Research on Vision-Based Recognition Algorithm for Roadside Cross-Field Targets

DOI: https://doi.org/10.62517/jbdc.202601207

Author(s)

Liangdong Zuo1,2,*, Jia Liu1, Jie Li3, Hejia Li1

Affiliation(s)

1Chongqing College of Architecture And Technology, Chongqing, China 2Chongqing Research Institute of Shanghai Jiao Tong University , Chongqing, China 3Chongqing University of Science and Technology, Chongqing, China *Corresponding Author

Abstract

The realization of intelligent transportation systems and vehicle-infrastructure cooperation demands robust road-side perception capabilities. However, vision-based recognition algorithms face significant performance degradation when deployed across different domains—varying geographical locations, weather conditions, lighting environments, and traffic scenarios. This paper investigates recognition algorithms for roadside cross-field targets, addressing the fundamental challenge of domain generalization in visual perception systems. We systematically analyze the limitations of conventional detection methods in cross-domain scenarios, propose a novel framework integrating domain adaptation techniques with multi-scale feature extraction, and present experimental validations using diverse roadside datasets. Our approach achieves substantial improvements in cross-domain recognition accuracy while maintaining real-time performance requirements for roadside deployment. The research con-tributes to the foundational technology for large-scale implementation of vehicle-infrastructure cooperative systems.

Keywords

Cross-Field Target Recognition; Roadside Perception; Domain Adaptation; Computer Vision; Intelligent Transportation Systems

References

[1] Li, H., Zhang, R., Yao, H., et al. DA-Mamba: Learning Domain-Aware State Space Model for Global-Local Alignment in Domain Adaptive Object Detection. In Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2026, pp. 1–10. [2] Kay, J., Haucke, T., Stathatos, S., et al. Align and Distill: Unifying and Improving Domain Adaptive Object Detection. Trans. Mach. Learn. Res., 2025. [3] Danish, M. S., Khan, M. H., Munir, M. A., et al. Improving Single Domain-Generalized Object Detection: A Focus on Diversification and Alignment. In Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2025, pp. 11805–11814. [4] Zhang, Y., Tian, X., Li, J., et al. Robust Domain Adaptive Object Detection With Unified Multi-Granularity Alignment. IEEE Trans. Pattern Anal. Mach. Intell., vol. 46, no. 12, pp. 4102–4116, Dec. 2024. [5] Hao, Y., Forest, F., Fink, O. Simplifying Source-Free Domain Adaptation for Object Detection: Effective Self-Training Strategies and Performance Insights. In Proc. Eur. Conf. Comput. Vis. (ECCV), 2024, pp. 1–17. [6] Yao, H., Zhao, S., Li, P., et al. Beyond Boundaries: Leveraging Vision Foundation Models for Source-Free Object Detection. In Proc. AAAI Conf. Artif. Intell. (AAAI), 2026, pp. 1–9. [7] Liu, Y., Wang, J., Zhang, X., et al. SGV3D: Toward Scenario Generalization for Vision-Based Roadside 3D Object Detection. IEEE Trans. Intell. Transp. Syst., vol. 27, no. 1, pp. 389–402, Jan. 2026. [8] Chen, L., Wu, Y., Zhao, Z., et al. Cross-domain Multi-step Thinking: Zero-shot Fine-grained Traffic Sign Recognition in the Wild. IEEE Trans. Intell. Transp. Syst., vol. 26, no. 8, pp. 4567–4578, Aug. 2025. [9] Zhang, R., Lee, J., Cai, X., et al. Revisiting Cross-Domain Problem for LiDAR-based 3D Object Detection in Autonomous Driving. IEEE Robot. Autom. Lett., vol. 9, no. 2, pp. 1345–1352, Feb. 2024. [10] Kim, S., Park, J., Lee, S. Multi-Scale Adversarial Cross-Domain Vehicle Detection from UAV to Satellite Imagery. In Proc. IEEE Int. Geosci. Remote Sens. Symp. (IGARSS), 2024, pp. 7890–7893. [11] Wang, X., Li, Y., Zhang, S., et al. DATR: Unsupervised Domain Adaptive Detection Transformer with Dataset-Level Adaptation and Prototypical Alignment. IEEE Trans. Image Process., vol. 34, pp. 2345–2358, 2025. [12] Zhao, H., Sun, P., Wu, Z., et al. Multi-Scale Feature Alignment for Cross-Domain Roadside Object Detection. IEEE Intell. Transp. Syst. Mag., vol. 16, no. 3, pp. 112–123, 2024. [13] Liu, Z., Lin, Y., Cao, Y., et al. Swin-DA: Domain Adaptive Object Detection with Swin Transformer and Multi-Scale Adversarial Learning. Neurocomputing, vol. 589, pp. 127658, 2025.