You are here


Project Lifetime: 
Aug 2009

Description: This project aims to improve the accuracy and processing speed for indoor 3D mapping using RGB-D data to produce a clean, light-weight, photo-realistic representation of indoor scenes of varying sizes, in real time.

We developed the RGB-D sensor depth uncertainty model, a new fast visual odometry pipeline, place recognition and global alignment agorithm, new data structure for scene reconstruction, and implemented the algorithms using Hokoyu Laser on board the MAV as well as the Google Tango devices.

Sponsors: This project is supported in part by Army Research Office under Grant No. W911NF0910565 and Google.