YOLO-TKD：用于自动驾驶的轻量化时序交互信号检测模型

胡振虓; 姚肖婷; 席治远; 王晓

doi:10.11959/j.issn.2096-6652.ZNKX26020

您当前的位置：

首页 >

文章列表页 >

YOLO-TKD：用于自动驾驶的轻量化时序交互信号检测模型

学术论文 | 更新时间：2026-07-29

- YOLO-TKD：用于自动驾驶的轻量化时序交互信号检测模型
- YOLO-TKD: lightweight temporal interactive signal detection model for autonomous driving
- 智能科学与技术学报 2026年8卷第2期页码：196-206
- 作者机构：
  
  1.安徽大学人工智能学院，安徽合肥 230601
  2.安徽大学具身智能研究院，安徽合肥 230031
- 作者简介：
  
  胡振虓（2000- ），男，安徽大学人工智能学院硕士生，主要研究方向为交互信号识别、计算机视觉、机器学习。
  姚肖婷（1998- ），女，安徽大学人工智能学院硕士生，主要研究方向为智能决策、轨迹预测、智能交通系统。
  王晓（1988- ），女，安徽大学人工智能学院教授，主要研究方向为社会计算、群体行为建模、无人自主系统及其平行测试。xiao.wang@ahu.edu.cn
- 基金信息：
  
  国家自然科学基金资助项目(62522601)
- DOI：10.11959/j.issn.2096-6652.ZNKX26020
  中图分类号： TP391
- 收稿：2026-02-10，
  
  修回：2026-06-03，
  
  录用：2026-06-05，
  
  网络首发：2026-07-27，
  
  纸质出版：2026-06-15
- 稿件说明：
移动端阅览
胡振虓,姚肖婷,席治远等.YOLO-TKD：用于自动驾驶的轻量化时序交互信号检测模型[J].智能科学与技术学报,2026,08(02):196-206.

Hu Zhenxiao,Yao Xiaoting,Xi Zhiyuan,et al.YOLO-TKD: lightweight temporal interactive signal detection model for autonomous driving[J].Chinese Journal of Intelligent Science and Technology,2026,08(02):196-206.
胡振虓,姚肖婷,席治远等.YOLO-TKD：用于自动驾驶的轻量化时序交互信号检测模型[J].智能科学与技术学报,2026,08(02):196-206. DOI： 10.11959/j.issn.2096-6652.ZNKX26020.

Hu Zhenxiao,Yao Xiaoting,Xi Zhiyuan,et al.YOLO-TKD: lightweight temporal interactive signal detection model for autonomous driving[J].Chinese Journal of Intelligent Science and Technology,2026,08(02):196-206. DOI： 10.11959/j.issn.2096-6652.ZNKX26020.

摘要

在自动驾驶的感知任务中，准确理解交通参与者发出的交互信号是实现安全拟人化决策的关键。当前高精度模型难以实现实时交互信号检测，且普遍缺乏对转向灯闪烁、行人手势等动态时序信号的有效建模，导致复杂场景下漏检误判频繁发生，构成安全隐患。为此，提出一种面向自动驾驶边缘设备的轻量-时序-蒸馏协同优化框架模型，即轻量化时序知识蒸馏YOLO（YOLO-TKD），旨在解决交互信号检测中时序建模需求与计算资源受限的矛盾。该模型以YOLOv8n为基线，首先采用MobileNetV2作为主干网络以压缩模型规模；继而引入时序卷积网络模块（TCN），捕捉连续帧之间的动态依赖关系；最后，结合知识蒸馏策略（KD），保持其识别性能。实验结果表明，本方法的精确率、召回率和平均精度均值分别达到87.5%、86.4%和84.3%，相较于基线YOLOv8n分别提升了3.2%、2.6%和1.1%。同时，模型参数量减少26.3%。所提YOLO-TKD在显著降低资源消耗的同时，有效提升了交互信号检测的精度与时序鲁棒性，为自动驾驶边缘设备提供了可行的技术路径。

Abstract

In autonomous driving perception

accurate understanding of interaction signals is the key to safe

human-like decisions. Real-time detection of these signals is found difficult to achieve with current high-precision models. Effective modeling of dynamic temporal signals was generally lacking. However

effective modeling of dynamic temporal signals

such as turn signal flashes and pedestrian gestures. As a result

missed detections and false alarms occur frequently in complex scenarios

creating safety risks. To address this

a lightweight-temporal-knowledge distillation collaborative optimization model for autonomous driving edge devices was proposed

named lightweight temporal knowledge-distilled YOLO (YOLO-TKD). The model was designed to resolve the conflict between the need for temporal modeling in interactive signal detection and the constraints of computational resources. Firstly

the YOLOv8n baseline was adopted

with MobileNetV2 utilized as the backbone network to compress the model size. Subsequently

a temporal convolutional network (TCN) module was introduced to capture dynamic dependencies between consecutive frames. Finally

a knowledge distillation (KD) strategy was incorporated to maintain recognition performance. Experimental results on the dataset show that the proposed method achieves precision

recall

and mean average precision of 87.5%

86.4%

and 84.3%

respectively

which represent improvements of 3.2%

2.6%

and 1.1% over the baseline YOLOv8n. At the same time

the model parameter count is reduced by 26.3%. The proposed YOLO-TKD not only significantly reduces resource consumption but also effectively improves the accuracy and temporal robustness of interaction signal detection

providing a viable technical solution for autonomous driving edge devices.

关键词

Keywords

references

Wang X, Jiang R, Li L, et al. Capturing car-following behaviors by deep learning[J]. IEEE Transactions on Intelligent Transportation Systems, 2018, 19(3): 910-920.

Wang X M, Zheng X H, Chen W, et al. Visual human-computer interactions for intelligent vehicles and intelligent transportation systems: the state of the art and future directions[J]. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2021, 51(1): 253-265.

Wang Z J, Wu Y, Niu Q Q. Multi-sensor fusion in automated driving: a survey[J]. IEEE Access, 2020, 8: 2847-2868.

刘泽禹, 张慧, 李浥东, 等. 面向自动驾驶的感知决策一体化综述[J]. 智能科学与技术学报, 2025, 7(1): 4-20.

Liu Z Y, Zhang H, Li Y D, et al. A review on perception-decision integration for autonomous driving[J]. Chinese Journal of Intelligent Science and Technology, 2025, 7(1): 4-20.

Wang S, Yuan Y, Wang X, et al. An overview of smart contract: architecture, applications, and future trends[C]//Proceedings of the 2018 IEEE Intelligent Vehicles Symposium (Ⅳ). Piscataway: IEEE Press, 2018: 108-113.

Diwan T, Anirudh G, Tembhurne J V. Object detection using YOLO: challenges, architectural successors, datasets and applications[J]. Multimedia Tools and Applications, 2023, 82(6): 9243-9275.

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 Press, 2014: 580-587.

He K M, Gkioxari G, Dollár P, et al. Mask R-CNN[C]//Proceedings of the 2017 IEEE International Conference on Computer Vision (ICCV). Piscataway: IEEE Press, 2017: 2980-2988.

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 (CVPR). Piscataway: IEEE Press, 2016: 779-788.

Liu W, Anguelov D, Erhan D, et al. SSD: single shot MultiBox detector[C]//Computer Vision - ECCV 2016. Berlin: Springer, 2016: 21-37.

Zhao Y F, Gao H, Wang S, et al. A novel approach for traffic signal control: a recommendation perspective[J]. IEEE Intelligent Transportation Systems Magazine, 2017, 9(3): 127-135.

杜泉成, 王晓, 李灵犀, 等. 行人轨迹预测方法关键问题研究: 现状及展望[J]. 智能科学与技术学报, 2023, 5(2): 143-162.

Du Q C, Wang X, Li L X, et al. Key problems and progress of pedestrian trajectory prediction methods: the state of the art and prospects[J]. Chinese Journal of Intelligent Science and Technology, 2023, 5(2): 143-162.

Du Q C, Wang X, Yin S G, et al. Social force embedded mixed graph convolutional network for multi-class trajectory prediction[J]. IEEE Transactions on Intelligent Vehicles, 2024, 9(9): 5571-5580.

Liang S Y, 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.

Carrasco D P, Rashwan H A, García M Á, et al. T-YOLO: tiny vehicle detection based on YOLO and multi-scale convolutional neural networks[J]. IEEE Access, 2023, 11: 22430-22440.

Xu L Z, Yan W, Ji J S. The research of a novel WOG-YOLO algorithm for autonomous driving object detection[J]. Scientific Reports, 2023, 13: 3699.

Li C, Wang Y D, Liu X M. An improved YOLOv7 lightweight detection algorithm for obscured pedestrians[J]. Sensors, 2023, 23(13): 5912.

Jensen M B, Nasrollahi K, Moeslund T B. Evaluating state-of-the-art object detector on challenging traffic light data[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). Piscataway: IEEE Press, 2017: 882-888.

余杨, 何刘. 基于YOLOv8和SVDD重识别的交通信号灯检测方法[J]. 汽车实用技术, 2024, 49(23): 49-54.

Yu Y, He L. Traffic light detection method based on YOLOv8 and re-identification using SVDD[J]. Automobile Applied Technology, 2024, 49(23): 49-54.

Bai S J, Kolter J Z, Koltun V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling[PP]. V2. (2018-04-19)[2026-02-10]. arXiv:arXiv.1803.01271.

