Seminar: DeReEs: Real-Time Registration of RGBD Images Using Image-Based Feature Detection And Robu

Sahand Seifimamaghani
M.Sc. Candidate

Supervisor: Dr. Oscar Meruvia-Pastor

DeReEs: Real-Time Registration of RGBD Images Using Image-Based Feature Detection And Robust 3D Correspondence Estimation and Refinement

Department of Computer Science
Tuesday, April 21, 2015, 11:00 a.m., Room EN 2022


Abstract

We present DeReEs, a real-time RGDB registration algorithm for the scenario where multiple RGBD images of the same scene are obtained from Depth-sensing Cameras placed at different viewpoints, but with partial overlap between images. DeReEs consists of a sequence of Detection, Rejection and Estimation steps, as a combination of 2D image-based feature detection algorithms and a false-pair rejection algorithm applied on 3D point clouds based on RANSAC. DeReEs not only perform in real-time, but also supports large transformation distances for both translations and rotations. We use DeRes for scene reconstruction by merging the RGDB images after registration, but DeReEs might also be useful for indoor localization, 3D object scanning and other tasks which involve registration of RGBD images. We present comparisons of DeReEs with other registration algorithms. Our results suggest that DeReEs provides better performance and accuracy.

 

Contact

Department of Computer Science

230 Elizabeth Ave

St. John's, NL A1B 3X9 CANADA

Tel: (709) 864-2530

Fax: (709) 864-2552

becomestudent@mun.ca