Monocular 3D Object Reconstruction in Autonomous Driving

3D object reconstruction is a crucial task for autonomous driving. Many important fields in autonomous driving such as prediction, planning, and motion control generally require a faithful representation of the 3D space around the ego vehicle. Reliable and accurate 3D reconstruction with one RGB video alone in autonomous driving is the goal of this project. This project will present a reliable and accurate method for perspective projection reconstruction of rigid and non-rigid objects from single-view and realistic videos. This method will overcome all of the limitations arising from the usage of orthographic camera model and the complexity and non-linearity issues of the perspective projection equation. Unlike traditional structure-from-motion methods, which have been studied only on synthetic datasets and controlled lab environments that require some prior constraints or estimations (such as manually segmented objects, limited rotations and occlusions, and full-length trajectories); this method will be used in realistic video sequences.

Faculty Supervisor:

Shahryar Rahnamayan

Student:

Partner:

Xerxes Solutions Inc

Discipline:

Computer science

Sector:

Professional, scientific and technical services

University:

Brock University; University of Ontario Institute of Technology

Program: