基于YOLO-EfficientNet双阶段网络的高精度实时弹目偏差测量方法

赵玉良; 颜邵俊; 李苗; 孙曼晖; 吕晓永

doi:10.12382/bgxb.2025.0856

您当前的位置：

首页 >

文章列表页 >

基于YOLO-EfficientNet双阶段网络的高精度实时弹目偏差测量方法

更新时间：2026-03-09

- 基于YOLO-EfficientNet双阶段网络的高精度实时弹目偏差测量方法
- High-accuracy Real-time Projectile-target Deviation Measurement via YOLO-EfficientNet Dual-stage Architecture
- 兵工学报 2026年
- 作者机构：
  
  1. 东北大学秦皇岛分校智能感知与光电工程学院,河北,秦皇岛,066004
  2. 陆军工程大学石家庄校区,河北,石家庄,050003
  3. 军事科学院战争研究院,北京,100091
- 作者简介：
- 基金信息：
  
  中央高校基本科研业务费(Z2024-086);河北省高等学校科学研究项目(Z2024-003);河北省“三三三人才工程”资助项目(C20231008)
- DOI：10.12382/bgxb.2025.0856
  中图分类号： TJ765
- 收稿：2025-09-17，
  
  网络首发：2026-03-09，
- 稿件说明：
移动端阅览
赵玉良，颜邵俊，李苗，等. 基于YOLO-EfficientNet双阶段网络的高精度实时弹目偏差测量方法[J/OL]. 兵工学报, 2026(2026-03-09). https://doi.org/10.12382/bgxb.2025.0856.

ZHAO Y L, YAN S J, LI M, et al. High-accuracy real-time projectile-target deviation measurement via yolo-efficientnet dual-stage architecture[J/OL]. Acta Armamentarii, 2026(2026-03-09). https://doi.org/10.12382/bgxb.2025.0856. (in Chinese)
赵玉良，颜邵俊，李苗，等. 基于YOLO-EfficientNet双阶段网络的高精度实时弹目偏差测量方法[J/OL]. 兵工学报, 2026(2026-03-09). https://doi.org/10.12382/bgxb.2025.0856. DOI：

ZHAO Y L, YAN S J, LI M, et al. High-accuracy real-time projectile-target deviation measurement via yolo-efficientnet dual-stage architecture[J/OL]. Acta Armamentarii, 2026(2026-03-09). https://doi.org/10.12382/bgxb.2025.0856. (in Chinese) DOI：

摘要

高精度实时弹目偏差测量是提升现代火力打击系统命中精度与闭环校射能力的关键技术。然而，现有方法普遍存在实时性差、精度低和环境鲁棒性不足等问题。为此，提出一种基于YOLO-EfficientNet的双阶段深度学习框架，实现弹目目标的快速检测与偏差精准回归。在目标检测阶段采用轻量化改进的YOLOv11网络实现弹目区域的实时鲁棒检测；在偏差回归阶段通过嵌入坐标注意力机制的剪枝版EfficientNetV2网络，实现亚像素级偏差回归；设计标准化感兴趣区域生成机制，抑制背景干扰，增强模型在复杂环境下的稳定性。实验结果表明：改进模型在检测任务中的mAP50-95达到82.6%，回归任务均方误差低至0.1683，并且整体网络帧率达107.9 FPS，有效满足了弹目偏差测量任务的实际需求，提供了一个新的弹目偏差测量范式。

Abstract

High-precision real-time projectile-target deviation measurement is a key technology for enhancing the hit accuracy and closed-loop fire correction capability of modern fire strike systems. However

existing methods generally suffer from poor real-time performance

low accuracy

and insufficient environmental robustness. To address these issues

this paper proposes a two-stage deep learning framework based on YOLO-EfficientNetfor rapid detection of projectile-target regions and accurate deviationregression.First

in the target detection stage

a lightweight improved YOLOv11 network is employed to achieve real-time and robust detection of the projectile-targetregion.Second

in the deviation regression stage

a pruned version of the EfficientNetV2 network embedded with a coordinate attention mechanism is utilized to accomplish sub-pixel-level deviationregression.Third

a standardized Region of Interest generation mechanism is designed to suppressbackground interference and enhance model stability in complex environments. Experimental results demonstrate that the improved model achieves an mAP50-95 of 82.6% in the detection task

a mean square error as low as 0.1683 in the regression task

and an overall network frame rate of 107.9 FPS. This effectively meets the practical requirements of projectile-target deviation measurement tasks and provides a new paradigm for projectile-target deviation measurement.

关键词

Keywords

references

崔勇平,邢清华.从俄乌战争看无人机对野战防空的挑战和启示[J].航天电子对抗,2022,38(04):1-3.

CUI Y P, XING Q H. The challenge and enlightenment of UAV to field air defense from the Russia Ukraine war[J]. Aerospace Electronic Countermeasures, 2022,38 (04): 1-3. (in Chinese)

陈浩,屈艺,吴盘龙,等.弹目偏差仿真系统的设计与实现[J].兵器装备工程学报,2020,41(09):154-159.

CHEN H, QU Y, WU P L, et al. Design and implementation of missile target deviation simulation system [J]. Journal of Weapon Equipment Engineering, 2020,41 (09): 154-159. (in Chinese)

胡国欣,周唯,邵翔宇.基于面阵式激光雷达的弹目交会参数解算仿真研究[J].舰船电子工程,2023,43(09):88-94.

