A FEASIBILITY STUDY ON USE OF GENERIC MOBILE LASER SCANNING SYSTEM FOR DETECTING ASPHALT PAVEMENT CRACKS

Xinqu Chen, Jonathan Li2016 · DOI

Abstract This study explores the use of 3D point clouds from generic Mobile Laser Scanning (MLS) systems for automated pavement crack detection. Moving away from traditional intensity-based methods, the proposed four-step pipeline—ground point filtering, high-pass convolution, matched filtering, and noise removal—utilizes elevation data to identify moderate-to-severe cracks. Results from a 200m urban road segment in Kingston, Ontario, demonstrate the effectiveness of the approach in detecting longitudinal, transverse, random, and alligator cracks, despite hardware resolution limitations for minor distresses.

Introduction

Maintaining aging transportation infrastructure is a significant financial and safety challenge, with billions spent annually on pavement upkeep. Traditional inspection methods often rely on manual "boots-on-the-ground" surveys, which are subjective, slow, and potentially hazardous for workers. While automated image-based systems have been developed, they frequently struggle with varying lighting conditions and shadows.

Mobile Laser Scanning (MLS) offers a robust 3D alternative. However, most research in this area has focused on high-end, specialized equipment. This study investigates whether a generic MLS system, such as the RIEGL VMX-450, can provide sufficient detail to detect cracks by leveraging the physical "dip" in elevation rather than relying solely on the reflective intensity of the crack.

Methodology

The researchers developed a computationally efficient, four-step automated detection pipeline designed to handle large-volume, mixed-density, unstructured MLS point clouds.

1. Ground Point Filtering

The raw point cloud is first segmented into blocks. A voxel-based upward growing method is applied to distinguish the road surface from surrounding objects like vehicles, trees, and curbs. This process results in a clean Digital Terrain Model (DTM) with a 4 cm grid cell resolution.

2. High-Pass Convolution

A 5x5 high-pass filter is applied to the DTM. Because healthy pavement is relatively smooth, this filter amplifies sudden, local elevation jumps—the geometric signature of a physical crack.

3. Matched Filtering

To delineate crack features, the study adapts a technique from medical imaging originally used to identify blood vessels in retinas. Twelve Gaussian-shaped kernels, rotated at 15° intervals, are used to "match" and extract crack features regardless of their orientation on the road surface.

4. Noise Removal

The final stage involves filtering out non-crack objects. Detected features that do not meet specific geometric thresholds for length or total area are removed to refine the final output.

Experimental Data and Results

The study utilized a real-world dataset collected from a 200-meter segment of Front Road in Kingston, Ontario, Canada. The RIEGL VMX-450 system collected over 25.5 million points with an average point spacing of 3.4 cm.

Key Findings

  • Elevation vs. Intensity: The study confirms that elevation jumps are a viable and robust metric for detecting physical distress, offering an alternative to intensity-based LiDAR detection which can be affected by surface moisture or material changes.
  • Versatility: The method successfully identified four major crack types: longitudinal, transverse, random, and alligator cracks.
  • Severity Detection: The system is highly effective at detecting cracks in the moderate-to-severe range (13 mm to 25 mm wide). Interestingly, the DTM convolution can detect some cracks narrower than the 4 cm grid size if they have sufficient length to be captured by multiple laser pulses.
  • Efficiency: The matched filtering process is highly optimized, processing a 2500 x 760 image in less than 10 seconds, outperforming many state-of-the-art methods like tensor voting.

Limitations and Future Outlook

The primary limitation of this approach is the hardware's resolution floor. "Slight" cracks (3 mm to 13 mm) often remain unclear because the 3.4 cm point spacing is too coarse to capture them reliably. Additionally, the high-pass filter can occasionally misidentify manholes, curbs, or thick road markings as cracks due to their sharp elevation changes.

Future work will likely benefit from next-generation MLS systems with higher point densities. Combining elevation data with intensity data also holds promise for reducing false positives and improving the detection of minor pavement distresses.

Summary updated

Abstract

This study aims to automatically detect pavement cracks on urban roads by employing the 3D point clouds acquired by a mobile laser scanning (MLS) system. Our method consists of four steps: ground point filtering, high-pass convolution, matched filtering, and noise removal. First, a voxel-based upward growing method is applied to construct Digital Terrain Model (DTM) of the road surface. Then, a high-pass filter convolutes the DTM to detect local elevation changes that may embed cracking information. Next, a two-step matched filter is applied to extract crack features. Lastly, a noise removal process is conducted to refine the results. Instead of using MLS intensity, this study takes advantages of the MLS elevation information to perform automated crack detection from large-volume, mixed-density, unstructured MLS point clouds. Four types of cracks including longitudinal, transvers, random, and alligator cracks are detected. Our results demonstrated that the proposed method works well with the RIEGL VMX-450 point clouds and can detect cracks in moderate-to-severe severity (13 - 25 mm) within a 200 m by 30 m urban road segment located in Kingston, Ontario, at one time. Due to the resolution capability, small cracks with slight severity remain unclear in the MLS point cloud.
BibTeX
@inproceedings{chen2016feasibility,
  author = {Xinqu Chen and Jonathan Li},
  title = {{A FEASIBILITY STUDY ON USE OF GENERIC MOBILE LASER SCANNING SYSTEM FOR DETECTING ASPHALT PAVEMENT CRACKS}},
  booktitle = {˜The},
  year = {2016},
  doi = {10.5194/isprsarchives-xli-b1-545-2016},
}
Figure