RO-YOLOv9车辆行人检测算法

doi:10.3778/j.issn.1002-8331.2409-0128

摘要/Abstract

摘要： 针对道路交通环境中车辆和行人目标较小或被遮挡导致的检测精度低以及误检、漏检问题，提出道路目标检测算法RO-YOLOv9。增加小目标检测层，增强算法对小目标的特征学习能力。设计双向与自适应尺度融合特征金字塔网络（bidirectional and adaptive scale fusion feature pyramid network，BiASF-FPN）结构，优化多尺度特征融合，保证算法有效捕捉从小尺度到大尺度目标的详细信息。提出OR-RepN4模块，通过重参数化策略，复杂算法结构简单化，提高推理速度。引用Shape-NWD（shape neighborhood weighted decomposition）损失函数，专注边界框形状与尺寸，采用归一化高斯Wasserstein距离平滑回归，实现跨尺度不变性，降低小尺度与遮挡目标的检测误差。实验结果表明，在优化后的SODA10M和BDD100K数据集下，RO-YOLOv9算法的mAP@0.5（mean average precision）分别达到68.1%和56.8%，比YLOLOv9算法提高5.6个百分点和4.4个百分点，并且检测帧率分别达到了55.3 帧/s和54.2 帧/s，达到检测精度和检测速度的平衡。

关键词: YOLOv9, 小目标检测, 双向与自适应尺度融合特征金字塔网络（BiASF-FPN）, OR-RepN4, Shape-NWD

Abstract: Aiming at the low detection accuracy, false detection and missed detection problems caused by small or occluded vehicle and pedestrian targets in road traffic environments, a road target detection algorithm RO-YOLOv9 is proposed. Firstly, the small target detection layer is added to enhance the algorithm’s feature learning ability for small targets. Secondly, the bidirectional and adaptive scale fusion feature pyramid network (BiASF-FPN) structure is designed to optimize multi-scale feature fusion and ensure that the algorithm effectively captures detailed information from small to large scale targets. Thirdly, the OR-RepN4 module is proposed to simplify the complex algorithm structure and improve the inference speed through the re-parameterization strategy. Finally, the shape neighborhood weighted decomposition (Shape-NWD) loss function is used to focus on the shape and size of the bounding box, and the normalized Gaussian Wasserstein distance smoothing regression is used to achieve cross-scale invariance and reduce the detection error of small-scale and occluded targets. The experimental results show that under the optimized SODA10M and BDD100K datasets, the mAP@ 0.5 (mean average precision) of RO-YOLOv9 algorithm reaches 68.1% and 56.8%, respectively, which is 5.6 percentage points and 4.4 percentage points higher than that of the YLOLOv9 algorithm, and the detection frame rates reach 55.3 and 54.2 frames per second, respectively, achieving a balance between detection accuracy and detection speed.

Key words: YOLOv9, small target detection, bidirectional and adaptive scale fusion feature pyramid network(BiASF-FPN), OR-RepN4, shape neighborhood weighted decomposition(Shape-NWD)

廖炎华, 万学俊, 赵周洲, 潘文林. RO-YOLOv9车辆行人检测算法[J]. 计算机工程与应用, 2025, 61(11): 144-155.

LIAO Yanhua, WAN Xuejun, ZHAO Zhouzhou, PAN Wenlin. RO-YOLOv9 Vehicle and Pedestrian Detection Algorithm[J]. Computer Engineering and Applications, 2025, 61(11): 144-155.

