ICP Registration Based on Overlapping Region

To solve the problem that ICP(Iterative Closest Point), a widely used fine registration algorithm, can degrade good initial registration results, this paper introduces a step of overlapping region extraction between the initial registration and fine registration to further enhance the accuracy of ICP. Furthermore, a method for determining overlapping regions based on grid partitioning is proposed, which resolves the issue of longer consumption of time for overlapping regions in large-scale point clouds. Experimental results indicate that the improved algorithm process reduces angular errors by approximately 50% and translational errors by approximately 60% compared to the pre-improvement process.