Traffic Sign Detection Based on Camera Imaging Apriority

Traffic sign detection is very important to the vehicle intelligent auxiliary driving system and the driverless system. However, traffic sign detection is still a challenging problem, and there is not a satisfactory solution until now. In this paper, we aim at improving the speed and accuracy of traffic sign detection. In order to improve the detection speed, we use the image-forming principle to select the scale of sliding windows instead of the standard sliding window scheme. This operation will reduce the computational complexity from

O

(

N

4

) to

O

(

N

2

). In order to improve the detection accuracy, we adopt the hierarchical detection scheme. In the first stage, we use the cascade GentleAdaBoost classifier combined with the Haar-like features; in the second stage, we use the GentleAdaboost classifier combined with the multiple features fusing the color cues. The hierarchical detection scheme greatly reduces the false positive rate. We implement our approach on the Swedish Traffic Signs Dataset, the experimental results demonstrate our approach is effective and our approach could greatly reduce the false positive rate while keeping the detection rate.