Abstract:Traffic Surveillance Systems (TSS) have become increasingly crucial in modern intelligent transportation systems, with vision-based technologies playing a central role for scene perception and understanding. While existing surveys typically focus on isolated aspects of TSS, a comprehensive analysis bridging low-level and high-level perception tasks, particularly considering emerging technologies, remains lacking. This paper presents a systematic review of vision-based technologies in TSS, examining both low-level perception tasks (object detection, classification, and tracking) and high-level perception applications (parameter estimation, anomaly detection, and behavior understanding). Specifically, we first provide a detailed methodological categorization and comprehensive performance evaluation for each task. Our investigation reveals five fundamental limitations in current TSS: perceptual data degradation in complex scenarios, data-driven learning constraints, semantic understanding gaps, sensing coverage limitations and computational resource demands. To address these challenges, we systematically analyze five categories of potential solutions: advanced perception enhancement, efficient learning paradigms, knowledge-enhanced understanding, cooperative sensing frameworks and efficient computing frameworks. Furthermore, we evaluate the transformative potential of foundation models in TSS, demonstrating their unique capabilities in zero-shot learning, semantic understanding, and scene generation. This review provides a unified framework bridging low-level and high-level perception tasks, systematically analyzes current limitations and solutions, and presents a structured roadmap for integrating emerging technologies, particularly foundation models, to enhance TSS capabilities.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is that in traffic surveillance systems (TSS), existing visual technology research usually only focuses on isolated aspects, lacking a comprehensive analysis between low - level perception tasks (such as object detection, classification, and tracking) and high - level perception applications (such as parameter estimation, anomaly detection, and behavior understanding), especially without fully considering the impact of emerging technologies. In addition, current research often lacks a detailed analysis of methodologies in task categories and an exploration of the revolutionary potential of fundamental models (i.e., large - scale models) in high - level perception tasks.
Specifically, the main contributions of the paper include:
1. **Providing a systematic review**: A systematic review of the application of vision - based technologies in TSS was carried out, which was divided into low - level and high - level perception tasks, and a detailed classification and performance analysis of the methodologies in each category was conducted.
2. **Identifying existing limitations**: By analyzing the limitations of current TSS technologies and applications, a systematic development roadmap was proposed, key challenges were pointed out, and specific technical innovation suggestions were put forward, providing practical guidance for researchers and practitioners.
3. **Exploring fundamental models**: The application of fundamental models in traffic perception was deeply studied, and their unique capabilities (such as zero - shot learning, semantic understanding, and scene generation) and their transformative potential in promoting TSS applications were analyzed.
### Overview of the paper structure
1. **Introduction**: The importance of TSS in intelligent transportation systems and the fundamental role of visual technologies in it were introduced.
2. **Low - level traffic perception tasks**:
- **Detection**: Including 2D and 3D detection, the evolution and classification of mainstream detection methods were introduced in detail.
- **Classification**: Including vehicle model recognition and vehicle re - identification (Re - ID), the progress of hand - crafted features and deep - learning methods was discussed.
- **Tracking**: Including single - object tracking (SOT) and multi - object tracking (MOT), the development and performance of related methods were analyzed.
3. **High - level traffic perception tasks**:
- **Parameter estimation**: Such as camera calibration, speed estimation, and vehicle counting.
- **Anomaly detection**: Covering weakly - supervised and unsupervised methods.
- **Behavior understanding**: Including vehicle behavior recognition, trajectory prediction, and intention prediction.
4. **Limitations analysis and future prospects**:
- **Limitations overview**: Five main limitations of current visual technologies in TSS were discussed, including perception data degradation, data - driven learning limitations, semantic understanding gaps, sensing coverage limitations, and computational resource requirements.
- **Potential solutions**: Five categories of potential solutions were proposed, including high - level perception enhancement, efficient learning paradigms, knowledge - enhanced understanding, collaborative perception frameworks, and efficient computing frameworks.
- **Fundamental model prospects**: The unique capabilities and transformative potential of fundamental models in TSS were explored, such as zero - shot learning, open - vocabulary detection, visual question answering, multi - modal complementarity, and physical - scene reasoning.
### Conclusion
Through a comprehensive analysis of visual technologies in TSS, this paper not only fills the gaps in existing research but also provides a clear direction and practical suggestions for future technological development. In particular, the paper emphasizes the great potential of fundamental models in improving TSS performance, providing an important reference for future scientific research and applications.