融合平移依赖性与双维度的小目标检测网络

doi:10.3778/j.issn.1002-8331.2403-0374

摘要/Abstract

摘要： 针对目标检测任务中小目标尺寸小、密集度高和排布复杂，易出现漏检错检的问题，提出一种融合平移依赖性与双维度的混合注意力机制算法。旨在保证模型轻量化同时提高小目标特征提取的完整性，以及对特征处理的针对性。设计位置信息自适应模块CIM（coordinate information module），为特征图添加坐标编码，完善目标的位置信息。设计了自适应批归一化模块BNM（batch normalization attention module）及优化-激励模块OEM（optimization-excitation module），对输入数据内部以及彼此之间的特征进行更具针对性的权重分配，帮助模型在处理复杂任务时集中注意力于最相关的信息。最后设计了以一条主分支两条次分支的结构，将三个模块结合成为融合平移依赖性与双维度的混合注意力算法TDHA（translation and dual-dimensional hybrid attention）。该算法可以兼容任意目标检测网络，具备通用性。为验证算法性能，在无人机数据集VisDrone2021上进行了实验，实验结果表明，加入该算法后较基线模型mAP0.5精确度提高1.6个百分点，mAP0.5：0.95提高1个百分点；较YOLOv8精确度mAP0.5提高了1.4个百分点。

关键词: 小目标检测, 平移依赖性, 注意力机制, 轻量化

Abstract: In response to the problem of missed detection and wrong detection in small object detection tasks due to their small size, high density, and complex arrangement, this paper proposes a translation and dual-dimensional hybrid attention. This algorithm aims to ensure lightweight models while improving the integrity of small object feature extraction and the pertinence of feature processing. It designs a coordinate information module (CIM) to add coordinate encoding to the feature map, improving the accuracy of object localization. Additionally, it designs a batch normalization attention module (BNM) and an optimization-excitation module (OEM) to allocate more targeted weights to the internal and interrelated features of the input data, helping the model focus on the most relevant information when handling complex tasks. Finally, it combines these three modules into a hybrid attention algorithm called translation and dual-dimensional hybrid attention (TDHA), which consists of a main branch and two sub-branches. TDHA can be compatible with any object detection network, making it highly versatile. To validate the algorithm’s performance, it conducts experiments on the VisDrone2021 dataset, which consists of drone imagery. The experimental results show that the inclusion of this algorithm improves the mAP0.5 accuracy of the baseline model by 1.6 percentage points and the mAP0.5：0.95 by 1 percentage points. Moreover, it achieves a 1.4 percentage points improvement in mAP0.5 precision compared to YOLOv8.

Key words: small object detection, translation dependency, attention mechanism, lightweight

于龙昆, 王子昊, 沈红, 占强波. 融合平移依赖性与双维度的小目标检测网络[J]. 计算机工程与应用, 2025, 61(13): 300-308.

YU Longkun, WANG Zihao, SHEN Hong, ZHAN Qiangbo. Fusion of Translation Dependency and Dual-Dimensional Small Object Detection Network[J]. Computer Engineering and Applications, 2025, 61(13): 300-308.

