Direct and Sparse Deformable Tracking

Deformable Monocular SLAM algorithms recover the localization of a camera in an unknown deformable environment. Current approaches use template-based tracking to pose and deformation map. These methods underlying global model. In this letter, we introduce novel method with local model for each point. Each map point is defined as single textured surfel that moves independently other points. Thanks direct photometric error cost function, can track position orientation without explicit our experiments, validate proposed system observe estimates more accurately targeted deformations both laboratory-controlled experiments in-body scenarios undergoing quasi-isometric deformations, changing topology or discontinuities.