关键词: 交通标志, 深度卷积神经网络, 目标检测, 图像分割, 图像标记

Abstract: In order to solve the problem that it is difficult to get traffic signs from natural scenes, an automatic traffic sign constructing method based on deep convolutional neural network(CNN) and graph cut is proposed. Firstly, an initial dataset is obtained by collecting seven main categories of traffic signs and their subclasses. The initial dataset is used to train CNN model for traffic sign detection. Then the traffic sign’s category, confidence level and location of images or videos are detected by the trained CNN model. Secondly, the detected traffic sign’s region which is greater than a given threshold is stored. The corresponding new region is obtained by localization algorithm based on graph cut. Thirdly, the new traffic sign’s region and its category is used as its ground truth. The new training set and testing set by adding the new detected traffic signs is applied to fine-tune the convolution network and obtain a new model. The experimental results show that the fine-tuned model has a higher mean average precision than the initial one. This method can automatically generate a new traffic sign data set by relocated the detected traffic sign from images or videos which are captured by the vehicular camera or the traveling data recorder.

Key words: traffic sign, deep convolutional neural network, object detection, image segmentation, image component labeling