Traffic Signs Classification Based on Local Characteristics and ELM
Wenju Li, Qi Chen, Tianzhen Dong, Lihua Wei, Qing Zhang · 2017
Traffic signs classification as a key problem in ITS. This paper presents a method about traffic signs recognition based on local characteristics and ELM neural network. This method combines a variety of local characteristics and ELM neural network. First, HOG characteristics and LBP characteristics were used to extract sign characteristics and formed a new eigenvector, which was inputted onto ELM neural network for training. Finally the trained ELM network was used to traffic signs recognition. In this paper the traffic sign samples extracted from the GTSRB database, including 780 training samples and 120 testing samples, the result shows that the recognition rate of test sample is up to 100% with 2.97 ms/frame.