改进YOLOv5的无人机航拍图像密集小目标检测算法

doi:10.3778/j.issn.1002-8331.2306-0289

摘要/Abstract

摘要： 无人机航拍图像中小目标实例多、尺寸变化剧烈且存在密集遮挡等问题，为解决现有目标检测算法难以检测到航拍图像中的小目标物体，提出了一种针对密集小目标的RDS-YOLOv5检测算法。在YOLOv5的三个检测层上新增一个小目标检测层，以保留更丰富的特征信息，增强网络对小目标特征的提取能力，并改善误检漏检情况；为了提高网络的多尺度特征表征能力以及抑制冲突的产生，设计了具有等级制的残差结构的多尺度特征提取模块C3Res2Block；使用解耦检测头Decoupled Head避免不同任务之间的差异所带来的预测偏差，提升了模型的定位精度和检测精度；采用软化非极大值抑制Soft NMS算法对候选框的置信度进行优化，提高模型对密集小目标的检测精度。通过VisDrone数据集的实验结果表明，与基准模型YOLOv5相比，RDS-YOLOv5在mAP0.5上提升了12.9个百分点，mAP0.5：0.95上提升了10.6个百分点，与目前主流的目标检测算法相比也取得更优的检测精度，能够有效完成无人机航拍图像的密集小目标检测任务。

关键词: 小目标检测层, 残差结构, 解耦, 软化非极大值抑制, YOLOv5

Abstract: UAV aerial images have many instances of small objects, drastic size changes and dense occlusions, etc. To solve the difficulty of existing object detection algorithms to detect small objects in aerial images, an RDS-YOLOv5 detection algorithm for dense small objects is proposed. Adding a new small object detection layer to the three detection layers of YOLOv5 to retain richer feature information, the ability of the network is enhanced to extract small object features and reduce false and miss detection. A multi-scale feature extraction module C3Res2Block with a hierarchical residual structure is designed to improve the multi-scale feature representation capability of network as well as to suppress the generation of conflicts. Decoupled detection head is used to avoid the prediction bias caused by the difference between different tasks, which improves the localization and detection accuracy. The confidence of the anchor box is optimized using the Soft NMS algorithm to improve the detection accuracy of model for dense small objects. The experimental results of VisDrone dataset show that RDS-YOLOv5 improves 12.9 percentage points on mAP0.5 and 10.6 percentage points on mAP0.5：0.95 compared with the baseline model YOLOv5, and achieves better detection accuracy compared with the current mainstream object detection algorithms, which can effectively accomplish the task of dense small object detection in UAV aerial images.

Key words: small object detection layer, residual structure, decoupled, soft non-maximum suppression (NMS) , YOLOv5

陈佳慧, 王晓虹. 改进YOLOv5的无人机航拍图像密集小目标检测算法[J]. 计算机工程与应用, 2024, 60(3): 100-108.

CHEN Jiahui, WANG Xiaohong. Dense Small Object Detection Algorithm Based on Improved YOLOv5 in UAV Aerial Images[J]. Computer Engineering and Applications, 2024, 60(3): 100-108.

参考文献

[1] KISANTAL M, WOJNA Z, MURAWSKI J, et al. Augmentation for small object detection[J]. arXiv:1902.07296, 2019.
[2] 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.
[3] LIU S, QI L, QIN H F, et al. Path aggregation network for instance segmentation[C]//IEEE Conference on Computer Vision and Pattern Recognition, 2018: 8759-8768.
[4] ZHU X, LYU S, WANG X, et al. TPH-YOLOv5: improved YOLOv5 based on transformer prediction head for object detection on drone-captured scenarios[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021: 2778-2788.
[5] WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision (ECCV), 2018: 3-19.
[6] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[J]. arXiv:1706.03762, 2017.
[7] YANG C, HUANG Z, WANG N. QueryDet: cascaded sparse query for accelerating high-resolution small object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022: 13668-13677.
[8] GAO S, CHENG M M, ZHAO K, et al. Res2net: a new multi-scale backbone architecture[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 43(2): 652-662.
[9] BODLA N, SINGH B, CHELLAPPA R, et al. Soft-NMS improving object detection with one line of code[C]//Proceedings of the IEEE International Conference on Computer Vision, 2017: 5562-5570.
[10] DU D, ZHU P, WEN L, et al. VisDrone-DET2019: the vision meets drone object detection in image challenge results[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019: 213-226.
[11] 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 UVA based on improved YOLOv5[J]. Acta Aeronautica et Astronautica Sinica, 2023, 44(7): 251-265.
[12] LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//Proceedings of the IEEE International Conference on Computer Vision, 2017: 2980-2988.
[13] CAI Z, VASCONCELOS N. Cascade R-CNN: delving into high quality object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 6154-6162.
[14] MA S, SONG Y, CHENG N, et al. Structured light detection algorithm based on deep learning[C]//IOP Conference Series: Earth and Environmental Science.[S.l.]: IOP Publishing, 2019: 042050.
[15] 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.
[16] LAW H, DENG J. Cornernet: detecting objects as paired keypoints[C]//Proceedings of the European Conference on Computer Vision (ECCV), 2018: 734-750.
[17] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//IEEE Conference on Computer Vision and Pattern Recognition, 2016.
[18] REDMON J, FARHADI A. Yolov3: an incremental improvement[J]. arXiv:1804.02767, 2018.
[19] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. Yolov4: optimal speed and accuracy of object detection[J]. arXiv:2004.10934, 2020.
[20] LI C, LI L, JIANG H, et al. YOLOv6: a single-stage object detection framework for industrial applications[J]. arXiv:2209.02976, 2022.
[21] WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 7464-7475.
[22] GIRSHICK R. Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision, 2015: 1440-1448.
[23] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot mutilbox detector[C]//European Conference on Computer Vision (ECCV). Amsterdam: Springer, 2016: 21-37.
[24] GE Z, LIU S, WANG F, et al. Yolox: exceeding yolo series in 2021[J]. arXiv:2107.08430, 2021.
[25] LIM J S, ASTRID M, YOON H J, et al. Small object detection using context and attention[C]//2021 International Conference on Artificial Intelligence in Information and Communication (ICAIIC), 2021: 181-186.
[26] SHI Z. Object detection algorithms: a comparison[C]//2022 IEEE 4th International Conference on Civil Aviation Safety and Information Technology (ICCASIT), 2022: 861-865.
[27] 徐光达, 毛国君. 多层级特征融合的无人机航拍图像目标检测[J]. 计算机科学与探索, 2023, 17(3): 635-645.
XU G D, MAO G J. Aerial image object detection of UAV based on multi-level feature fusion[J]. Journal of Frontiers of Computer Science and Technology, 2023, 17(3): 635-645.
[28] 谢椿辉, 吴金明, 徐怀宇. 改进YOLOv5的无人机影像小目标检测算法[J]. 计算机工程应用, 2023, 59(9): 198-206.
XIE C H, WU J M, XU H Y. Small object detection algorithm based on improved YOLOv5 in UAV image[J]. Computer Engineering and Applications, 2023, 59(9): 198-206.