A Bayesian approach to traffic light detection and mapping

作者：Hosseinyalamdary Siavash^*; Yilmaz Alper

来源：ISPRS Journal of Photogrammetry and Remote Sensing, 2017, 125: 184-192.

DOI：10.1016/j.isprsjprs.2017.01.008

• Traffic light detection and mapping
• Conic section geometry
• Spatio-temporal consistency
• Bayesian inference

Automatic traffic light detection and mapping is an open research problem. The traffic lights vary in color, shape, geolocation, activation pattern, and installation which complicate their automated detection. In addition, the image of the traffic lights may be noisy, overexposed, underexposed, or occluded. In order to address this problem, we propose a Bayesian inference framework to detect and map traffic lights. In addition to the spatio-temporal consistency constraint, traffic light characteristics such as color, shape and height is shown to further improve the accuracy of the proposed approach. The proposed approach has been evaluated on two benchmark datasets and has been shown to outperform earlier studies. The results show that the precision and recall rates for the KITTI benchmark are 95.78% and 92.95% respectively and the precision and recall rates for the LARA benchmark are 98.66% and 94.65%.

• 出版日期2017-3