Metrics for Aerial, Urban LiDAR Point Clouds

TL;DR

This paper proposes five new metrics for evaluating the density and accuracy of aerial urban LiDAR point clouds, focusing on vertical surfaces and the effects of flight pass overlap, to improve urban scene analysis.

Contribution

It introduces novel metrics for vertical surface density and accuracy, and analyzes their relevance in modern urban LiDAR data collection and processing.

Findings

Vertical surface density is influenced by sensor position and wall height.

Cross-pass registration significantly affects vertical surface error.

Vertical density improves with optimized flight missions, but accuracy may decrease.

Abstract

This paper introduces five new density and accuracy metrics for aerial point clouds that address the complexity and objectives of modern, dense laser scans of urban scenes. The five metrics describe (1) vertical surface density (points per area on vertical surfaces); (2) vertical density as a function of horizontal density; (3) vertical surface accuracy; and a decomposition of error into (4) within-pass and (5) cross-pass components. Specifically considered is vertical surface coverage and the practice of overlapping flight passes to reduce the occlusions and achieve the vertical density needed for twenty-first-century use cases (e.g. curb and window detection). The application of these metrics to a quartet of recent urban flyovers demonstrates their relevance by establishing (1) the efficacy of considering sensor position and wall height when predicting point density on vertical surfaces; (2) that cross-pass registration accounts for a disproportionate amount of the vertical surface error (but not horizontal) and provides a meaningful parameter to compare high-density, urban point clouds; and (3) that compared to horizontal density and accuracy, the vertical counterparts are disproportionately impacted (positively for density and negatively for accuracy) by modern, optimized flight missions.