摘要
针对于汽车轮胎规格字符识别效率低、准确率低等问题,提出一种改进YOLOv5 网络的轮胎规格字符识别方法.首先,将YOLOv5 中的耦合头改为解耦头,提高网络的泛化能力;其次,提出C3-Faster模块,替换YOLOv5 的Backbone和Head中的部分C3 模块,提高网络的计算速度;最后,选用WIoU损失函数替换YOLOv5 的CIoU损失函数,优化网络.通过对比实验,验证了C3-Faster和WIoU损失函数的有效性,在消融实验中,改进后的网络训练时间减少,mAP提高了3.7%,Preci-sion提升2.1%.实验结果表明,该方法在汽车轮胎规格字符识别的有效性,提高了识别的准确性.
Abstract
A method for improving the recognition of characters indicating specifications for automobile tires is proposed to address the issues of low efficiency and accuracy in character recognition.This method involves improving the YOLOv5 network by replacing the coupled head with a decoupled head to enhance the network's generalization ability.Additionally,a C3-Faster module is proposed to replace some C3 mod-ules in the YOLOv5 backbone and head,thereby increasing the network's computational speed.Finally,the WIoU loss function is selected to replace the CIoU loss function in YOLOv5 to optimize the network.Com-parative experiments have been conducted to verify the effectiveness of the C3-Faster and WIoU loss func-tion.The improved network reduced training time and increased mAP by 3.7%and precision by 2.1%in ablative experiments.The experimental results show that this method is effective in recognizing characters indicating specifications for automobile tires,improving recognition accuracy.
基金项目
辽宁省教育厅2021年度科学研究经费面上项目(LJKZ0535)
辽宁省教育厅2021年度科学研究经费面上项目(LJKZ0526)
2021年度本科教育教学综合改革项目(JGLX2021020)
2021年度本科教育教学综合改革项目(JCLX2021008)
NETL
NSTL
维普
万方数据