Search papers, labs, and topics across Lattice.
This paper introduces a topologically stable Hough Transform for line detection in point clouds, replacing the discrete voting scheme with a continuous score function. By analyzing the persistent homology of this score function, the method identifies candidate lines that are robust to noise and parameter variations. The authors provide an efficient algorithm for computing these topologically-derived candidate lines, demonstrating a more stable and reliable approach to line detection compared to the classical Hough transform.
Ditch the brittle Hough Transform: persistent homology offers a noise-resistant way to detect lines in point clouds.
We propose an alternative formulation of the well-known Hough transform to detect lines in point clouds. Replacing the discretized voting scheme of the classical Hough transform by a continuous score function, its persistent features in the sense of persistent homology give a set of candidate lines. We also devise and implement an algorithm to efficiently compute these candidate lines.
