[1]黄诚惕,曾智浩,王飞鹏,等.基于改进YOLOv5n模型的靶面弹孔识别技术方案[J].华侨大学学报(自然科学版),2025,46(5):569-580.[doi:10.11830/ISSN.1000-5013.202508037]
　HUANG Chengti,ZENG Zhihao,WANG Feipeng,et al.Bullet Hole Recognition Technology Scheme Based on Improved YOLOv5n Model[J].Journal of Huaqiao University(Natural Science),2025,46(5):569-580.[doi:10.11830/ISSN.1000-5013.202508037]
点击复制

基于改进YOLOv5n模型的靶面弹孔识别技术方案()
分享到：

《华侨大学学报（自然科学版）》[ISSN:1000-5013/CN:35-1079/N]

卷:
第46卷
期数:
2025年第5期
页码:
569-580
栏目:
出版日期:
2025-09-20

文章信息/Info

Title:
Bullet Hole Recognition Technology Scheme Based on Improved YOLOv5n Model
文章编号:
1000-5013(2025)05-0569-12
作者:
黄诚惕;  曾智浩;  王飞鹏;  朱建清
华侨大学 工学院, 福建 泉州 362021
Author(s):
HUANG Chengti;  ZENG Zhihao;  WANG Feipeng;  ZHU Jianqing
College of Engineering, Huaqiao University, Quanzhou 362021, China
关键词:
透视校正;  YOLOv5n;  靶面标定;  边缘计算;  目标检测
Keywords:
perspective correction;  YOLOv5n;  target surface calibration;  edge computing;  target detection
分类号:
TP391.4
DOI:
10.11830/ISSN.1000-5013.202508037
文献标志码:
A
摘要:
针对现有靶面弹孔环数自动识别与标定技术存在的精度不足、抗干扰能力弱及维护成本高等问题,提出一种基于改进YOLOv5n模型的靶面弹孔识别技术方案。在图像预处理阶段,通过多尺度模板匹配定位靶面区域,结合Laplacian算子筛选清晰图像,并利用形态学处理提取胸靶有效区域;进一步实施了Mosaic数据增强优化、骨干网络优化、Neck结构和检测头优化、注意力机制融合及损失函数优化等改进措施,使改进后的YOLOv5n模型的mAP@0.5提升至97.43%,浮点运算数仅2.2×1012 s-1;通过透视校正矩阵实现弹孔定位,结合环线半径计算完成环值标定。通过构建312张图像数据集,将模型部署于RK3588平台进行测试。结果表明:识别速率为21 帧·s-1,文中方法有效平衡了精度与实时性需求,为靶面弹孔环数自动识别与标定提供了可靠的技术支持。
Abstract:
To address the issues of insufficient accuracy, weak anti-interference capability, and high maintenance cost in existing automatic recognition and calibration technologies for bullet hole rings on target surfaces, a bullet hole recognition scheme based on improved YOLOv5n model is proposed. In the image preprocessing stage, multi-scale template matching is employed to locate the target surface region, the Laplacian operator is used to filter images, and morphological processing is applied to extract the effective area of the chest target. Several improvements are further implemented, including Mosaic data augmentation optimization, backbone network optimization, Neck structure and detection head refinement, attention mechanism integration, and loss function optimization. As a result, the improved YOLOv5n model achieves an mAP@0.5 of 97.43% with only 2.2×1012 s-1 floating point operations. Perspective correction matrices are applied for bullet hole localization, and ring values are calibrated using semi-arcradius calculations. A dataset containing 312 images is constructed,and the model is deployed and tested on the RK3588 platform. The results show that the proposed method achieves a recognition speed of 21 frames per second, effectively balances the requirements of accuracy and real-time performance, and providing reliable technical support for the automatic bullet hole ring recognition and calibration on target surface.