参考文献

[1] VIOLA P, JONES M. Rapid object detection using a boosted cascade of simple features[C]//Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2001.
[2] DALAL N, TRIGGS B. Histograms of oriented gradients for human detection[C]//Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition , 2005: 886-893.
[3] LOWE D G. Distinctive image features from scale-invariant keypoints[J]. International Journal of Computer Vision, 2004, 60: 91-110.
[4] 包本刚. 融合多特征的目标检测与跟踪方法[J]. 电子测量与仪器学报, 2019, 33(9): 93-99.
BAO B G. Target detection and tracking based on multi-feature fusion[J]. Journal of Electronic Measurement and Instrumentation, 2019, 33(9): 93-99.
[5] GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2014: 580-587.
[6] GIRSHICK R. Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision, 2015: 1440-1448.
[7] REN S, HE K, 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] 曹磊, 王强, 史润佳, 等. 基于改进RPN的Faster-RCNN网络SAR图像车辆目标检测方法[J]. 东南大学学报(自然科学版), 2021, 51(1): 87-91.
CAO L, WANG Q, SHI R J, et al. Method for vehicle target detection on SAR image based on improved RPN in Faster-RCNN[J]. Journal of Southeast University(Natural Science Edition), 2021, 51(1): 87-91.
[9] MA J, WAN H, WANG J, et al. An improved one-stage pedestrian detection method based on multi-scale attention feature extraction[J]. Journal of Real-Time Image Processing, 2021, 18: 1965-1978.
[10] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector[C]//Proceedings of the 14th European Conference on Computer Vision, 2016: 21-37.
[11] 徐彦威, 李军, 董元方, 等. YOLO系列目标检测算法综述[J]. 计算机科学与探索, 2024, 18(9): 2221-2238.
XU Y W, LI J, DONG Y F, et al. Survey of development of YOLO object detection algorithms[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(9): 2221-2238.
[12] 宋宇博, 高嘉振. 改进YOLOv3算法的交通多目标检测方法[J]. 北京邮电大学学报, 2022, 45(5): 103-108.
SONG Y B, GAO J Z. Improved YOLOv3 algorithm for multi-target detection of traffic[J]. Journal of Beijing University of Posts and Telecommunications, 2022, 45(5): 103-108.
[13] FU P B, ZHANG X, YANG H R. Answer sheet layout analysis based on YOLOv5s-DC and MSER[J]. The Visual Computer, 2024, 40(9): 6111-6122.
[14] 杜娟, 崔少华, 晋美娟, 等. 改进YOLOv7的复杂道路场景目标检测算法[J]. 计算机工程与应用, 2024, 60(1): 96-103.
DU J, CUI S H, JIN M J, et al. Improved complex road scene object detection algorithm of YOLOv7[J]. Computer Engineering and Applications, 2024, 60(1): 96-103.
[15] 盛帅, 段先华, 胡维康, 等. Dynamic-YOLOX: 复杂背景下的苹果叶片病害检测模型[J]. 计算机科学与探索, 2024, 18(8): 2118-2129.
SHENG S, DUAN X H, HU W K, et al. Dynamic-YOLOX: detection model for apple leaf disease in complex background[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(8): 2118-2129.
[16] AN R, ZHANG X, SUN M, et al. GC-YOLOv9: Innovative smart city traffic monitoring solution[J]. Alexandria Engineering Journal, 2024, 106: 277-287.
[17] SHI Y, LI S, LIU Z, et al. MTP-YOLO: you only look once based maritime tiny person detector for emergency rescue [J]. Journal of Marine Science and Engineering, 2024, 12(4): 669.
[18] TAN M, PANG R, LE Q V. Efficientdet: scalable and efficient object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 10781-10790.
[19] KANG M, TING C M, TING F F, et al. ASF-YOLO: a novel YOLO model with attentional scale sequence fusion for cell instance segmentation[J]. Image and Vision Computing, 2024, 147: 105057.
[20] HU M, FENG J, HUA J, et al. Online convolutional re-parameterization[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022: 568-577.
[21] MA Q, JIN S, BIAN G, et al. Multi-scale marine object detection in side-scan sonar images based on BES-YOLO [J]. Sensors, 2024, 24(14): 4428.
[22] ZENG Y, ZHANG T, HE W, et al. YOLOv7-UAV: an unmanned aerial vehicle image object detection algorithm based on improved YOLOv7[J]. Electronics, 2023, 12(14): 3141.