A Real-time Traffic Sign Detection Model Based on Improved YOLOv3
Xiongfei Liu, Fan Xiong · IOP Conference Series Materials Science and Engineering · 2020
Abstract Advanced driver assistance system, which can quickly and accurately detect driver assistance personnel, ensure traffic safety, and play an important role in guiding road traffic signs. However, traditional detection models often have false detection, missed detection, and predicted low accuracy of the target box, especially for the Identification of small targets, its accuracy is difficult to meet the requirements. Therefore, based on the target detection algorithm model YOLOv3, this paper replaces the common two-dimensional convolution with dilated convolution and expands the receptive field, At the end of the model, a three-layer spatial pyramid cascade structure is added to make the multi-scale feature map deep fusion. finally, the improved model is tested for performance on the GTSDB dataset. The results show that the improved model implemented in this paper can effectively solve some problems in the actual road traffic sign detection, and meet the requirements of real-time detection.