Time-of-Flight Cameras in Space: Pose Estimation with Deep Learning Methodologies

摘要

Recently introduced 3D Time-of-Flight (ToF) cameras have shown a huge potential for mobile robotic applications, proposing a smart and fast technology that outputs 3D point clouds, lacking however in measurement precision and robustness. With the development of this low-cost sensing hardware, 3D perception gathers more and more importance in robotics as well as in many other fields, and object registration continues to gain momentum. Registration is a transformation estimation problem between a source and a target point clouds, seeking to find the transformation that best aligns them. This work aims at building a full pipeline, from data acquisition to transformation identification, to robustly detect known objects observed by a ToF camera within a short range, estimating their 6 degrees of freedom position. We focus this work to demonstrating the capability of detecting a part of a satellite floating in space, to support in-orbit servicing missions (e.g. for space debris removal). Experiments reveal that deep learning techniques can obtain higher accuracy and robustness w.r.t. classical methods, handling significant amount of noise while still keeping real-time performance and low complexity of the models themselves.