HU G X, ZHOU W, SHAO X Y. Simulation study of missile target rendezvous parameters based on array laser radar [J]. Naval Electronic Engineering, 2023,43 (09): 88-94. (in Chinese)

毕月,冯玉光.关于改进导弹脱靶量测量方法的研究[J].电光与控制,2021,28(08):71-76.

BI Y, FENG Y G. Research on improving the measurement method of missile miss distance [J]. Electro Optics and Control, 2021,28 (08): 71-76. (in Chinese)

牛龙飞.基于LFMCW雷达测距的空中目标轨迹测量方法[J].火力与指挥控制,2022,47(06):171-175.

NIU L F. Measurement method of air target trajectory based on LFMCW radar ranging [J]. Fire and Command and Control, 2022,47 (06): 171-175. (in Chinese)

邓桂福，刘海良，高节．基于LFMCW雷达的标量脱靶量测量系统[J]．雷达科学与技术，2018,16(06):671-675.

DENG G F, LIU H L, GAO J. Scalar miss distance measurement system based on LFMCW radar [J]. Radar Science and Technology, 2018,16 (06): 671-675. (in Chinese)

毕月,冯玉光.基于FMCW的导弹脱靶量测量系统设计与仿真[J].兵器装备工程学报,2021,42(04):193-197.

BI Y, FENG Y G. Design and simulation of missile miss distance measurement system based on FMCW [J]. Journal of Weapon Equipment Engineering, 2021,42 (04): 193-197. (in Chinese)

张飞猛,马春茂.对空射击声学靶脱靶量测试系统的精度分析[J].兵工学报,2000,(01):23-26.

ZHANG F M, MA C M. Accuracy analysis of miss distance measurement system for acoustic target in air shooting [J]. Journal of Military Engineering, 2000, (01): 23-26. (in Chinese)

王玉龙,刘艳红,陈均辉,等.超音速弹丸斜入射脱靶量计算方法[J].电子测量技术,2020,43(04):71-76.

WANG Y L, LIU Y H, CHEN J H, et al. Calculation method for miss distance of supersonic projectiles at oblique incidence [J]. Electronic Measurement Technology, 2020,43 (04): 71-76. (in Chinese)

王激扬,白风宇,孙晓峰.基于“脱靶管”原理的防空导弹脱靶量估算方法[J].现代防御技术,2023,51(02):49-54.

WANG J Y, BAI F Y, SUN X F. Estimation method of miss distance of air defense missile based on "miss tube" principle [J]. Modern Defense Technology, 2023,51 (02): 49-54. (in Chinese)

韩先平．靶场光学测量空间脱靶量高精度计算方法[J]．电子测量技术，2019,42(17):60-64.

HAN X P. High precision calculation method for space miss distance of optical measurement in shooting range [J]. Electronic Measurement Technology, 2019,42 (17): 60-64. (in Chinese)

崇元,艾葳,王玉坤.空中脱靶量光学测量精度评估[J].战术导弹技术,2024,(03):41-46.

CHONG Y, AI W, WANG Y H. Evaluation of optical measurement accuracy of air miss distance [J]. Tactical Missile Technology, 2024, (03): 41-46. (in Chinese)

RABADAN J, GUERRA V, GUERRA C, et al. A novel ranging technique based on optical camera communications and time difference of arrival[J]. Applied Sciences, 2019, 9(11): 2382.

张琦,迟明祎,李佩军,等.同测试平台脱靶量测试方法研究[J].火力与指挥控制,2023,48(02):167-172.

ZHANG Q, CHI M Y, LI P J, et al. Research on miss distance test method of the same test platform [J]. Fire and Command and Control, 2023,48 (02): 167-172. (in Chinese)

BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4:optimal speed and accuracy of object detection[J]. arXiv:2004.10934, 2020.

LI C, LI L, JIANG H, et al. YOLOv6:a single-stage object detection framework for industrial applications[J]. arXiv:2209.02976, 2022.

WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7:trainable bag-of-freebies sets new state-of-the-art for realtime object detectors[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023:7464-7475.

WANG C Y, YEH I H, LIAO H Y M. YOLOv9:learning what you want to learn using programmable gradient information[J]. arXiv:2402.13616, 2024.

WANG A, CHEN H, LIU L, et al. YOLOv10:real-time endto-end object detection[J]. arXiv:2405.14458, 2024.

TAN M X, LE Q V. Efficientnetv2:Smaller Models and Faster Training [EB/OL].(2021-04-01).https://arxiv.org/abs/2104.00298.

QIN D, LEICHNER C, DELAKIS M, et al. MobileNetV4-universal models for the mobile ecosystem[J]. arXiv preprint arXiv:2404.10518, 2024.

TAN M, PANG R, LE Q V. EfficientDet: scalable and efficient object detection[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2020: 10781-10790.

孟莉莎,杨贤昭,刘惠康.基于CA-EfficientNetV2的蘑菇图像分类算法研究[J].激光与光电子学进展,2022,59(24):56-63.

MENG L S, YANG X Z, LIU H K. Research on mushroom image classification algorithm based on CA Efficient NetV2 [J]. Progress in Laser and Optoelectronics, 2022,59 (24): 56-63. (in Chinese)

HOU Q B, ZHOU D Q, FENG J S. Coordinate attention for efficient mobile network design[C]∥2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR), June 20-25, 2021, Nashville, TN, USA.New York:IEEE Press, 2021:13708-13717.

LIN T Y, DOLLÁR P, GIRSHICK R B, et al. Feature pyramid networks for object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York, NY, US:IEEE, 2017:936-944.

浏览量

下载量

CNKI被引量

文章被引用时，请邮件提醒。

提交

工具集

关联资源

暂无数据