Optimization of Traffic Sign Detection and Classification Based on Faster R-CNN
Qiao Kun, Hanzhou Gu, Jiaming Liu, Pei Liu · 2017 International Conference on Computer Technology, Electronics and Communication (ICCTEC) · 2017
Traffic sign detection and recognition is a key area of research on intelligent transportation, which has significant theoretical value and an expansive market application prospect. As a crucial part, the algorithm of traffic sign detection and classification has great impact on subsequent procedures. In this way, implementing a faster and robust algorithm is what most researchers are pursuing in this area. However, sometimes, such a great variety of signs are hard to be detected or classified especially if they are spoiled or the driving environment is complicated. Traditional methods are mostly based on extracting features like color or shape, which need higher quality of images and may sometimes lead to a poor precision and robustness. This paper provides an optimization based on Faster R-CNN combining with ZF and VGG network. This algorithm improves validation accuracy and robustness, which also reduces the requirements of quality of images and related computation.