Innovative sensing technology enhances driverless car perception
Our Computer Science research is helping to put driverless cars on the road to reality.
Research by Professor Toby Breckon is paving the way on new technological development that allows cars to be driven autonomously on public roads.
Ground-breaking research
Professor Breckon’s research team is primarily focused on the use of automated image understanding techniques and addresses two key algorithmic tasks relating to how vehicles perceive the road environment upon which they are driving – “Where am I?” (known as the task of localisation) and “What is around me?” (known as the task of scene understanding).
Using stereo vision, a technique that uses a pair of cameras to measure object distances similarly to our own pair of eyes, the vehicle can readily estimate the distances of objects around it in both in an urban and rural setting.
Additionally, using recent advances in convolutional neural networks, his team have been able to enable the vehicle to both detect objects in the near vicinity such as pedestrians and other vehicles for obstacle avoidance whilst also recognising varying terrain types for off-road driving.
Leading AI technology
Furthermore, the team’s pioneering research enables the use of 360° panoramic imagery that offers both 3D object detection and highly detailed 3D scene depth t around the entire vehicle.
His team specifically target the use of low-cost camera sensors and aim achieve superior sensing capabilities via pioneering AI algorithms such that this technology can be made commercially available in affordable domestic vehicles.
Related research has led to major improvements in a car being able to accurately pinpoint its location and visualise surroundings under varying weather conditions.
Find out more
- Learn more about the work of Toby Breckon, Professor in our Department of Computer Science and Department of Engineering,
- Visit Professor Toby Breckon’s website.
- Read related journal articles:
- Multi-Modal Learning for Real-Time Automotive Semantic Foggy Scene Understanding via Domain Adaptation
- Eliminating the Blind Spot: Adapting 3D Object Detection and Monocular Depth Estimation to 360° Panoramic Imagery
- From On-Road to Off: Transfer Learning within a Deep Convolutional Neural Network for Segmentation and Classification of Off-Road Scenes
- Interested in studying at Durham? Explore our undergraduate and postgraduate courses in Department of Computer Science.