改进YOLOv8n的无人机对地多尺度目标检测算法及实现

doi:10.3778/j.issn.1002-8331.2504-0095

摘要/Abstract

摘要： 针对无人机视角下，检测目标尺寸小、尺度跨度大、复杂场景导致的互相遮挡等检测难点，提出一种基于YOLOv8n的改进目标检测网络。引入空间深度卷积SPD-Conv并对其进行轻量化改进，得到SPDs-Conv，在提升小目标检测性能的同时有效降低模型的参数量；设计C2f_SKAttention模块进行多尺度信息的动态调整，实现更好的信息融合；采用一种全新的动态检测头策略，将各类目标的检测头进行了有机统一，降低由复杂场景产生的遮挡等情况对检测的影响。在VisDrone2019数据集上实验得出，改进后的模型在6.9×106参数下实现了49.7%的平均精度，较基准算法提高了15.1个百分点；并具有41?FPS的推理速度，能够实现实时检测，同时相较于其他算法，该算法能够达到更高的检测精度。改进后的模型在Jetson AGX Orin设备上也达到了49.3%的平均精度，推理速度达到28.7?FPS，验证了其效率和对实时嵌入式边缘平台目标检测任务的适用性。

关键词: 无人机, 目标检测, SPDs-Conv, C2f_SKAttention模块, 动态检测头

Abstract: In view of the challenges in target detection from an unmanned aerial vehicle (UAV) perspective, such as small target sizes, large scale variations, and mutual occlusion in complex scenarios, an improved target detection network based on YOLOv8n is proposed. the spatial depth convolution (SPD-Conv) is introduced and lightweight improvements are made to it, resulting in the SPDs-Conv. This not only enhances the performance of small target detection but also effectively reduces the number of model parameters. The C2f_SKAttention module is designed to dynamically adjust multi- scale information, achieving better information fusion. A novel dynamic detection head strategy is adopted. This strategy organically unifies the detection heads for various types of targets, reducing the impact of occlusion and other issues caused by complex scenarios on detection. Experiments on the VisDrone2019 dataset show that the improved model achieves an average precision of 49.7% with 6.9×106 parameters, which is 15.1?percentage points higher than the baseline algorithm. It also has an inference speed of 41?FPS, enabling real-time detection. Compared with other algorithms, the proposed algorithm can achieve higher detection accuracy. The improved model also achieves an average precision of 49.3% on the Jetson AGX Orin device, with an inference speed of 28.7 FPS, verifying its efficiency and applicability to real-time object detection tasks on embedded edge platforms.

Key words: unmanned aerial vehicle (UAV), object detection, SPDs-Conv, C2f_SKAttention module, dynamic detection head

吴思, 黄丹丹, 刘智, 王惠绩, 田成军. 改进YOLOv8n的无人机对地多尺度目标检测算法及实现[J]. 计算机工程与应用, 2025, 61(21): 105-116.

WU Si, HUANG Dandan, LIU Zhi, WANG Huiji, TIAN Chengjun. Improved UAV Multi-Scale Object Detection Algorithm Based on YOLOv8n for Ground Targets and Implementation[J]. Computer Engineering and Applications, 2025, 61(21): 105-116.

