Information extraction from complex circular charts

The paper is concerned with the recognition of circular charts (disks). The application domain is tachograph disks. These disks exhibit typical properties of circular charts but are more complex as different types of information are present at the same time and there are serious artefacts that hinder the recognition process. The method starts by analysing the characteristics of the greyscale image to separate the background from the recorded information. The position of the disk and its centre are then established. Finally, the orientation of the disk is determined and the information (recorded in a circular manner) is read and converted into a linear representation. Experimental results of testing against professionally prepared ground-truth show an average accuracy of 94% in the recognition of the target information.