DefSLAM: Tracking and Mapping of Deforming Scenes From Monocular Sequences

Monocular simultaneous localization and mapping (SLAM) algorithms perform robustly when observing rigid scenes; however, they fail the observed scene deforms, for example, in medical endoscopy applications. In this article, we present DefSLAM, first monocular SLAM capable of operating deforming scenes real time. Our approach intertwines Shape-from-Template (SfT) Non-Rigid Structure-from-Motion (NRSfM) techniques to deal with exploratory sequences typical SLAM. A deformation tracking thread recovers pose camera map, at frame rate, by means SfT processing a template that models shape-at-rest. runs parallel update template, keyframe an isometric NRSfM batch full perspective keyframes. our experiments, DefSLAM processes close-up scenes, both laboratory-controlled experiment sequences, producing accurate 3-D respect moving camera.