基于改进YOLOv5的轻量级铁水罐号车号检测算法  

Lightweight hot metal tank number detection algorithm based on improved YOLOv5

作　　者：张继凯 周亚辉 梁勇 柴轶凡[2] ZHANG Jikai;ZHOU Yahui;LIANG Yong;CHAI Yifan(Information Engineering School,Inner Mongolia University of Science and Technology,Baotou 014010,China;Materials and Metallurgy School,Inner Mongolia University of Science and Technology,Baotou 014010,China)

机构地区：[1]内蒙古科技大学信息工程学院,内蒙古包头014010 [2]内蒙古科技大学材料与冶金学院,内蒙古包头014010

出　　处：《内蒙古科技大学学报》2023年第3期265-270,共6页Journal of Inner Mongolia University of Science and Technology

基　　金：国家自然科学基金资助项目(51904161);内蒙古自治区自然科学基金资助项目(2019BS06005);内蒙古自治区高等学校科学研究资助项目(NJZY20095);内蒙古自治区科技计划资助项目(2019GG138).

摘　　要：为解决车号罐号识别中因环境恶劣、字符较小导致的准确率偏低且实时性较差等问题,提出一种基于改进YOLOv5的轻量级检测方法.首先通过二阶段检测并增加小目标检测层,进一步采用大尺寸图像输入和数据均衡方法,提升模型检测效果;其次在骨干网络的最后一层引入CA坐标注意力,并制作掩码实现感兴趣区域检测,提升复杂场景下的车号字符检测精度.最后,通过采用GhostNet模块替换骨干网络模块,使模型进一步轻量化.实验结果表明:YOLO⁃MGCA模型,相较于基线模型map提高了1.4%,模型精度增加了3%,模型参数量减少了40%.In order to solve the problems of low accuracy and poor real⁃time performance in vehicle number and tank number recogni⁃tion due to bad environment and small characters,this paper proposes a lightweight detection method based on improved YOLOv5.Firstly,two⁃stage detection and small target detection layer were added,and then large image input and data balancing methods were a⁃dopted to improve the model detection effect.Secondly,the CA coordinate attention was introduced in the last layer of the backbone network,and the mask was made to realize the detection of the region of interest,so as to improve the detection accuracy of vehicle number characters in complex scenes.Finally,the model was further lightened by replacing the backbone network module with Ghost⁃Net module.The experimental results show that the map value of the YOLO⁃MGCA model in this paper is improved by 1.4%.In com⁃parison with that of the baseline model the model accuracy is increased by 3%and the amount of model parameters is reduced by 40%.

关 键 词：号码识别 YOLOv5 轻量级目标检测 感兴趣区域 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]