Spatial and Distance Discrepancies in Target-Based and Visual-Aligned LiDAR Models

Today, static, terrestrial, Light Detection and Ranging (LiDAR) instruments are becoming more affordable and ubiquitous in numerous professional areas, including civil engineering, construction, architecture, surveying, archeological sites, and manufacturing, to name a few. Individual applications often require a greater or lesser level of accuracy, and many can present tolerance constraints that may rule out some available methodologies. Since most applications require to stitch (register) multiple scans into a common coordinate system, it is critical to examine various existing popular registration approaches to form expectations of the accuracies they will yield. This work studies the relative accuracy of two final point-cloud models of the same spatial conditions. Each model uses a different scan-stitching approach, the Target-Based (TB) registration and the Visual- Aligned (VA) registration. For comparison purposes, both models are georeference into the same State-Plane coordinate system. Point positions and distances, extracted from both final LiDAR models, are compared against the same positions and distances acquired in the field by an accurate Robotic Total-Station (RTS) instrument. This article presents those discrepancies and concludes that the TB registration is, approximately, 3 to 4 times more accurate than the VA one, when modeling a 3.5-acre site, via a closed scanning loop, under the conditions described in this project.