From Propagation to Prediction: Point-level Uncertainty Evaluation of MLS Point Clouds under Limited Ground Truth

TL;DR

The paper tackles the costly ground-truth limitation in evaluating MLS point-cloud uncertainty by introducing a learning-based framework that maps local geometric features to point-level uncertainty quantified by the cloud-to-cloud distance $C2C$. It integrates optimal neighborhood estimation with 26 geometric features plus the neighborhood size $OptN$, and compares RF and XGBoost on a real-world indoor dataset, finding that XGBoost achieves comparable accuracy to RF but with about three times faster runtime. Feature-importance analyses via SHAP and permutation measures reveal height variation, density, and roughness as primary drivers of uncertainty, offering interpretable links between geometry and error. The results suggest that MLS uncertainty is learnable and that the proposed approach provides a practical, flexible alternative to GT-heavy evaluation, with broad implications for uncertainty-aware MLS workflows in AEC and related fields.

Abstract

Evaluating uncertainty is critical for reliable use of Mobile Laser Scanning (MLS) point clouds in many high-precision applications such as Scan-to-BIM, deformation analysis, and 3D modeling. However, obtaining the ground truth (GT) for evaluation is often costly and infeasible in many real-world applications. To reduce this long-standing reliance on GT in uncertainty evaluation research, this study presents a learning-based framework for MLS point clouds that integrates optimal neighborhood estimation with geometric feature extraction. Experiments on a real-world dataset show that the proposed framework is feasible and the XGBoost model delivers fully comparable accuracy to Random Forest while achieving substantially higher efficiency (about 3 times faster), providing initial evidence that geometric features can be used to predict point-level uncertainty quantified by the C2C distance. In summary, this study shows that MLS point clouds' uncertainty is learnable, offering a novel learning-based viewpoint towards uncertainty evaluation research.

Figures (8)

• Figure 1: Proposed learning-based point-level uncertainty evaluation framework.
• Figure 2: Overview of the scanned area. Red line indicates the scanning path.
• Figure 3: Predicted results from XGBoost model.
• Figure 4: Detailed comparison of XGBoost prediction on selected objects.
• Figure 5: Absolute error and residual of the predicted C2C distance with respect to the GT from XGBoost model.