面向空基侦察平台的轻量化多通道军事小目标检测方法

赵阳; 王宪楠; 郭策安; 张健

doi:10.12382/bgxb.2025.0938

您当前的位置：

首页 >

文章列表页 >

面向空基侦察平台的轻量化多通道军事小目标检测方法

更新时间：2026-05-07

- 面向空基侦察平台的轻量化多通道军事小目标检测方法
- A Multi-Channel Lightweight Small Object Detection Algorithm for Low-Altitude Reconnaissance Scenarios
- 兵工学报 2026年
- 作者机构：
  
  沈阳理工大学装备工程学院,辽宁,沈阳,110159
- 作者简介：
- 基金信息：
- DOI：10.12382/bgxb.2025.0938
  中图分类号： TP391.41
- 收稿：2025-10-21，
  
  网络首发：2026-05-07，
- 稿件说明：
移动端阅览
赵阳，王宪楠，郭策安，等. 面向空基侦察平台的轻量化多通道军事小目标检测方法[J/OL]. 兵工学报, 2026(2026-05-07). https://doi.org/10.12382/bgxb.2025.0938.

ZHAO Y, WANG X N, ZHANG J. A multi-channel lightweight small object detection algorithm for low-altitude reconnaissance scenarios[J/OL]. Acta Armamentarii, 2026(2026-05-07). https://doi.org/10.12382/bgxb.2025.0938. (in Chinese)
赵阳，王宪楠，郭策安，等. 面向空基侦察平台的轻量化多通道军事小目标检测方法[J/OL]. 兵工学报, 2026(2026-05-07). https://doi.org/10.12382/bgxb.2025.0938. DOI：

ZHAO Y, WANG X N, ZHANG J. A multi-channel lightweight small object detection algorithm for low-altitude reconnaissance scenarios[J/OL]. Acta Armamentarii, 2026(2026-05-07). https://doi.org/10.12382/bgxb.2025.0938. (in Chinese) DOI：

摘要

空基侦察平台在复杂战场环境下军事目标通常具有尺寸小、分布密集、背景复杂及尺度变化显著等特征，加之空基与无人机平台受算力与载荷约束，现有检测方法难以同时满足精度与实时性要求。为此，提出一种面向空基与遥感场景的多通道轻量化小目标检测方法。通过改进网络主干结构，引入灰度辅助通道构建彩色-灰度(Red Green Blue Gray

RGB-Gray)四通道输入，以增强对小目标纹理与边缘细节的特征提取能力；在特征融合阶段启用高分辨率的P2检测层，并裁剪低分辨率的P5层，以在保证精度的同时降低模型复杂度；设计基于门控的自适应轻量双分支C3k2模块(Gated-based Adaptive Lightweight Dual-branch Cross Stage Partial Bottleneck with 2 Convolutions

Conditional-C3k2)模块用于替代传统特征提取单元，引入门控机制动态融合主分支与捷径路径，实现特征信息的自适应分配；进一步采用轻量化跨阶段聚合结构以提升多尺度特征整合效率；此外，改进动态检测头，并提出面积加权损失(Varifocal-SIoU-Area-weighted Loss

VSA-Loss)，以优化边界框回归与分类性能。实验结果表明：所提GCSD-YOLOv11n算法在VisDrone2019与军用车辆细粒度检测遥感图像数据集（Military Vehicle Remote Sensing Dataset

MVRSD）上平均精度均值较基准模型分别提升了13.0%与7.4%，其中MVRSD全类精度达到86.4%，同时模型参数量大幅压缩29.4%；该算法在MAR20军事飞机数据集上亦展现出较强的泛化能力，验证了其对复杂军事成像场景下细粒度目标的探测鲁棒性。

Abstract

Military targets in reconnaissance and remote sensing are characterized by small sizes

dense distributions

complex backgrounds

and significant scale variations

while limited computing resources ofunmanned aerial vehicle(UAV)platforms hinder achieving both high accuracy and real-time performance. This paper proposes a multi-channel lightweight detection algorithm for such scenarios. A grayscale auxiliary channel is introduced to construct ared green blue gray(RGB-Gray)four-channel input

enhancing fine-grained texture and edge feature extraction. High-resolution P2 is enabled and low-resolution P5 is pruned to reduce model complexity. Agated-based adaptive lightweight dual-branch cross stage partial bottleneck with 2 convolutions(Conditional-C3k2)module with dynamic gating is designed for adaptive feature fusion

and alightweight cross-stage aggregation structure is adopted to improve multi-scale integration efficiency. Anenhanced dynamic headand avarifocal-SIoU-area-weighted loss(VSA-Loss)function with an area-weighted mechanism optimize performance. GCSD-YOLOv11n improves Mean Average Precisionby 13.0% and 7.4% on VisDrone2019 andmilitary vehicle remote sensing dataset (MVRSD)

respectively

with MVRSD accuracy reaching 86.4% and parameters reducing by 29.4%. Strong generalization on the MAR20 dataset verifies robustness in complex military imaging

providing an effective solution for real-time airborne reconnaissance tasks.

关键词

Keywords

references

张永梅,赖裕平,马健喆,等.基于视频的装甲车和飞机检测跟踪及轨迹预测算法[J].兵工学报,2021,42(3):545-554.

