University of Bristol > Visual Information Laboratory > Jack Greenhalgh > Real Time Detection and Recognition...

Real Time Detection and Recognition of Road Traffic Signs

We are currently researching automatic traffic sign detection and recognition, which is an important part of an advanced driver assistance system (ADAS). The purpose of this research is to implement algorithms for the real time detection and recognition of ideogram based road signs in real traffic scenes. Traffic symbols have several distinguishing features, such as shape and colour, which may be used for their detection and identification. These features can also be used to place road signs into specific semantic groups.

Our system consists of two stages; detection of candidate regions, which is performed using a novel application of maximally stable extremal regions (MSERs), and recognition, which is performed with HOG features classified using Random Forests. All training data used for the training of classifiers is synthetically generated from template images, taken from an online database. This approach allows classification to be performed on all possible road signs, and also avoids the tedious process of hand-labelling large datasets.