Analyzing automatically captured traffic sign data from the Vionice system
Abstract
The Finnish startup company Vionice has developed a system for automatically capturing data about road signs. The system uses mobile phones with camera and GPS placed in the windscreen of NPRA vehicles, and computer vision and machine learning algorithms to recognize and position signs. This system has been used to capture data about all speed limit and road works signs that the vehicles are passing, and the results have been analysed.
The analysis shows that the system looks very promising, but since the actual sign locations from NVDB, which are used to validate the system, are shown to be inaccurate, the system still lacks a solid validation.