Enhanced Scan Registration Through Recursive ICP and Bias Filtering for Geoengineering Applications

Remote sensing–based model generation for geoengineering applications requires a robust computational framework to quantify uncertainty and evaluate thresholds of measurable intervals. The combination of partially mutually inclusive scans induces inherent and quantifiable uncertainty for engineering measurements. Iterative closest point (ICP) registration with recursive weighing of excessively far-off datapoints presents an acknowledged solution for scan registration. The resulting quality of conventional ICP is severely impacted by localized mutual exclusions (i.e., relative occlusions). The present work proposes a recursive ICP (R-ICP) followed by systematic excessive bias filtering loops to enhance the quality of scan registration. The method offers an intuitive and explicit perspective on quantifiable registration quality. The method is exemplified with a series of examples, followed by validation schemes to quantify the limitations of the proposed method.