ORTHOGRAPHIC REFLECTANCE IMAGE FOR PLANAR TARGET LOCALIZATION IN LOW DENSITY TLS POINT CLOUDS

Dazhou Guo, Deping Yu, Yubin Liang, Chen Feng2019 · DOI

Abstract

Point cloud registration is an important and essential task for terrestrial laser scanning applications. Point clouds acquired at different positions exhibit significant variation in point density. Most registration methods implicitly assume dense and uniform distributed point clouds, which is hardly the case in large-scale surveying. The accuracy and robustness of feature extraction are greatly influenced by the point density, which undermines the feature-based registration methods. We show that the accuracy and robustness of target localization dramatically decline with decreasing point density. A methodology for localization of artificial planar targets in low density point clouds is presented. An orthographic image of the target is firstly generated and the potential position of the target center is interactively selected. Then the 3D position of the target center is estimated by a non-linear least squares adjustment. The presented methodology enables millimeter level accuracy of target localization in point clouds with 30mm sample interval. The robustness and effectiveness of the methodology is demonstrated by the experimental results.

Introduction

Terrestrial Laser Scanning (TLS) is a staple for 3D reconstruction in engineering and archaeology. To create a complete model, multiple scans from different positions must be "registered" (aligned) into a single coordinate system. This often relies on artificial targets—like black-and-white paper squares—placed throughout the scene. The "achilles heel" of this process is point density. As a scanner gets further from a target or the angle becomes oblique, the number of points hitting that target drops significantly. Most registration algorithms and commercial tools (like Leica Cyclone) assume dense, uniform data. When data becomes sparse, feature extraction becomes unstable, leading to poor alignment or total failure to recognize targets.

Methodology

The proposed methodology overcomes the limitations of low-density data through a two-stage refinement process.

Orthographic Reflectance Image Generation

First, Principal Component Analysis (PCA) is performed on the target's point cloud to identify the target's plane. The 3D points are then projected onto this 2D plane to create a grayscale "ortho-image." Crucially, this process utilizes the reflectance (intensity) values of the laser returns. As seen in the figure below, the visual pattern of the black-and-white target remains recognizable even when the geometric points are sparse.

Low-density reflectance image of a planar target showing the discernible checkerboard pattern despite sparse data.

Non-linear Least Squares Adjustment

After a user interactively selects a potential center on the ortho-image, the 3D coordinates are refined using a Gauss-Newton iterative approach. The model minimizes the 3D Euclidean distance between the target center and its neighboring points. By linearizing this distance function using a Taylor series expansion, the solution iteratively converges to a precise 3D coordinate.

Experimental Setup and Results

The researchers established a controlled test field using a RIEGL VZ-1000 laser scanner and 30 planar targets uniformly distributed on a hallway wall. A high-density ground truth scan of nearly 40 million points was processed in Leica Cyclone 6.0.3 for comparison.

Experimental setup showing the RIEGL VZ-1000 scanner and the distribution of targets in the test hallway.

To simulate low-density conditions, the original point cloud was decimated using 3D gridding at sample intervals ranging from 5mm to 45mm. The results highlighted a clear "failure cliff" for commercial software. While Leica Cyclone performed well at high densities (5-15mm), its performance dropped sharply beyond 20mm. At a 30mm sample interval, Cyclone could only localize 2 out of 30 targets. In contrast, the proposed method successfully localized all 30 targets with millimeter-level accuracy (RMSE of approximately 3-4mm across X, Y, and Z axes).

Conclusion

This study demonstrates that reflectance data provides a robust signal for target localization when geometric density is insufficient for standard algorithms. While the current workflow requires a human-in-the-loop for initial center selection, it offers a significant improvement in reliability for large-scale surveys. Future work will focus on automating the initial center selection and adapting the methodology for other types of features in low-density environments.

Summary updated

Abstract

Point cloud registration is important and essential task for terrestrial laser scanning applications. Point clouds acquired at different positions exhibit significant variation in point density. Most registration methods implicitly assume dense and uniform distributed point clouds, which is hardly the case in large-scale surveying. The accuracy and robustness of feature extraction are greatly influenced by the point density, which undermines the feature-based registration methods. We show that the accuracy and robustness of target localization dramatically decline with decreasing point density. A methodology for localization of artificial planar targets in low density point clouds is presented. An orthographic image of the target is firstly generated and the potential position of the target center is interactively selected. Then the 3D position of the target center is estimated by a non-linear least squares adjustment. The presented methodology enables millimeter level accuracy of target localization in point clouds with 30mm sample interval. The robustness and effectiveness of the methodology is demonstrated by the experimental results.
BibTeX
@inproceedings{guo2019orthographic,
  author = {Dazhou Guo and Deping Yu and Yubin Liang and Chen Feng},
  title = {{ORTHOGRAPHIC REFLECTANCE IMAGE FOR PLANAR TARGET LOCALIZATION IN LOW DENSITY TLS POINT CLOUDS}},
  booktitle = {˜The},
  year = {2019},
  doi = {10.5194/isprs-archives-xlii-2-w13-995-2019},
}
Figure