Automatic in-situ radiometric calibration for mobile laser scanning: Compensating for distance and angle of incidence effects

Abstract. Laser scanner intensity data provide valuable insights into material properties, enabling applications such as point cloud segmentation and material probing. However, extracting meaningful information is challenging due to the influence of the measurement configuration represented by the angle of incidence (AOI) and distance. Although existing methods for radiometric calibration in terrestrial laser scanning (TLS) exist, they rely on either overlapping scans from discrete positions or on manual segmentation. This limits their applicability to mobile laser scanning (MLS), which typically produces very large datasets (requiring automation) with little or no overlap, and from continuously changing positions. This study presents an approach for adapting an automatic in-situ radiometric calibration method originally developed for TLS that applies to MLS. Building on our previous work, we introduce techniques to estimate AOI and distance influence compensation functions with little or without overlap, as well as non-discrete scan stations, and propose two strategies for AOI influence compensation - global and local. The global method computes one best-fitting AOI compensation function for the entire scan. It uses local reflectance estimation, which relies on a modified filtering technique, accommodating the unique characteristics of MLS data. The local method computes the best-fitting AOI compensation function per segment, ideally containing a single material. We use machine learning for point cloud semantic segmentation with additional instance segmentation to automatically obtain a material proxy segment. We evaluate the proposed methods on four datasets captured by two different MLS systems, demonstrating their ability to reduce measurement configuration related influences on intensities and enhance following point cloud segmentation.