Traffic Sign Detection From Video: A Fast Approach With Tracking

This paper proposes a fast approach for traffic sign detection from video. First, we modify the image-based detector HHVCas to improve its accuracy and speed, then apply it to video-based detection with further acceleration by tracking. For the image-based detector, by optimizing the parameters in the cascade using an unsupervised approach, we achieve performance comparable to the state-of-the-art while keeping the speed advantage. Parallelizing some steps in the HHVCas detector leads to 1.5x speedup and 20 fps detection. In video, the detector achieves 2.8x speedup and performs 35 fps by tracking every other frame. It also obtains significant precision increase by 5 similar to 8% at high recall when exploiting temporal coherence of results in multiple frames.