AUTHOR=Zhang Chengsi , Czarnuch Stephen 

TITLE=Point cloud completion in challenging indoor scenarios with human motion

JOURNAL=Frontiers in Robotics and AI

VOLUME=Volume 10 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/robotics-and-ai/articles/10.3389/frobt.2023.1184614

DOI=10.3389/frobt.2023.1184614

ISSN=2296-9144

ABSTRACT=Combining and completing point cloud data from two or more sensors with arbitrarily relative perspectives in a dynamic, cluttered and complex environment is challenging, especially when the two sensors have significant perspective difference while the large overlap ratio and feature rich scene cannot be guaranteed. We are creating a novel approach that is targeting this challenge scenario by registering two cameras’ captures in time series with unknown perspectives and human movements so that our system can be easily used in real life scene. In our approach, we first reducing the 6 unknowns of 3D point cloud completion to 3 unknowns by aligning ground planes which is found by our previous perspective independent 3D ground plane estimation algorithm. Next, we use histogram based approach to identify and extract all the humans from each frame generating 3D human walking sequence in time series. In order to enhance the accuracy and performance, we then convert 3D human walking sequences to lines by calculating Center of Mass point of each human body and connecting them together. Finally, we match the walking paths in different data trials by minimizing the Frechet distance between two walking paths, and using 2D iterative closest point to find the rest 3 unknowns in the overall transformation matrix for the final alignment. By using this approach, we are able to successfully register the corresponding walking path of the human between the two cameras’ captures and estimate the transformation matrix between two sensors