Design and research of bridge collision avoidance system based on camera calibration technology and motion detection

Area of Science:

• Computer Vision
• Artificial Intelligence
• Infrastructure Safety

Background:

• Bridge collisions, especially from over-height vehicles, endanger public infrastructure, economy, and safety.
• Existing systems lack the robustness to handle dynamic environments and complex collision scenarios.

Purpose of the Study:

• To develop and validate an intelligent, vision-based Bridge Collision Avoidance System (BCAS).
• To proactively detect and mitigate potential collisions using advanced algorithms and real-time risk assessment.

Main Methods:

• Utilized high-resolution video feeds with precise intrinsic/extrinsic camera calibration for 2D to 3D spatial transformation.
• Employed a hybrid motion detection and object segmentation approach (background subtraction, YOLOv11, Vision Transformers).
• Implemented a risk evaluation model based on spatial thresholds, velocity vectors, and confidence scores for real-time alerts via edge-cloud frameworks.

Main Results:

• Achieved high accuracy (95.7%) and a low false alarm rate (3.2%) in diverse conditions (occlusion, night, dense traffic).
• Demonstrated superior performance compared to traditional systems with an average response latency of 162 ms.
• Validated the system's effectiveness in real-world scenarios, including challenging environmental and traffic conditions.

Conclusions:

• The developed BCAS offers a modular, scalable, and fault-tolerant solution for enhancing bridge safety.
• The system provides a significant advancement in proactive collision avoidance for smart urban infrastructures.
• This research contributes a reliable method for mitigating risks associated with vehicle-bridge interactions.