参考文献/References:

[1] 陈广庆,孙建,王吉岱,等.基于激光靶的野外自动报靶系统设计[J].电子技术应用,2015,41(5):32-34.DOI:10.16157/j.issn.0258-7998.2015.05.004.
[2] 谭庆,周爱国,龙红,等.一种激光网靶式椭圆弹孔检测装置设计研究[J].电子技术与软件工程,2021(22):97-99.DOI:10.20109/j.cnki.etse.2021.22.036.
[3] 秦向军,周汉昌.激光光幕自动报靶系统的设计与实现[J].计量与测试技术,2009,36(3):30-31.DOI:10.3969/j.issn.1004-6941.2009.03.015.
[4] 张潮,李范鸣.一种改进的模板匹配方法在弹点识别中的应用[J].红外,2005(12):13-17.DOI:10.3969/j.issn.1672-8785.2005.12.004.
[5] 黄应清,陈晓明,谢志宏,等.一种基于计算机视觉的胸环靶弹孔检测方法[J].兵器装备工程学报,2021,42(11):157-162.DOI:10.11809/bqzbgcxb2021.11.025.
[6] QIANG Ren,ZHENG Yongbin,SUN Peng,et al.A robust and accurate end-to-end template matching method based on the siamese network[J].IEEE Geoscience and Remote Sensing Letters,2021(99):1-5.DOI:10.1109/LGRS.2021.3094046.
[7] HAN Hu,KLARE B F,BONNEN K,et al.Matching composite sketches to face photos: A component-based approach[J].IEEE Transactions on Information Forensics & Security,2013,8(1):191-204.DOI:10.1109/TIFS.2012.2228856.
[8] LEI Fei,WEI Ren,LU Lü.Image matching algorithm based on improved harris[C]//2015 International Conference on Computational Intelligence and Communication Networks.Jabalpur:IEEE Press,2015:378-381.DOI:10.1109/CICN.2015.81.
[9] PARIS S,HASINOFF S W,KAUTZ J.Local Laplacian filters: Edge-aware image processing with aLaplacian pyramid[J].ACM Transactions on Graphics,2011,30(4):68.DOI:10.1145/1276377.1276506.
[10] HAN Kai,WANG Yunhe,TIAN Qi,et al.GhostNet: More features from cheap operations[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Seattle:IEEE Press,2020:1580-1589,DOI:10.1109/CVPR42600.2020.00165.
[11] WANG C Y,LIAO Hongyuan,WU Y H,et al.CSPNet: A new backbone that can enhance learning capability of CNN[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops.Seattle:IEEE Press,2020:390-391.DOI:10.1109/CVPRW50498.2020.00203.
[12] LI Hulin,LI Jun,WEI Hanbing,et al.Slim-neck by GSConv: A lightweight-design for real-time detector architectures[J].Journal of Real-Time Image Processing,2024,21(3):62.DOI:10.1007/s11554-024-01436-6.
[13] ALAEDDINE H,JIHENE M.Deep residual network in network[J].Computational Intelligence and Ne-Uroscience,2021,2021:1-9.DOI:10.1155/2021/6659083.
[14] WANG Qilong,WU Banggu,ZHU Pengfei,et al.ECA-Net: Efficient channel attention for deep convolutional neuralnetworks[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Seattle:IEEE Press,2020:11534-11542.DOI:10.1109/CVPR42600.2020.01155.
[15] DU Shuangjiang,ZHANG Baofu,ZHANG Pin,et al.An improved bounding box regression loss function based on CIOU loss for multi-scale object detection[C]//IEEE 2nd International Conference on Pattern Recognition and Machine Learning.Chengdu:IEEE Press,2021:92-98.DOI:10.1109/PRML52754.2021.9520717.
[16] WANG Jinwang, XU Chang, YANG Wen,et al. A normalized Gaussian Wasserstein distance for tiny object detection[EB/OL].(2022-07-14)[2025-07-28] .https://arxiv.org/pdf/2110.13389.
[17] PANARETOS V M,ZEMEL Y.Statistical aspects of Wasserstein distances[J].Annual Review of Statistics and Its Application,2019,6:405-431.DOI:10.1146/annurev-statistics-030718-104938.
[18] DING Jian,XUE Nan,XIA Guisong,et al.Object detection in aerial images: A large-scale benchmark and challenges[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2021,44(11):7778-7796.DOI:10.1109/TPAMI.2021.3117983.

备注/Memo

备注/Memo:
收稿日期: 2025-08-03
通信作者: 黄诚惕(1980-),男,讲师,博士,主要从事物联网技术、智能数据处理、电池健康状态预测的研究。E-mail:qzhct@hqu.edu.cn。
基金项目: 福建省自然科学基金杰出青年基金资助项目(2022J06023); 福建省科技兴警研究计划项目(2024Y 0064); 福建省泉州市高层次人才创新创业项目(2023C013)https://hdxb.hqu.edu.cn/
更新日期/Last Update: 2025-09-20