A key problem in the management of a large rail network is the ability to maintain an accurate and up-to-date map of rail assets, such as signal boxes, signage and track junctions.
This project aims to develop a novel, flexible, compact, portable system for mapping assets using enhanced 3D imaging & visual analytics. This system will be sufficiently compact to enable it to be mounted on any operating train and be able to survey at train operating speeds.
An automated imaging system will be implemented to recognise assets and their properties from high definition 2D images captured using forward facing cameras.
The project will use a semi-supervised approach to develop classifiers leveraging human experience to train a flexible and accurate system for detecting new types of assets.
Using a combination of sensors including lasers and cameras, we will build a navigable map of the environment. We can then localise trains and detected assets within this map.
A key challenge of this project will be to find a robust solution for estimating the extrinsic calibration and synchronisation between different sensor modalities.
Using all of the above elements combined, we will be able to identify assets and make measurements on their attributes (such as size, orientation, etc.), whilst locating them within a map of the environment. MRG has a suite of publications related to the research areas outlined above.