改进YOLOv5的无人机小目标检测方法研究

doi:10.3778/j.issn.1002-8331.2309-0454

摘要/Abstract

摘要： 无人机拍摄图像的小目标具有特征信息不明显、背景复杂和部分目标遮挡的特点，导致检测时会出现误检和漏检的问题。针对这些问题，提出一种改进YOLOv5的无人机小目标检测方法SDT-YOLOv5。首先对数据集图像进行切片处理，使得每个切片中小目标的占比变高，提高小目标的识别能力。其次采用动态解耦检测头，引入动态卷积和自适应感受野机制，同时将检测头的分类和回归分支解耦，以实现更强的特征表达和提取能力。最后提出了基于最小点距离的交并比损失的最佳传输分配方法。最小化预测边界框与真实边界框之间的左上角和右下角点距离，再基于边界框的位置信息和距离度量，以最小化总成本为目标，找到最优的真实边界框与预测边界框匹配方案，提高检测的准确性。在VisDrone2019数据集上进行实验，结果显示改进YOLOv5的mAP50值达到了58.5%，相比于原始YOLOv5的mAP50值提高了23.2个百分点。这表明改进方法有效地提高了无人机小目标的检测精度，能够更准确地检测到小目标。

关键词: YOLOv5, 小目标检测, 检测头, 损失函数

Abstract: Small objects in images captured by UAV have the characteristics of unclear feature information, complex backgrounds, and partial target occlusion, which leads to problems of false detection and missed detection during detection. In response to these problems, an improved UAV small target detection method SDT-YOLOv5 is proposed. Firstly, the dataset images are sliced to increase the proportion of small objects in each slice and improve the recognition ability of small objects. Secondly, a dynamic decoupled detection head is used to introduce dynamic convolution and adaptive receptive field mechanism while decoupling the classification and regression branches of the detection head to achieve stronger feature expression and extraction capabilities. Finally, an optimal transmission allocation method based on intersection-union ratio loss of minimum point distance is proposed. The distance between the upper left corner and lower right corner of the predicted bounding box and the real bounding box is minimized, and then based on the position information and distance measurement of the bounding box, with the goal of minimizing the total cost, the optimal real bounding box and predicted bounding box matching scheme is found to improve detection accuracy. Through experimental results on the VisDrone2019 dataset, the mAP50 value of the improved YOLOv5 reaches 58.5%, which is 23.2 percentage points higher than the mAP50 value of the original YOLOv5. This shows that the improved method effectively improves the detection accuracy of UAV small objects and can detect small objects more accurately.

Key words: YOLOv5, small object detection, detection head, loss function

白宇, 周艳媛, 安胜彪. 改进YOLOv5的无人机小目标检测方法研究[J]. 计算机工程与应用, 2024, 60(10): 276-284.

BAI Yu, ZHOU Yanyuan, AN Shengbiao. Research on UAV Small Object Detection Method Improved by YOLOv5[J]. Computer Engineering and Applications, 2024, 60(10): 276-284.

