Article Details

Abstract

To address the problems of low detection accuracy and difficult multisource feature fusion in vehicle following behavior detection in the Internet of Vehicles (IoV), this paper proposes a topological representation and detection scheme based on a Graph Neural Network. Firstly, the scheme constructs a mapping mechanism between multisource heterogeneous features and graph nodes, converting structured data such as vehicle positioning and temporal interactions, as well as unstructured visual and facial data, into unified graph node attributes to achieve effective fusion of heterogeneous features. Second, a dynamic spatiotemporal graph incorporating real-time interactions among vehicles is constructed, and an image-capture count calculation function integrating temporal decay and distance weighting is designed to define graph edge weights and accurately characterize the dynamic association strength between vehicles. Finally, a temporal graph convolutional network model is proposed to accurately identify following behavior by capturing the spatiotemporal evolution information of the graph structure. Experimental results verify the accuracy and stability of the proposed scheme, which effectively overcomes the technical bottlenecks of traditional methods and provides a technical approach and methodological support for following behavior detection in vehicle safe-driving monitoring.

Keywords

Graph Neural Network; Internet of Vehicles; following behavior; topological modeling; spatiotemporal graph

References

1. [1] Li Jiaxing Reinforcement learning based scheduling method for space-based data relay tasks in the Internet of Vehicles [D]. Changsha: Central South University, 2023
2. [2] Ge Yumu Multi perspective vehicle collision prediction model based on adaptive convolution [D]. Shanghai: Shanghai University of Finance and Economics, 2022
3. [3] Zou Zhihong Quantitative Study on the Economic Efficiency of Commercial Vehicle Driver Driving Behavior Based on Connected Vehicle Data [D]. Guilin: Guilin University of Electronic Science and Technology, 2022
4. [4] Zhang Yanyu, Zhang Zhiming, Liu Chunyang, etc Electric Vehicle Charging Load Prediction Based on Dynamic Adaptive Graph Neural Network [J]. Power System Automation, 2024, 48 (7): 86-93
5. [5] Liao Yong, Yin Zisong, Tian Xiaoyi GNN based SC-FDMA intelligent channel estimation in V2I scenarios of connected vehicles [J]. Acta Sinica, 2024, 52 (3): 772-782
6. [6] Jia Yiwei, Qi Yong, Wu Lanfen Research on the Integration and Evolution of Artificial Intelligence and Connected Vehicle Technology from a Patent Perspective [J]. Technological Progress and Countermeasures, 2022, 39 (22): 20-29
7. [7] Zhang Rui Intrusion Detection Model Based on Convolutional Neural Network for Connected Vehicles [J]. Telecommunication Science, 2024, 40 (12): 51-62
8. [8] Li Ya, Zhang Yu Interrupt Performance Prediction of Collaborative Connected Vehicles Based on GRNN [J]. Wireless Communication, 2023, 13 (1): 1-11
9. [9] Lv Xinchen, Zhang Chenyu Edge intelligent multi-source data processing for vehicle networking autonomous driving [J]. Journal of Beijing University of Posts and Telecommunications, 2021, 44 (2): 102-108
10. [10] Wen Huiying, Zhang Weigang, Zhao Sheng Vehicle lane changing trajectory prediction model based on generative adversarial network [J]. Journal of South China University of Technology (Natural Science Edition), 2020, 48 (5): 32-40

About the authors

Lin Jianmin (January 1983-), male, Han ethnicity, from Shaoxing, Zhejiang Province, is a general manager, artificial intelligence engineer, and a master's student with research interests in new materials and artificial intelligence.