无人机航拍视角下密集场景非机动车小目标检测方法

doi:10.16097/j.cnki.1009-6744.2025.04.015

交通运输系统工程与信息 ›› 2025, Vol. 25 ›› Issue (4): 147-161.DOI: 10.16097/j.cnki.1009-6744.2025.04.015

• 智能交通系统与信息技术 • 上一篇下一篇

无人机航拍视角下密集场景非机动车小目标检测方法

郑展骥¹，冯昌奎¹，赵杨洋²，凃强^1,3，张河山^*1，徐进¹

1. 重庆交通大学，交通运输学院，重庆400074；2.公安部交通管理科学研究所，江苏无锡214151； 3. 重庆市城投金卡信息产业(集团)股份有限公司，博士后科研工作站，重庆400074

收稿日期:2025-01-28 修回日期:2025-05-18 接受日期:2025-06-10 出版日期:2025-08-25 发布日期:2025-08-25
作者简介:郑展骥(1990—)，男，浙江瑞安人，副教授，博士。
基金资助:
中国博士后科学基金面上项目(2022M710547)；教育部人文社会科学研究青年基金项目 (24YJCZH412)。

Non-motorized Small Target Detection Method for Dense Scenes Under UAV Aerial Photography Perspective

ZHENG Zhanji¹, FENG Changkui¹, ZHAO Yangyang², TU Qiang^1,3, ZHANG Heshan^*1, XU Jin¹

1. School of Traffic and Transportation, Chongqing Jiaotong University, Chongqing 400074, China; 2. Traffic Management Research Institute of the Ministry of Public Security, Wuxi 214151, Jiangsu, China; 3. Postdoctoral Research Station, Chongqing City Investment & Golden Card Information Industry (Group) Co Ltd, Chongqing 400074, China

Received:2025-01-28 Revised:2025-05-18 Accepted:2025-06-10 Online:2025-08-25 Published:2025-08-25
Supported by:
China Postdoctoral Science Foundation(2022M710547)；Youth Fund Project for Humanities and Social Sciences Research of China's Ministry of Education (24YJCZH412)。

摘要/Abstract

摘要： 针对无人机航拍图像中背景丰富，多个小目标聚集导致的误检、漏检，以及置信度不高等问题，本文提出一种改进YOLOX(YouOnlyLookOnceX)的目标检测算法。首先，设计一种提取高频特征信息的注意力机制(LE-MSA)，避免小目标特征消失在冗余信息中；其次，为防止样本类别不均衡导致检测效果差的问题，引入VarifocalLoss损失函数，与BCEWithLogitsLoss损失函数共同参与分类准确度和目标框定位精度的提升；最后，提出一种包含自适应小目标增强和区域增强方法在内的多策略数据增强方法，提高模型的泛化能力。试验结果表明，LE-YOLOX算法表现出良好的检测能力，其检测精度达到90.78%，优于Faster R-CNN(71.30%)、YOLOv5(88.15%)、YOLOv8(87.63%)、YOLOv10(86.1%)和YOLOX(87.82%)；同时，改进YOLOX在实际检测时能够有效解决无人机航拍图像下密集小目标的误检和漏检问题，具有较强的小目标识别和密集目标处理能力。

关键词: 智能交通, 密集小目标检测, YOLOX, 无人机航拍图像, 注意力机制

Abstract: Aiming at the problems of misdetection, omission and low confidence caused by the rich background and the aggregation of multiple small targets in UAV aerial images, this paper proposes a target detection algorithm to improve YOLOX. First, an attention mechanism (LE-MSA) is designed to avoid small target features disappearing into redundant information by extracting high-frequency feature information. Second, in order to prevent the problem of poor detection effect caused by the imbalance of sample categories, a VarifocalLoss loss function is introduced to participate in the improvement of classification accuracy and target frame localization accuracy, together with the BCEWithLogitsLoss loss function. Finally, a multi-strategy data enhancement method, including adaptive small target enhancement and region enhancement methods, is proposed to improve the generalization ability of the model. The experimental results show that the LE-YOLOX algorithm exhibits good detection ability with a detection accuracy of 90.78%, which is better than Faster R-CNN (71.30%), YOLOv5 (88.15%), YOLOv8 (87.63%), YOLOv10 (86.1%) and YOLOX (87.82%). Meanwhile, the improved YOLOX is able to solve the problem of misdetection and missed detection of dense small targets under UAV aerial images in actual detection effectively, and it has strong small target recognition and dense target processing capabilities.

Key words: intelligent transportation, dense small target detection, YOLOX, UAV aerial images, attention mechanism

中图分类号:

U491

郑展骥, 冯昌奎, 赵杨洋, 凃强, 张河山, 徐进. 无人机航拍视角下密集场景非机动车小目标检测方法[J]. 交通运输系统工程与信息, 2025, 25(4): 147-161.

ZHENG Zhanji, FENG Changkui, ZHAO Yangyang, TU Qiang, ZHANG Heshan, XU Jin. Non-motorized Small Target Detection Method for Dense Scenes Under UAV Aerial Photography Perspective[J]. Journal of Transportation Systems Engineering and Information Technology, 2025, 25(4): 147-161.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: http://www.tseit.org.cn/CN/10.16097/j.cnki.1009-6744.2025.04.015

http://www.tseit.org.cn/CN/Y2025/V25/I4/147

参考文献

[1] FASTER R. Towards real-time object detection with region proposal networks[J]. Advances in Neural Information Processing Systems, 2015(28): 2969239 2969250.

[2] REDMON J. You only look once: Unified, real-time object detection[C]. Las Vegas: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016.

[3] TERVEN J, CORDOVA-ESPARZA D M, ROMERO GONZALEZ J A. A comprehensive review of YOLO architectures in computer vision: From YOLOV1 to YOLOV8 and YOLO-NAS[J]. Machine Learning and Knowledge Extraction, 2023, 5(4): 1680-1716.

[4] LONG X, DENG K, WANG G, et al. PP-YOLO: An effective and efficient implementation of object detector [EB/OL]. arXiv: 2007.12099, (2020-07-23) [2024-09 16]. https://arxiv.org/abs/2007.12099

[5] GE Z, LIU S, WANG F, et al. Yolox: Exceeding YOLO series in 2021[EB/OL]. arXiv: 2107.08430, (2021-08 06) [2024-09-16]. https://arxiv.org/abs/2107.08430.

[6]张河山,范梦伟,谭鑫,等.基于改进YOLOX的无人机航拍图像密集小目标车辆检测[J/OL].吉林大学学报(工学版), (2023-12-28) [2024-09-16]. https://doi.org/ 10.13229/j.cnki.jdxbgxb.20230779. [ZHANG H S, FAN MW, TAN X, et al. Dense small object vehicle detection in UAV aerial images using improved YOLOX[J/OL]. Journal of Jilin University(Engineering and Technology Edition), (2023-12-28) [2024-09-16]. https://doi.org/ 10.13229/j.cnki.jdxbgxb.20230779.]

[7] HOU Q, ZHOU D, FENG J. Coordinate attention for efficient mobile network design[C] Piscataway: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021.

[8] SHEN C, MA C, GAO W. Multiple attention mechanism enhanced YOLOX for remote sensing object detection[J]. Sensors, 2023, 23(3): 1261-1261.

[9] 罗为明,李旭,孙正良,等.基于高效卷积注意力特征融合的道路目标检测[J].东南大学学报(自然科学版), 2024, 54(4): 1005-1013. [LUO W M, LI X, SUN Z L, et al. Object detection in road based on efficient convolutional attention feature fusion[J]. Journal of Southeast University(Natural Science Edition), 2024, 54 (4): 1005-1013.]

[10] 沈瑜, 李阳阳,李博昊,等.基于感知增强与多尺度融合的小目标车辆检测[J/OL].北京航空航天大学学报, (2024-07-19) [2024-09-16]. https://doi.org/10.13700/j. bh.1001-5965.2024.0124. [SHEN Y, LI Y Y, LI B H, et al. Small target vehicle detection based on perceptual enhancement and multi scale fusion[J/OL]. Journal of Beijing University of Aeronautics and Astronautics, (2024-07-19) [2024-09-16]. https://doi.org/10.13700/j. bh.1001-5965.2024.0124.]

[11] ZHANG K, ZHANG Z, LI Z, et al. Joint face detection and alignment using multitask cascaded convolutional networks[J]. IEEE Signal Processing Letters, 2016, 23 (10): 1499-1503.

[12] 宋存利,柴伟琴,张雪松.基于改进YOLOv5算法的道路小目标检测[J].系统工程与电子技术,2024,46(10): 3271-3278. [SONG C L, CHAI W Q, ZHANG X S. Road small target detection based on improved YOLO v5 algorithm[J]. Systems Engineering and Electronics, 2024, 46(10): 3271-3278.]

[13] 张河山, 谭鑫,范梦伟,等.无人机高空航拍视角下小尺度车辆精确检测方法[J].交通运输系统工程与信息, 2024, 24(3): 299-309. [ZHANG H S, TAN X, FAN MW, et al. Accurate detection method of small-scale vehicles from perspective of unmanned aerial vehicle High-altitude aerial photography[J]. Journal of Transportation Systems Engineering and Information Technology, 2024, 24(3): 299-309.]

[14] ANDREW W, GREATWOOD C, BURGHARDT T. Aerial animal biometrics: Individual friesian cattle recovery and visual identification via an autonomous UAV with onboard deep inference[C]. Macau: 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), IEEE, 2019.

[15] 艾青林,杨佳豪,崔景瑞.基于自适应增殖数据增强与全局特征融合的小目标行人检测[J].浙江大学学报(工学版), 2023, 57(10): 1933-1944, 1976. [AI Q L, YANG J H, CUI J R. Small target pedestrian detection based on adaptive proliferation data enhancement and global feature fusion[J]. Journal of Zhejiang University (Engineering Science), 2023, 57(10): 1933-1944, 1976.]

[16] KISANTAL M. Augmentation for small object detection [J/OL]. arXiv: 1902.07296, (2019-02-19) [2024-09 16]. https://arxiv.org/abs/1902.07296.

[17] HU J, SHEN L, SUN G. Squeeze-and-excitation networks [C]. Salt Lake City: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, IEEE, 2018.

[18] WOO S, PARK J, LEE J Y, et al. Cbam: Convolutional block attention module[C]. Munich: Proceedings of the European Conference on Computer Vision (ECCV), 2018.

[19] LIU Y, SHAO Z, TENG Y, et al. NAM: Normalization based attention module[J/OL]. arXiv: 2111.12419, (2021-11-24) [2024-09-16]. https://arxiv.org/abs/ 2111.12419.

[20] SHORTEN C, KHOSHGOFTAAR T M. A survey on image data augmentation for deep learning[J]. Journal of Big Data, 2019, 6(1): 1-48.

[21] FRID-ADAR M, KLANG E, AMITAI M, et al. Synthetic data augmentation using GAN for improved liver lesion classification[C]. Washington: 2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI 2018), IEEE, 2018.

[22] INOUE H. Data augmentation by pairing samples for images classification[J]. arXiv Preprint arXiv: 1801.02929, 2018.

[23] WONG S C, GATT A, STAMATESCU V, et al. Understanding data augmentation for classification: When to warp?[C]. Queensland: 2016 International Conference on Digital Image Computing: Techniques and Applications (DICTA), IEEE, 2016.

无人机航拍视角下密集场景非机动车小目标检测方法

Non-motorized Small Target Detection Method for Dense Scenes Under UAV Aerial Photography Perspective

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

[1]	姚振兴, 刘贤, 赵一飞, 王亮, 王彦琛. 手机信令不均匀定位下出行端点自适应识别方法[J]. 交通运输系统工程与信息, 2025, 25(4): 44-52.
[2]	张鹏, 李兴旺, 姬炳豪, 孙超, 李文权. 路口重复放行的公交与社会车辆协同绿波优化模型[J]. 交通运输系统工程与信息, 2025, 25(4): 53-62.
[3]	代亮, 杜鹏飞, 黄自彬, 杨朋博. 基于深度强化学习的城市交通信号分层协同控制方法[J]. 交通运输系统工程与信息, 2025, 25(4): 63-72.
[4]	王庞伟, 王思淼, 雷方舒, 徐京辉, 王子鹏, 王力. 混合动作表示强化学习下的城市交叉口智慧信控方法[J]. 交通运输系统工程与信息, 2025, 25(4): 73-83.
[5]	王连震, 沈超文, 王宇萍, 薛淑祺. 网联高速公路合流区基于间隙优化的车辆协同控制方法[J]. 交通运输系统工程与信息, 2025, 25(4): 84-95.
[6]	王维锋, 黄建鑫, 王晓全, 吴昕韩, 卞子馨. 基于无锚旋转框的航拍图像车辆全向检测方法[J]. 交通运输系统工程与信息, 2025, 25(4): 104-115.
[7]	陈峥, 张景, 陈博闻, 李春宇, 郭凤香, 魏福星. 基于异构多图时空融合的长时域车辆轨迹预测[J]. 交通运输系统工程与信息, 2025, 25(4): 126-136.
[8]	王祥, 任浩, 谭国真, 李健平, 王珏, 王妍力. 大语言模型协同强化学习的自动驾驶决策方法[J]. 交通运输系统工程与信息, 2025, 25(4): 137-146.
[9]	吴剑凡, 谢征宇, 秦勇, 王力, 王佳丽. 基于计算机视觉的地铁车站内乘客异常行为检测模型[J]. 交通运输系统工程与信息, 2025, 25(4): 162-174.
[10]	宋翠颖, 丁杰, 张春波. 模块化公交车辆调度研究综述[J]. 交通运输系统工程与信息, 2025, 25(4): 175-192.
[11]	谢秉磊, 冯健茜, 秦筱然. 多特征融合的网约车拼车起讫点需求时空预测[J]. 交通运输系统工程与信息, 2025, 25(4): 193-205.
[12]	陈喜群, 祝文琪, 吕朝锋. 融合轨迹时序与行为修正的车辆冲突风险预测[J]. 交通运输系统工程与信息, 2025, 25(4): 219-229.
[13]	高远, 付金龙, 冯文文. 考虑时空特征动态耦合的车辆轨迹预测方法[J]. 交通运输系统工程与信息, 2025, 25(3): 107-116.
[14]	赵霞, 李之红, 刘剑锋, 杨静, 吴梦琳, 秦伊萌. 行为模式时空动态超图聚类的公共交通异常团体检测[J]. 交通运输系统工程与信息, 2025, 25(3): 132-141.
[15]	常文文, 芦家磊, 黄霄, 闫光辉. 融合递归图的脑电驾驶行为分类方法研究[J]. 交通运输系统工程与信息, 2025, 25(3): 152-162.