ZHANG Y M, LAI Y P, MA J Z, et al. Detection, tracking, and trajectory prediction algorithm for armored vehicles and aircraft based on video[J]. Acta Armamentarii, 2021, 42(3): 545–554. (in Chinese)

赵海丽,包大泱,张从豪,等.基于改进SDU-YOLOv8的军事飞机目标检测算法[J].兵工学报, 2026, 47(1):250294.

ZHAO H L, BAO D Y, ZHANG C H, et al. Military aircraft target detection algorithm based on improved SDU-YOLOv8[J]. Acta Armamentarii, 2026, 47(1):250294. (in Chinese)

刘克顺,左晓桐,张玉华,等.基于深度学习的无人机可见光目标检测研究综述[J].航空兵器,2024,31(6):23-35.

LIU K S, ZUO X T, ZHANG Y H, et al. Research review on UAV visible-light target detection based on deep learning[J]. Aeronautical Armament, 2024, 31(6): 23–35. (in Chinese)

DAI X Y, CHEN Y P, XIAO B, et al. Dynamic head: unifying object detection heads with attentions[C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville, TN, US:IEEE, 2021: 7369–7378.

XIONG Y W, LI Z Q, CHEN Y T, et al. Efficient deformable ConvNets: rethinking dynamic and sparse operator for vision applications[C]//Proceedings of the 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle, WA, US:IEEE, 2024:5652-5661.

ZHANG H Y, WANG Y, DAYOUB F, et al. VarifocalNet: an IoU-aware dense object detector[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville, TN, US:IEEE, 2021: 8510-8519.

GEVORGYAN Z. SIoU loss: more powerful learning for bounding box regression: arXiv:2205.12740[R/OL]. Ithaca, NY, US:Cornell University, 2022(2022-05-25). https://arxiv.org/abs/2205.12740.

ZHU P F, WEN L Y, DU D W, et al. Detection and tracking meet drones challenge[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 44(11): 7380–7399.

柏栋,于英,宋亮,等.面向军用车辆细粒度检测的遥感图像数据集构建与验证[J].中国图象图形学报,2024,29(12):3564-3577.

BAI D, YU Y, SONG L, et al. Construction and validation of remote sensing image dataset for fine-grained detection of military vehicles[J]. Journal of Image and Graphics, 2024, 29(12): 3564-3577. (in Chinese)

禹文奇,程塨,王美君,等.MAR20：遥感图像军用飞机目标识别数据集[J].遥感学报,2023,27(12):2688-2696.

YU W Q, CHENG G, WANG M J, et al. MAR20: a benchmark for military aircraft recognition in remote sensing images[J]. Journal of Remote Sensing, 2023, 27(12): 2688–2696. (in Chinese)

杨俭, 陈一剑, 耿樊骏喆. 乌克兰危机无人机作战运用及启示[C]// 第十三届中国指挥与控制大会论文集. 北京:兵器工业出版社, 2025: 205-208.

YANG J, CHEN Y J, GENG F J Z. Operational employment of unmanned aerial vehicles in the Ukraine crisis and its implications[C]// Proceedings of the 13th China Conference on Command and Control. Beijing: The Publishing House of Ordnance Industry,2025: 205–208. (in Chinese)

LOH Y P, CHAN C S. Getting to know low-light images with the Exclusively Dark dataset[J]. Computer Vision and Image Understanding, 2019, 178: 30-42.

TIAN Y J, Ye Q X, DOERMANN D. YOLOv12: Attention-centric real-time object detectors[C]//Advances in Neural Information Processing Systems (NeurIPS). La Jolla: Neural Information Processing Systems Foundation, 2025.

TANG Y H, HAN K, GUO J Y, et al. GhostNetV2: enhance cheap operation with long-range attention[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(8): 10144-10157.

MA N N, ZHANG X Y, ZHENG H T, et al. ShuffleNet V2: practical guidelines for efficient CNN architecture design[C]// Computer Vision – ECCV 2018. Cham, Switzerland:Springer, 2018, 11218: 122-138.

QIN D F, LEICHNER C, DELAKIS M, et al. MobileNetV4: Universal models for the mobile ecosystem[C]// Computer Vision – ECCV 2024. Cham, Switzerland:Springer, 2024, 15098: 78-96.

WANG C C, HE W, NIE Y, et al. Gold-YOLO: efficient object detection via gather-and-distribute mechanism: arXiv:2309.11331 [J/OL].Ithaca, NY, US:Cornell University, , 2023. https://arxiv.org/abs/2309.11331.

冯迎宾,郭枭尊,晏佳华.基于多尺度注意力机制的无人机小目标检测算法[J].兵工学报,2025,46(1):231124.

FENG Y B, GUO X Z, YAN J H. UAV small object detection algorithm based on multi-scale attention mechanism[J]. Acta Armamentarii, 2025, 46(1): 231124. (in Chinese)

ZHAO Y, LV W Y, XU S L, et al. DETRs beat YOLOs on real-time object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway: IEEE, 2024: 22781-22790.

ZHANG H X, LIU K, GAN Z X, et al. UAV-DETR: Efficient end-to-end object detection for unmanned aerial vehicle imagery[J/OL]. arXiv:2501.01855, 2025.

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

基于改进DeepLabV3+的海上船舶目标分割方法