively improve the detection accuracy.

关 键 词：石窟壁画 图像识别 YOLO算法 注意力机制 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]