Mobile mapping systems camera–LiDAR data registration for mitigating GNSS/INS trajectory perturbations

Abstract

Wheeled Mobile Mapping Systems (MMS), equipped with LiDAR, cameras, and integrated GNSS/INS units, are widely used in urban planning, map generation, and infrastructure monitoring. Accurate registration between wheeled MMS camera and LiDAR data, relying on precise system calibration and GNSS/INS trajectory, is crucial for effective data fusion to address the needs of these applications. However, environmental factors can degrade sensor calibration and GNSS/INS accuracy, leading to misalignment between imagery and LiDAR data. Calibration parameters, such as mounting parameters and sensors’ Interior Orientation Parameters (IOP), can be affected by sensor aging and environmental conditions, while data collection along transportation corridors may suffer from GNSS signal occlusions due to interference from traffic, bridges, and buildings. GNSS/INS trajectory errors are more frequent than calibration errors. This research addresses these trajectory issues by analyzing image-LiDAR misalignments and proposes a novel registration approach. The method establishes an appropriate transformation function, automatically extracts lane markings as common primitives, and develops a similarity measure tailored to these primitives. These elements are integrated into an automated optimization strategy that estimates transformation function parameters. The proposed learning-based algorithm is effective in both urban and highway environments, offering a robust solution for camera-LiDAR alignment. Additionally, an analysis of stereo camera poses before and after registration identifies misalignment causes, whether due to GNSS/INS errors or calibration inaccuracy. The proposed algorithm, evaluated using the mean of minimum Euclidean distances and Intersection over Union (IoU), demonstrates significant improvements, reducing misalignment to less than a few pixels and achieving IoU improvements exceeding 50%.