Hinton G, Vinyals O, Dean J. Distilling the knowledge in a neural network[PP]. V1. (2015-03-09)[2026-02-10]. arXiv:arXiv.1503.02531.

Zhang X Y, Zhou X Y, Lin M X, et al. ShuffleNet: an extremely efficient convolutional neural network for mobile devices[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2018: 6848-6856.

Howard A, Sandler M, Chen B, et al. Searching for MobileNetV3[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision (ICCV). Piscataway: IEEE Press, 2019: 1314-1324.

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 Press, 2018: 4510-4520.

Zhou D Q, Hou Q B, Chen Y P, et al. Rethinking bottleneck structure for efficient mobile network design[C]//Computer Vision - ECCV 2020. Berlin: Springer, 2020: 680-697.

Tian S S, Li L S, Li W J, et al. A survey on few-shot class-incremental learning[J]. Neural Networks, 2024, 169: 307-324.

Shu C Y, Liu Y F, Gao J F, et al. Channel-wise knowledge distillation for dense prediction[C]//Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision (ICCV). Piscataway: IEEE Press, 2021: 5291-5300.

Rasouli A, Kotseruba I, Kunic T, et al. PIE: a large-scale dataset and models for pedestrian intention estimation and trajectory prediction[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision (ICCV). Piscataway: IEEE Press, 2019: 6261-6270.

Rasouli A, Kotseruba I, Tsotsos J K. Are they going to cross? a benchmark dataset and baseline for pedestrian crosswalk behavior[C]//Proceedings of the 2017 IEEE International Conference on Computer Vision Workshops (ICCVW). Piscataway: IEEE Press, 2017: 206-213.

Zhu X K, Lyu S C, 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 Workshops (ICCVW). Piscataway: IEEE Press, 2021: 2778-2788.

Li C Y, Li L L, Jiang H L, et al. YOLOv6: a single-stage object detection framework for industrial applications[PP]. V1. (2022-09-07)[2026-02-10]. arXiv: arXiv.2209.02976.

Wang C Y, Bochkovskiy A, Liao H 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 (CVPR). Piscataway: IEEE Press, 2023: 7464-7475.

Yaseen M. What is YOLOv8: an in-depth exploration of the internal features of the next-generation object detector[PP]. V1. (2024-08-28)[2026-02-10]. arXiv: arXiv.2408.15857.

Wang A, Chen H, Liu L H, et al. YOLOv10: real-time end-to-end object detection[PP]. V2. arXiv (2024-10-30)[2026-02-10]. arXiv: arXiv.2405.14458.

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 (CVPR). Piscataway: IEEE Press, 2024: 16965-16974.

KÖPÜKLÜ O, Wei X, Rigoll G. You only watch once: a unified CNN architecture for real-time spatiotemporal action localization[PP]. V5. (2021-10-18)[2026-02-10]. arXiv: arXiv.1911.06644.

Zhu F H, Lv Y S, Chen Y Y, et al. Parallel transportation systems: toward IoT-enabled smart urban traffic control and management[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 21(10): 4063-4071.

Du Q C, Xu L L, Wu Q, et al. SafeCrossNet:multi-modal fusion with social-aware for pedestrian crossing intention prediction[J]. Information Fusion, 2026, 126: 103609.

Cui Y D, Huang S C, Zhong J M, et al. DriveLLM: charting the path toward full autonomous driving with large language models[J]. IEEE Transactions on Intelligent Vehicles, 2024, 9(1): 1450-1464.

Jang D, Lee S, Choi G, et al. Energy-based domain adaptation without intermediate domain dataset for foggy scene segmentation[J]. IEEE Transactions on Image Processing, 2024, 33: 6143-6157.

Shen Y, Wei M, Wang Y, et al. Rethinking real-world image deraining via an unpaired degradation-conditioned diffusion model[PP]. V4. (2024-05-01)[2026-02-10]. arXiv: arXiv.2301.09430.

Qin H D, Lu X, Wang L C, et al. Foggy-aware teacher: an unsupervised domain adaptive learning framework for object detection in foggy scenes[J]. IEEE Robotics and Automation Letters, 2025, 10(7): 7508-7515.

刘文华, 李浥东, 王涛, 等. 基于高维特征表示的交通场景识别[J]. 智能科学与技术学报, 2019, 1(4): 392-399.

Liu W H, Li Y D, Wang T, et al. Transportation scene recognition based on high level feature representation[J]. Chinese Journal of Intelligent Science and Technology, 2019, 1(4): 392-399.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

注意力机制增强的输煤传送带异物检测

UAVAI-YOLO：无人机航拍图像的小目标检测模型

基于深度学习的无人机航拍输电线路侵入异物检测综述

K-12 AI课程的自我调节学习重塑：基于LLM智能体的学习辅助框架

人工智能的知识本源