LMFI-YOLO：复杂场景下的轻量化行人检测算法

doi:10.3778/j.issn.1002-8331.2503-0097

摘要/Abstract

摘要： 针对当前行人检测算法在复杂场景下存在误检、漏检及模型复杂度高等问题，提出一种改进YOLO11的轻量化行人检测算法——LMFI-YOLO融合RepConv来改进C3k2模块，构建RS-C3k2结构，以增强网络对行人特征的学习与捕捉能力。设计全新的颈部结构MBFPN，结合高效上采样模块与多尺度卷积模块，以强化特征融合并提升行人的特征表达能力，大幅提高检测精度。设计任务交互检测头TD-Detect，通过共享卷积与任务交互机制显著减少参数量和模型大小。为进一步提高检测精度，采用Focaler-GIoU作为边界框回归损失函数，通过聚焦不同回归样本来提升目标定位与整体性能。实验结果表明，所提算法在CityPersons数据集上mAP50提升8.5个百分点，模型参数量降至1.8×106，模型大小压缩至4.1 MB；在TinyPerson与CrowdHuman数据集上的泛化性实验表明，该算法在小尺寸目标和遮挡场景下的mAP50分别提升6.0和4.0个百分点。综合来看，LMFI-YOLO在保证检测精度显著提升的同时大幅降低了模型复杂度。

关键词: 行人检测, 小目标行人, 遮挡行人, 深度卷积, 任务交互

Abstract: Aiming at the problems of false detection, missing detection and high model complexity in the current pedestrian detection algorithm in complex scenes, an improved YOLO11 lightweight pedestrian detection algorithm LMFI-YOLO is proposed. RepConv is integrated to improve the C3k2 module, and RS-C3k2 structure is constructed to enhance the learning and capturing ability of pedestrian features. A new neck structure MBFPN is designed, which combines efficient upsampling module and multi-scale convolution module to strengthen feature fusion and enhance the feature expression ability of pedestrians, and greatly improve the detection accuracy. TD-Detect, a task interaction detection header, is designed to significantly reduce the number of parameters and model size by sharing the convolution and task interaction mechanism. In order to further improve the detection accuracy, Focaler-GIoU is used as the bounding box regression loss function to improve the target positioning and overall performance by focusing different regression samples. The experimental results show that the proposed algorithm increases mAP50 by 8.5 percentage points on the CityPersons dataset, reduces the number of model parameters to 1.8×106, and compacts the model size to 4.1 MB. Generalization experiments on TinyPerson and CrowdHuman datasets show that the mAP50 of the proposed algorithm can be improved by 6.0 and 4.0 percentage points in small-size targets and occlusion scenes, respectively. In summary, LMFI-YOLO significantly reduces the complexity of the model while significantly improving the detection accuracy.

Key words: pedestrian detection, small target pedestrian, occluded pedestrians, depth-wise convolution, task interaction

袁婷婷, 赖惠成, 汤静雯, 张晞, 高古学. LMFI-YOLO：复杂场景下的轻量化行人检测算法[J]. 计算机工程与应用, 2025, 61(15): 111-123.

YUAN Tingting, LAI Huicheng, TANG Jingwen, ZHANG Xi, GAO Guxue. LMFI-YOLO: Lightweight Pedestrian Detection Algorithm in Complex Scenes[J]. Computer Engineering and Applications, 2025, 61(15): 111-123.

