Scan Registration
The process of aligning multiple individual laser scans into a single unified point cloud by identifying common reference points or overlapping geometry between scans.
Scan registration is the process of combining multiple individual laser scans — each captured from a different position — into a single unified coordinate system. Since each scan exists in its own local coordinate system (centred on the scanner position), they must be spatially aligned before they can be used as a single dataset.
Registration uses either target-based alignment (physical targets of known geometry — typically spheres or checkerboard-like flat targets — are placed in the overlap zone between adjacent scans and identified in each scan) or cloud-to-cloud alignment (algorithms like ICP — Iterative Closest Point — find the optimal transformation that minimises the distance between overlapping scan data).
Registration error — the residual misalignment after the best-fit transformation is applied — determines the global accuracy of the registered point cloud. Target-based registration typically achieves residuals of 2–5 mm; cloud-to-cloud registration 5–15 mm depending on overlap and surface complexity.