参考文献

[1] 潘玮, 韦超, 钱春雨, 等. 面向无人机视角下小目标检测的YOLOv8s改进模型[J]. 计算机工程与应用, 2024, 60(9): 142-150.
PAN W, WEI C, QIAN C Y, et al. Improved YOLOv8s model for small object detection from perspective of drones[J]. Computer Engineering and Applications, 2024, 60(9): 142-150.
[2] 赵侃, 汪慧兰, 郭娇娇, 等. 基于DTA-FSAF的无人机小目标检测研究[J]. 计算机技术与发展, 2024, 34(4): 101-108.
ZHAO K, WANG H L, GUO J J, et al. Research on small object detection of UAV based on DTA-FSAF[J]. Computer Technology and Development, 2024, 34(4): 101-108.
[3] 肖进胜, 赵陶, 周剑, 等. 基于上下文增强和特征提纯的小目标检测网络[J]. 计算机研究与发展, 2023, 60(2): 465-474.
XIAO J S, ZHAO T, ZHOU J, et al. Small target detection network based on context augmentation and feature refinement[J]. Journal of Computer Research and Development, 2023, 60(2): 465-474.
[4] 赵亮, 刘世鹏. 全局与局部图像特征自适应融合的小目标检测算法[J]. 控制与决策, 2023, 38(4): 935-943.
ZHAO L, LIU S P. Small object detection algorithm based on adaptive fusion of global and local image features[J]. Control and Decision, 2023, 38(4): 935-943.
[5] 韩俊, 袁小平, 王准, 等. 基于YOLOv5s的无人机密集小目标检测算法[J]. 浙江大学学报(工学版), 2023, 57(6): 1224-1233.
HAN J, YUAN X P, WANG Z, et al. UAV dense small target detection algorithm based on YOLOv5s[J]. Journal of Zhe- jiang University (Engineering Science), 2023, 57(6): 1224-1233.
[6] GIRSHICK R. Fast R-CNN[C]//Proceedings of the 2015 IEEE International Conference on Computer Vision. Piscataway: IEEE, 2015: 1440-1448.
[7] REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.
[8] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2016: 779-788.
[9] LI C, LI L, JIANG H, et al. YOLOv6: a single-stage object detection framework for industrial applications[J]. arXiv:2209. 02976, 2022.
[10] REDMON J, FARHADI A. YOLO9000: better, faster, stronger[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 6517-6525.
[11] LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//Proceedings of the 2017 IEEE International Conference on Computer Vision. Piscataway: IEEE, 2017: 2999-3007.
[12] REDMON J, FARHADI A. YOLOv3: an incremental improvement[J]. arXiv:1804.02767, 2018.
[13] LIN T Y, DOLLáR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 2117-2125.
[14] BOCHKOVSKIY A, WANG C Y, LIAO H M. YOLOv4: optimal speed and accuracy of object detection[J]. arXiv:2004. 10934, 2020.
[15] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 7132-7141.
[16] WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision, 2018: 3-19.
[17] CAO Y, XU J R, LIN S, et al. GCNet: non-local networks meet squeeze-excitation networks and beyond[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision Workshop. Piscataway: IEEE, 2019: 1971-1980.
[18] WANG X L, GIRSHICK R, GUPTA A, et al. Non-local neural networks[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 7794-7803.
[19] CHEN Y P, KALANTIDIS Y, LI J S, et al. A2-nets: double attention networks[J]. arXiv:1810.11579, 2018.
[20] LIU J J, HOU Q B, CHENG M M, et al. Improving convolutional networks with self-calibrated convolutions[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 10093-10102.
[21] GAO Z L, XIE J T, WANG Q L, et al. Global second-order pooling convolutional networks[C]//Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 3019-3028.
[22] HUANG Z L, WANG X G, HUANG L C, et al. CCNet: criss-cross attention for semantic segmentation[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2019: 603-612.
[23] HOU Q B, ZHOU D Q, FENG J S. Coordinate attention for efficient mobile network design[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 13708-13717.
[24] LIU R, LEHMAN J, MOLINO P, et al. An intriguing failing of convolutional neural networks and the CoordConv solution[J]. arXiv:1807.03247, 2018.
[25] LIU Y C, SHAO Z R, TENG Y Y, et al. NAM: normalization-based attention module[J]. arXiv:2111.12419, 2021.
[26] IOFFE S, SZEGEDY C. Batch normalization: accelerating deep network training by reducing internal covariate shift[C]//Proceedings of the International Conference on Machine Learning, 2015: 448-456.
[27] RAMACHANDRAN P, ZOPH B, LE Q V. Searching for activation functions[J]. arXiv:1710.05941, 2017.
[28] BA J L, KIROS J R, HINTON G E. Layer normalization[J]. arXiv:1607.06450, 2016.
[29] 奉志强, 谢志军, 包正伟, 等. 基于改进YOLOv5的无人机实时密集小目标检测算法[J]. 航空学报, 2023, 44(7): 251-265.
FENG Z Q, XIE Z J, BAO Z W, et al. Real-time dense small object detection algorithm for UAV based on improved YOLOv5[J]. Acta Aeronautica et Astronautica Sinica, 2023, 44(7): 251-265.
[30] LIU Y C, SHAO Z R, HOFFMANN N. Global attention mechanism: retain information to enhance channel-spatial interactions[J]. arXiv:2112.05561, 2021.
[31] HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2016: 770-778.