AUTOMATED VISIBILITY FIELD EVALUATION OF TRAFFIC SIGN BASED ON 3D LIDAR POINT CLOUDS

Shuoxun Zhang, Cheng Wang, Ming Cheng, Jonathan Li2019 · DOI

Abstract

Maintaining high visibility of traffic signs is critical for road safety. Manual inspection and removal of occlusions are labor-intensive tasks for traffic management departments. This paper presents an automated method to detect occlusions and quantitatively estimate traffic sign visibility across entire road surfaces using Mobile Laser Scanning (MLS) systems. We introduce the concept of a "visibility field" and utilize a retinal imaging area model to align the evaluation with human physiological vision. Validation using 2D-3D registration demonstrates high consistency between point cloud occlusion ratios and real-world photographs, proving the method's efficiency in large-scale environments.

Introduction

Traffic signs are vital components of road infrastructure, yet they are often obscured by vegetation or urban development. Traditional manual inspections are subjective and costly. While 2D image-based or simulator-based methods exist, they often fail to capture the continuous, 3D complexity of real-world driving environments. This research addresses these gaps by proposing a continuous visibility evaluation framework based on high-fidelity 3D LiDAR data.

Methodology

The proposed framework transitions from simple geometric area measurements to a human-centric vision model.

  1. Data Acquisition: Mobile Laser Scanning (MLS) captures dense 3D point clouds of the environment alongside the vehicle's precise trajectory.
  2. Retinal Imaging Model: To simulate human perception, 3D sign points are projected onto a virtual "retina" positioned 17mm from a virtual eye. This accounts for distance and viewing angle, ensuring that a sign's "size" in the model reflects its actual visibility to a driver.
  3. Occlusion Detection (The Ray Method): Virtual rays are cast from various road viewpoints toward the sign. Points that intercept these rays but do not belong to the sign (e.g., tree branches) are identified as occlusion points.
  4. The Visibility Field: By calculating visibility scores from hundreds of viewpoints along a trajectory, the system generates a "Visibility Field"—a spatial distribution representing how well a sign is seen from different locations on the road.
3D visualization of the visibility field concept, showing viewpoints relative to a traffic sign.

Experimental Validation

The method was validated using MLS data and 144 high-resolution photos from Xiamen, China. Each photo was paired with centimeter-level GPS coordinates to ensure the 3D viewpoint matched the camera's real-world position.

Validation of occlusion detection by comparing 3D LiDAR point clouds with registered 2D photographs.

The algorithm achieved an average occlusion ratio accuracy of 94.89% compared to manually segmented ground truth photos. As shown in the figure above, the 3D-generated boundaries for signs (yellow) and occlusions (red) align closely with the visual evidence in the photographs.

Key Insights and Results

  • Human-Centric Metric: Using the retinal imaging area acknowledges that distance and orientation are as important as physical size for visibility.
  • Continuous Evaluation: The "Visibility Field" allows authorities to pinpoint exactly where a sign becomes obscured, rather than relying on a single snapshot.
  • Robustness: The system successfully handled complex, irregular occlusions (like tree branches) in winding mountain road environments.

Conclusion and Future Work

This automated approach provides a scalable, objective tool for traffic safety management. Future improvements will focus on integrating environmental factors like weather (fog, rain) and lighting conditions, as well as accounting for driver-specific cognitive loads.

Summary updated

Abstract

Maintaining high visibility of traffic signs is very important for traffic safety. Manual inspection and removal of occlusion in front of traffic signs is one of the daily tasks of the traffic management department. This paper presents a method that can automatically detect the occlusion and continuously quantitative estimate the visibility of traffic sign cover all the road surface based on Mobile Laser Scanning (MLS) systems. The concept of traffic sign’s visibility field is proposed in this paper. One of important innovation of this paper is that we use retinal imaging area to evaluate the visibility of a traffic sign. And this makes our method is in line with human vision. To validate the reasonable and accuracy of our method, we use the 2D and 3D registration technology to observe the consistence of the occlusion ratio in point clouds with it in photo. Experiment of implementation on large scale traffic environments show that our method is feasible and efficient.
BibTeX
@inproceedings{zhang2019automated,
  author = {Shuoxun Zhang and Cheng Wang and Ming Cheng and Jonathan Li},
  title = {{AUTOMATED VISIBILITY FIELD EVALUATION OF TRAFFIC SIGN BASED ON 3D LIDAR POINT CLOUDS}},
  booktitle = {˜The},
  year = {2019},
  doi = {10.5194/isprs-archives-xlii-2-w13-1185-2019},
}
Figure