面向交通执法中车辆使用远光灯检测的YOLOv10-ECC算法研究

张立立, 张珂, 杨康, 魏薇, 李晶, 谭洪鑫

计量学报 ›› 2025, Vol. 46 ›› Issue (5) : 659-666.

PDF(2330 KB)
PDF(2330 KB)
计量学报 ›› 2025, Vol. 46 ›› Issue (5) : 659-666. DOI: 10.3969/j.issn.1000-1158.2025.05.06
光学计量

面向交通执法中车辆使用远光灯检测的YOLOv10-ECC算法研究

作者信息 +

Research on YOLOv10-ECC Algorithm for Vehicle High-beam Detection for Traffic Enforcement Scenarios

Author information +
文章历史 +

摘要

为了快速准确检测车辆使用远光灯的情况,提高交通管理部门执法效率,提出了一种端到端的实时YOLOv10-ECC检测算法。首先,在YOLOv10的骨干网络中设计了PSA_EMA模块,利用高效多尺度注意力模块(EMA)在保持参数效率的同时捕获更丰富的特征信息;其次,在颈部网络里设计了卷积特征重组模块(CCM),实现了上采样特征增强,构建了更为细微的特征图。在自采的远光灯车辆数据集上的实验结果表明:YOLOv10-ECC的均值平均精度mAP@0.5达到了93.2%,参数量为2.86×106,单张照片检测时间为2.2 ms,计算复杂度为9.0 GFLOP/s,能够实时、高精度地检测远光灯车辆。

Abstract

To rapid and accurate detection of vehicles using high-beam headlights can assist traffic management departments in efficient law enforcement. An end-to-end real-time YOLOv10-ECC detection algorithm is proposed. Firstly, the PSA_EMA module is designed in the backbone of YOLOv10, which uses the efficient multi-scale attention module (EMA) to capture more abundant feature information while maintaining the parameter efficiency. Secondly, the convolutional feature recombination module (CCM) is designed in the neck, which realizes the up-sampling feature enhancement and constructs the more subtle feature map. The experimental results on the self-collected high-beam vehicles dataset show that the mean average accuracy mAP@0.5 of YOLOv10-ECC reaches 93.2%, the number of parameters is 2.86×106, the detection time of a single photo is 2.2 ms, and the computational complexity is 9.0 GFLOP/s, which can detect high-beam vehicles in real time and with high accuracy.

关键词

机械视觉测量 / 远光灯检测 / YOLOv10-ECC / EMA模块 / CCM模块 / 上采样算子 / 端到端

Key words

mechanical vision measurement / high-beam detection / YOLOv10-ECC / EMA module / CCM module / CARAFE / end-to-end

引用本文

导出引用
张立立, 张珂, 杨康, . 面向交通执法中车辆使用远光灯检测的YOLOv10-ECC算法研究[J]. 计量学报. 2025, 46(5): 659-666 https://doi.org/10.3969/j.issn.1000-1158.2025.05.06
ZHANG Lili, ZHANG Ke, YANG Kang, et al. Research on YOLOv10-ECC Algorithm for Vehicle High-beam Detection for Traffic Enforcement Scenarios[J]. Acta Metrologica Sinica. 2025, 46(5): 659-666 https://doi.org/10.3969/j.issn.1000-1158.2025.05.06
中图分类号: TB96   

参考文献

1
吴云强,华佳峰,陈鹰,等.机动车不按规定使用远光灯治理难点及对策[J].中国公共安全(学术版)2018(1):78-81.
WU Y Q HUA J F CHEN Y, et al. Difficulties and Counter-measures for illegal use of driving beam headlamp of motor vehicles[J]. China Public Security( Academy Edition)2018(1): 78-81.
2
SU J Y ZHANG W W WU X C, et al. The driving beam recognition of vehicle based on multi-structure feature extraction and trajectory tracking[J]. Journal of Electronic Measurement and Instrumentation201832(10): 103-110.
3
龚航, 刘培顺. 夜间行驶车辆远光灯检测方法[J]. 计算机科学202148(12): 256-263.
GONG H LIU P S. Detection method of high beam of vehicles driving at night[J]. Computer science202148(12): 256-263.
4
CHU Y C GONG H WANG X F, et al. Research on detection methods of high beam in night driving vehicles[C]//Thirteenth International Conference on Graphics and Image Processing. Kunming, China, 2021: 2954-2961.
5
赵伟强, 孙航. 一种远光灯检测、预警系统及方法: 202111255664.9[P]. 2022-01-25.
6
王晓雷, 毕宵飞, 吕彤辉 等. 一种用于远光灯检测的车辆行驶场景识别区分方法: 202210586458.4[P]. 2022-05-27.
7
米增, 连哲.面向通用目标检测的YOLO方法研究综述[J/OL]. 计算机工程与应用: 1-19.
8
WANG A CHEN H LIU L, et al. Yolov10: Real-time end-to-end object detection[J]. arXiv preprint arXiv: 2405. 14458, 2024.
9
TAN H LIU X YIN B, et al. MHSA-Net: Multihead self-attention network for occluded person re-identification[J]. IEEE Transactions on Neural Networks and Learning Systems202234(11): 8210-8224.
10
SUN Y WANG Y FU K, et al. FMA-ETA: Estimating travel time entirely based on FFN with attention[C]// IEEE International Conference on Acoustics, Speech and Signal Processing. Toronto, Canada, 2021: 3355-3359.
11
OUYANG D HE S ZHANG G, et al. Efficient multi-scale attention module with cross-spatial learning[C]//ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) IEEE. Rhodes, Greece, 2023: 1-5.
12
WANG J CHEN K XU R, et al. Carafe: Content-aware reassembly of features[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Seoul, South Korea, 2019: 3007-3016.
13
NORKOBIL S S ABDUSSALOMOV A JAMIL M K, et al. A YOLOv6-based improved fire detection approach for smart city environments[J]. Sensors202323(6): 3161-3172.
14
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 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver, Canada, 2023: 7464-7475.
15
WANG X GAO H JIA Z, et al. BL-YOLOv8: An improved road defect detection model based on YOLOv8[J]. Sensors202323(20): 8361-8373..

基金

宁夏自然科学基金(2022AAC03757)
北京市数字教育研究课题(BDEC2022619048)
北京市教育委员会科研计划项目(KM202410017006)
北京市高等教育学会课题(MS2022144)

PDF(2330 KB)

Accesses

Citation

Detail

段落导航
相关文章

/