Student research opportunities

3D Object Detection and Classification in High Quality Geo-referenced LIDAR and Vision Data

Project Code: CECS_871

This project is available at the following levels:
Engn4200, Engn R&D, Honours, Summer Scholar
Please note that this project is only for undergraduate students.

Keywords:

Computer Vision, LIDAR, Mapping, Machine Learning

Supervisors:

Dr Lars Petersson
Dr Gary Overett

Outline:

Are you a talented Engineer or Software Engineer looking for a cool Honours project?
Are you interested in working with Terabytes of Real World Data collected using cutting edge LIDAR and Imaging technology?
Do you think a pool of NVidia GTX590's are useful for more than just gaming?
Would you like to practise SCRUM/Agile Software Engineering best practises?
Does the idea of having your work used in real world business applications or presented internationally excite you?
Are you considering an R&D career or PhD Research at ANU?
Do you have strong mathematical and analytic skills?

If you answered yes to several of the above questions you may be interested in pursuing an Honours Project within NICTA's Continent Scale Scene Understanding Project. Participating students will be provided with access to powerful computing resources, access to rich and large scale real world datasets, working space at NICTA Canberra Research Laboratory located adjacent to the ANU Campus, and access to top academics, engineers and researchers.

Goals of this project

This project ultimately has many facets but the main goal is to discover new ways of detecting and classifying objects of interest within the 3D data. This may involve work in finding features which are well suited to particular detections problems or creating datatypes which allow for faster search and processing of this very large dataset.

Requirements/Prerequisites

Success in this project will (depending on the problem you tackle and how you proceed) likely require a fair amount of programming in C++ and possibly Matlab. Unfortunately, we cannot in the time allowed first teach programming and then tackle an interesting problem so we are primarily seeking students with some significant programming experience or demonstrated ability.

This is a more practical than theoretical project however, you will need to have some analytical and mathematical skills to traverse certain problems.

Student Gain

- Chance to work with rich high value image and LIDAR sensor data
- Can explore the possibilities of a research career in academia or an R&D career in business
- Possibility of seeing their work commercialised as part of the wider project goals
- Gain valuable experience on challenging real-world problems
- Experience working with SCRUM/Agile software development practises

Background Literature

If you are interested take a look at all the cool stuff already happening with the PointCloud library (www.pointclouds.org). They have some tutorials there that give some idea of the programming level required. The related web links below may give some additional information.