参考文献

[1] ZOU Z, SHI Z, GUO Y, et al. Object detection in 20 years: a survey[J]. arXiv:1905.05055, 2019.
[2] WU X, SAHOO D, HOI S C H. Recent advances in deep learning for object detection[J]. Neurocomputing, 2020, 396: 39-64.
[3] ZAIDI S S A, ANSARI M S, ASLAM A, et al. A survey of modern deep learning based object detection models[J]. Digital Signal Processing, 2022: 103514.
[4] JIAO L, ZHANG F, LIU F, et al. A survey of deep learning-based object detection[J]. IEEE Access, 2019, 7: 128837-128868.
[5] ZHAO Z Q, ZHENG P, XU S, et al. Object detection with deep learning: a review[J]. IEEE Transactions on Neural Networks and Learning Systems, 2019, 30(11): 3212-3232.
[6] 高新波, 莫梦竟成, 汪海涛, 等.小目标检测研究进展[J].数据采集与处理, 2021, 36(3): 391-417.
GAO X B, MO M J C, WANG H T, et al. Recent advances in small object detection[J]. Journal of Data Acquisition and Processing, 2021, 36(3): 391-417.
[7] CHENG G, YUAN X, YAO X, et al. Towards large-scale small object detection: survey and benchmarks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(11): 13467-13488.
[8] ZHAO Y, JU Z, SUN T, et al. TGC-YOLOv5: an enhanced YOLOv5 drone detection model based on transformer, GAM & CA attention mechanism[J]. Drones, 2023, 7(7): 446.
[9] WU Y, LI J. YOLOv4 with deformable-embedding-transformer feature extractor for exact object detection in aerial imagery[J]. Sensors, 2023, 23(5): 2522.
[10] JUNG H K, CHOI G S. Improved YOLOv5: efficient object detection using drone images under various conditions[J]. Applied Sciences, 2022, 12(14): 7255.
[11] BETTI A, TUCCI M. YOLO-S: a lightweight and accurate YOLO-like network for small object detection in aerial imagery[J]. Sensors, 2023, 23(4): 1865.
[12] PEI W, SHI Z, GONG K. Small object detection with remote sensing images based on an improved YOLOv5 algorithm[J]. Frontiers in Neurorobotics, 2023, 16: 1074862.
[13] 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]. 计算机工程与应用, 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.
[14] 陈卫彪, 贾小军, 朱响斌, 等. 基于DSM-YOLO v5的无人机航拍图像目标检测[J]. 计算机工程与应用, 2023, 59(18): 226-233.
CHEN W B, JIA X J, ZHU X B, et al. Target detection for UAV image based on DSM-YOLO v5[J]. Computer Engineering and Applications, 2023, 59(18): 226-233.
[15] 张朝阳, 张上, 王恒涛, 等. 多尺度下遥感小目标多头注意力检测[J]. 计算机工程与应用, 2023, 59(8): 227-238.
ZHANG Z Y, ZHANG S, WANG H T, et al. Multi-head attention detection of small targets in remote sensing at multiple scales[J]. Computer Engineering and Applications, 2023, 59(8): 227-238.
[16] 杨晨, 佘璐, 杨璐, 等. 改进YOLOv5的遥感影像目标检测算法[J]. 计算机工程与应用, 2023, 59(15): 76-86.
YANG C, SHE L, YANG L, et al. Improved YOLOv5 object detection algorithm for remote sensing images[J]. Computer Engineering and Applications, 2023, 59(15): 76-86.
[17] 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 J H, WANG X H. Dense small object detection algorithm based on improved YOLOv5 in UAV aerial images[J]. Computer Engineering and Applications, 2024, 60(3): 100-108.
[18] 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 2021 IEEE/CVF International Conference on Computer Vision, 2021: 2778-2788.
[19] LIANG S, WU H, ZHEN L, et al. Edge YOLO: real-time intelligent object detection system based on edge-cloud cooperation in autonomous vehicles[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(12): 25345-25360.
[20] SILIANG M, YONG X. MPDIoU: a loss for efficient and accurate bounding box regression[J]. arXiv:2307.07662, 2023.
[21] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: optimal speed and accuracy of object detection[J]. arXiv: 2004.10934, 2020.
[22] DU D, ZHU P, WEN L, et al.VisDrone-DET2019: the vision meets drone object detection in image challenge results[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision Workshops, 2019: 213-226.
[23] GE Z, LIU S, WANG F, et al. YOLOx: exceeding YOLO series in 2021[J]. arXiv:2107.08430, 2021.
[24] PEYRé G, CUTURI M. Computational optimal transport: with applications to data science[J]. Foundations and Trends? in Machine Learning, 2019, 11(5/6): 355-607.
[25] ARISTODEMO A, GEMIGNANI L. Accelerating the Sinkhorn-Knopp iteration by Arnoldi-type methods[J]. Calcolo, 2020, 57(1): 10.
[26] REN S, HE K, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[C]//Advances in Neural Information Processing Systems 28, 2015.
[27] LAW H, DENG J. CornerNet: detecting objects as paired keypoints[C]//Proceedings of the 15th European Conference on Computer Vision, 2018: 734-750.
[28] 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 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 7464-7475.