Fusion of digital map traffic signs and camera-detected signs

In this paper, we present several fusion approaches to merge speed limits reported in digital maps with detected speed limit signs using an onboard camera. Digital maps holding speed limits signs are required to be updated to cover speed limit changes and are unable to support variable speed limits. On the other hand, a camera system placed onboard a vehicular can detect variable speed limits as well as temporary speed limits at construction sites. However, an onboard camera cannot detect implicit speed limits. As such, a combination of digital map and camera system can provide more accurate speed limit information for driver assistance and support vehicle safety features. In this paper, the digital map and camera system are fused to obtain the desired more accurate speed limit information. The fused speed limits as well as those from the individual sources are compared with ground truth data obtained from an extensive measurement campaign spanning over 15000 km driving distance in five European countries. Specifically, five fusion approaches are defined, modeled and evaluated. Four of them are based on prioritizing information while the remaining approach is based on the classical Dempster-Shafer data fusion technique. The performance of the approaches are presented in terms of the percentage of correct speed limit detection with respect to the driving distance. The obtained results clearly show that fusion techniques can significantly increase the amount of correctly detected speed limits. A MATLAB based graphical user interface was designed to load test data, evaluate and present the results as fast and as efficient as possible.