参考文献

[1] SRIVASTAVA S, NARAYAN S, MITTAL S. A survey of deep learning techniques for vehicle detection from UAV images[J]. Journal of Systems Architecture, 2021, 117: 102152.
[2] 侯琛, 董俞伯. 基于深度学习的无人机目标识别与反制[J]. 软件, 2024, 45(5): 161-164.
HOU C, DONG Y B. Deep learning based UAV target recognition and countermeasures[J]. Software, 2024, 45(5): 161-164.
[3] GUAN J X, MA W M. Ultra-deep channel sand body target recognition method based on improved deep learning under UAV cluster[J]. Open Geosciences, 2024, 16: 20220612.
[4] 李佳一, 闫振纲, 闫克丁, 等. 基于注意力与通道重排的无人机对地目标检测算法[J]. 兵器装备工程学报, 2024, 45(3): 306-313.
LI J Y, YAN Z G, YAN K D, et al. Detection algorithm of ground target based on attention and channel rearrangement for UAV[J]. Journal of Ordnance Equipment Engineering, 2024, 45(3): 306-313.
[5] DU X M, ZHANG S W, LIU L. UAV trajectory optimization for sensing exploiting target location distribution map[C]//Proceedings of the 2024 IEEE 99th Vehicular Technology Conference. Piscataway: IEEE, 2024: 1-5.
[6] HUANG Y Q, HUANG H, NIU M B, et al. UAV complex-scene single-target tracking based on improved re-detection staple algorithm[J]. Remote Sensing, 2024, 16(10): 1768.
[7] XU W Y, SUN H J, WANG S. SAT: spectrum-adaptive transformer with spatial awareness for UAV target tracking[J]. Remote Sensing, 2025, 17(1): 52.
[8] HUANG J Y, XIE J W, ZHAI H L, et al. An adaptive constant acceleration model for maneuvering target tracking[J]. Remote Sensing, 2025, 17(5): 850.
[9] GONG Y Q, YU X H, DING Y, et al. Effective fusion factor in FPN for tiny object detection[C]//Proceedings of the 2021 IEEE Winter Conference on Applications of Computer Vision. Piscataway: IEEE, 2021: 1159-1167.
[10] ZHANG Y. Detection and tracking of human motion targets in video images based on camshift algorithms[J]. IEEE Sensors Journal, 2020, 20(20): 11887-11893.
[11] CHEN X L, YU X H, HUANG Y, et al. Adaptive clutter suppression and detection algorithm for radar maneuvering target with high-order motions via sparse fractional ambiguity function[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2020, 13: 1515-1526.
[12] MANTAU A J, WIDAYAT I W, LEU J S, et al. A human-detection method based on YOLOv5 and transfer learning using thermal image data from UAV perspective for surveillance system[J]. Drones, 2022, 6(10): 290.
[13] 郝紫霄, 王琦.基于YOLO-v7的无人机航拍图像小目标检测改进算法[J].软件导刊, 2024, 23(1): 167-172.
HAO Z X, WANG Q. Enhanced algorithm for small target detection in UAV aerial images based on YOLO-v7[J]. Software Guide, 2024, 23(1): 167-172.
[14] 单慧琳, 王硕洋, 童俊毅, 等. 增强小目标特征的多尺度光学遥感图像目标检测[J]. 光学学报, 2024, 44(6): 0628006.
SHAN H L, WANG S Y, TONG J Y, et al. Multi-scale optical remote sensing image target detection based on enhanced small target features[J]. Acta Optica Sinica, 2024, 44(6): 0628006.
[15] 鞠默然, 罗江宁, 王仲博, 等. 融合注意力机制的多尺度目标检测算法[J]. 光学学报, 2020, 40(13): 1315002.
JU M R, LUO J N, WANG Z B, et al. Multi-scale target detection algorithm based on attention mechanism[J]. Acta Optica Sinica, 2020, 40(13): 1315002.
[16] LAN Z Y, ZHUANG F Y, LIN Z J, et al. MFO-net: a multiscale feature optimization network for UAV image object detection[J]. IEEE Geoscience and Remote Sensing Letters, 2024, 21: 6006605.
[17] 吴明杰, 云利军, 陈载清, 等. 改进YOLOv5s的无人机视角下小目标检测算法[J]. 计算机工程与应用, 2024, 60(2): 191-199.
WU M J, YUN L J, CHEN Z Q, et al. Improved YOLOv5s small object detection algorithm in UAV view[J]. Computer Engineering and Applications, 2024, 60(2): 191-199.
[18] TAHIR N U A, LONG Z, ZHANG Z P, et al. PVswin-YOLOv8s: UAV-based pedestrian and vehicle detection for traffic management in smart cities using improved YOLOv8[J]. Drones, 2024, 8(3): 84.
[19] 胡峻峰, 李柏聪, 朱昊, 等. 改进YOLOv8的轻量化无人机目标检测算法[J]. 计算机工程与应用, 2024, 60(8): 182-191.
HU J F, LI B C, ZHU H, et al. Improved YOLOv8 lightweight UAV target detection algorithm[J]. Computer Engineering and Applications, 2024, 60(8): 182-191.
[20] LOU H T, DUAN X H, GUO J M, et al. DC-YOLOv8: small-size object detection algorithm based on camera sensor[J]. Electronics, 2023, 12(10): 2323.
[21] 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, 2022, 44(11): 7380-7399.
[22] CAI Z W, VASCONCELOS N. Cascade R-CNN: delving into high quality object detection[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 6154-6162.
[23] 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.
[24] DUAN K W, BAI S, XIE L X, et al. CenterNet: keypoint triplets for object detection[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2020: 6568-6577.
[25] CAI Z S, HONG Z Y, YU W H, et al. CNXResNet: a light-weight backbone based on PP-YOLOE for drone-captured scenarios[C]//Proceedings of the 2023 8th International Conference on Signal and Image Processing. Piscataway: IEEE, 2023: 460-464.
[26] ZHANG Z X. Drone-YOLO: an efficient neural network method for target detection in drone images[J]. Drones, 2023, 7(8): 526.
[27] WANG G, CHEN Y F, AN P, et al. UAV-YOLOv8: a small-object-detection model based on improved YOLOv8 for UAV aerial photography scenarios[J]. Sensors, 2023, 23(16): 7190.