Open AccessTraffic sign detection and recognition based on CNN-ELM
Author(s) -
Wenju Li,
Xinyuan Na,
Pan Su,
Qing Zhang
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1848/1/012106
Subject(s) - extreme learning machine , traffic sign recognition , computer science , artificial intelligence , pattern recognition (psychology) , feature (linguistics) , sign (mathematics) , image (mathematics) , traffic sign , artificial neural network , mathematics , mathematical analysis , linguistics , philosophy
Aiming at the problem of low traffic sign recognition rate and slow speed, a traffic sign recognition algorithm combining CNN and Extreme Learning Machine is proposed. First, the ResNet50 network is used to extract image features, and then the Region Proposal Network (RPN) is used to generate proposals from the extracted image feature maps. Finally, the extreme learning machine is used to classify the generated proposals, and the fully connected layer is used for regression prediction. The experiment shows that compared with the Faster R-CNN model, the CNN+ELM improves the recognition accuracy on the TT-100K dataset 7.7% and reduces the training time per epoch by 32 seconds.