参考文献

[1] GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2014: 580-587.
[2] GIRSHICK R. Fast R-CNN[C]//Proceedings of the 2015 IEEE International Conference on Computer Vision. Piscataway: IEEE, 2015: 1440-1448.
[3] 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.
[4] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot MultiBox detector[C]//Proceedings of the 14th European Conference on Computer Vision. Cham: Springer International Publishing, 2016: 21-37.
[5] 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.
[6] DONG C, LUO X S. Research on a pedestrian detection algorithm based on improved SSD network[J]. Journal of Physics: Conference Series, 2021, 1802(3): 032073.
[7] LAW H, DENG J. CornerNet: detecting objects as paired keypoints[C]//Proceedings of the European Conference on Computer Vision. Cham: Springer, 2018: 765-781.
[8] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2016: 779-788.
[9] 聂源, 赖惠成, 高古学. 改进YOLOv7+Bytetrack的小目标检测与追踪[J]. 计算机工程与应用, 2024, 60(12): 189-202.
NIE Y, LAI H C, GAO G X. Improved small target detection and tracking with YOLOv7+Bytetrack[J]. Computer Engineering and Applications, 2024, 60(12): 189-202.
[10] LI C, WANG Y D, LIU X M. An improved YOLOv7 lightweight detection algorithm for obscured pedestrians[J]. Sensors, 2023, 23(13): 5912.
[11] 汤静雯, 赖惠成, 王同官. 远距离情形下的改进YOLOv8行人检测算法[J]. 计算机工程, 2025, 51(4): 303-313.
TANG J W, LAI H C, WANG T G. Improved YOLOv8 pedestrian detection algorithm for long-distance situations[J]. Computer Engineering, 2025, 51(4): 303-313.
[12] TANG J W, LAI H C, GAO G X, et al. PFEL-Net: a lightweight network to enhance feature for multi-scale pedestrian detection[J]. Journal of King Saud University-Computer and Information Sciences, 2024, 36(8): 102198.
[13] 李琳, 靳志鑫, 俞晓磊, 等. Haar小波下采样优化YOLOv9的道路车辆和行人检测[J]. 计算机工程与应用, 2024, 60(20): 207-214.
LI L, JIN Z X, YU X L, et al. Road vehicle and pedestrian detection based on YOLOv9 for haar wavelet downsampling[J]. Computer Engineering and Applications, 2024, 60(20): 207-214.
[14] 何湘杰, 宋晓宁. YOLOv4-Tiny的改进轻量级目标检测算法[J]. 计算机科学与探索, 2024, 18(1): 138-150.
HE X J, SONG X N. Improved YOLOv4-tiny lightweight target detection algorithm[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(1): 138-150.
[15] 王春梅, 刘欢. YOLOv8-VSC: 一种轻量级的带钢表面缺陷检测算法[J]. 计算机科学与探索, 2024, 18(1): 151-160.
WANG C M, LIU H. YOLOv8-VSC: lightweight algorithm for strip surface defect detection[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(1): 151-160.
[16] 胡晓伟, 闫奕昕, 王大为, 等. 基于YOLOM算法的路面病害轻量化检测方法[J]. 中国公路学报, 2024, 37(12): 381-391.
HU X W, YAN Y X, WANG D W, et al. Lightweight pavement disease detection based on YOLOM algorithm[J]. China Journal of Highway and Transport, 2024, 37(12): 381-391.
[17] 张浩晨, 张竹林, 史瑞岩, 等. YOLO-CDC: 优化改进YOLOv8的车辆目标检测算法[J]. 计算机工程与应用, 2025, 61(13): 124-137.
ZHANG H C, ZHANG Z L, SHI R Y, et al. YOLO-CDC: improved YOLOv8 vehicle object detection algorithm[J]. Computer Engineering and Applications, 2025, 61(13): 124-137.
[18] TAN M X, PANG R M, LE Q V. EfficientDet: scalable and efficient object detection[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 10778-10787.
[19] SANDLER M, HOWARD A, ZHU M L, et al. MobileNetV2: inverted residuals and linear bottlenecks[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 4510-4520.
[20] FENG C J, ZHONG Y J, GAO Y, et al. TOOD: task-aligned one-stage object detection[C]//Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2021: 3490-3499.
[21] ZHANG H, ZHANG S J. Focaler-IoU: more focused intersection over union loss[J]. arXiv:2401.10525, 2024.
[22] ZHANG S S, BENENSON R, SCHIELE B. CityPersons: a diverse dataset for pedestrian detection[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 4457-4465.
[23] YU X H, GONG Y Q, JIANG N, et al. Scale match for tiny person detection[C]//Proceedings of the 2020 IEEE Winter Conference on Applications of Computer Vision. Piscataway: IEEE, 2020: 1246-1254.
[24] SHAO S, ZHAO Z J, LI B X, et al. CrowdHuman: a benchmark for detecting human in a crowd[J]. arXiv:1805.00123, 2018.
[25] ZHU P F, DU D W, WEN L Y, et al. VisDrone-VID2019: the vision meets drone object detection in video challenge results[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision Workshop. Piscataway: IEEE, 2019: 227-235.
[26] ZHAO Y A, LV W Y, XU S L, et al. DETRs beat YOLOs on real-time object detection[C]//Proceedings of the 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2024: 16965-16974.
[27] WU G N, WU Q H. Enhancing steel surface defect detection: a hyper-YOLO approach with ghost modules and hyper FPN[J]. IAENG International Journal of Computer Science, 2024, 51(9): 1321-1330.
[28] WEI J, CHE K, GONG J Y, et al. Fast and accurate detection of dim and small targets for smart micro-light sight[J]. Electronics, 2024, 13(16): 3301.
[29] ZHANG F F, LEONG L V, YEN K S, et al. An enhanced lightweight model for small-scale pedestrian detection based on YOLOv8s[J]. Digital Signal Processing, 2025, 156: 104866.
[30] 江旺玉, 王乐, 姚叶鹏, 等. 多尺度特征聚合扩散和边缘信息增强的小目标检测算法[J]. 计算机工程与应用, 2025, 61(7): 105-116.
JIANG W Y, WANG L, YAO Y P, et al. Multi-scale feature aggregation diffusion and edge information enhancement small object detection algorithm[J]. Computer Engineering and Applications, 2025, 61(7): 105-116.
[31] 黄俊杰, 胡畅, 包嘉琪, 等. 轻量型密集行人检测算法研究[J]. 计算机仿真, 2024, 41(5): 183-188.
HUANG J J, HU C, BAO J Q, et al. Research on lightweight dense pedestrian detection algorithm[J]. Computer Simulation, 2024, 41(5): 183-188.
[32] 王泽宇, 徐慧英, 朱信忠, 等. 基于YOLOv8改进的密集行人检测算法: MER-YOLO[J]. 计算机工程与科学, 2024, 46(6): 1050-1062.
WANG Z Y, XU H Y, ZHU X Z, et al. An improved dense pedestrian detection algorithm based on YOLOv8: MER-YOLO[J]. Computer Engineering & Science, 2024, 46(6): 1050-1062.
[33] 蒋源, 朱高峰, 朱凤华, 等. 基于复合缩放的动态航拍小目标检测算法[J]. 航空兵器, 2025, 32(2): 104-112.
JIANG Y, ZHU G F, ZHU F H, et al. Dynamic aerial small target detection algorithm based on compound zoom scaling[J]. Aero Weaponry, 2025, 32(2): 104-112.
[34] 董一兵, 曾辉, 侯少杰. LMUAV-YOLOv8: 低空无人机视觉目标检测轻量化网络[J]. 计算机工程与应用, 2025, 61(3): 94-110.
DONG Y B, ZENG H, HOU S J. LMUAV-YOLOv8: lightweight network for object detection in low-altitude UAV vision[J]. Computer Engineering and Applications, 2025, 61(3): 94-110.