Stereo (Multi-view) DepthIQ and its Application in SLAM

Our partner organization has developed a new type of camera, one where depth information is part of a single lens/sensor construction, meaning that there are no occlusions or shadows which plague stereo cameras, there is no requirement to light the scene as well time-of-flight cameras, and the expense is low compared to light-field cameras. Currently these cameras are limited to a near-range depth field of around 2m, this project aims to improve through several methods. We shall work with our partner organization to develop a stereo setup (including a calibration process) that will greatly extend that range, and then incorporate a deep learning method that will be able to use the actual stereo method to infer the stereo back from a single camera. To validate this process we shall develop a Simultaneous Location and Mapping (SLAM) method using both cameras and determine the errors produced.

Faculty Supervisor:

Marzieh Amini;Chris Joslin

Student:

Partner:

Airy3D Inc

Discipline:

Engineering

Sector:

Professional, scientific and technical services

University:

Carleton University

Program: