Simulation Tools and Techniques. 11th International Conference, SIMUtools 2019, Chengdu, China, July 8–10, 2019, Proceedings

Research Article

Simulated Traffic Sign Classification Using Cross-Connected Convolution Neural Networks Based on Compressive Sensing Domain

  • @INPROCEEDINGS{10.1007/978-3-030-32216-8_57,
        author={Jiping Xiong and Lingfeng Ye and Fei Wang and Tong Ye},
        title={Simulated Traffic Sign Classification Using Cross-Connected Convolution Neural Networks Based on Compressive Sensing Domain},
        proceedings={Simulation Tools and Techniques. 11th International Conference, SIMUtools 2019, Chengdu, China, July 8--10, 2019, Proceedings},
        proceedings_a={SIMUTOOLS},
        year={2019},
        month={10},
        keywords={Compressive sensing domain Convolution neural networks CS measurements Simulated traffic sign recognition},
        doi={10.1007/978-3-030-32216-8_57}
    }
    
  • Jiping Xiong
    Lingfeng Ye
    Fei Wang
    Tong Ye
    Year: 2019
    Simulated Traffic Sign Classification Using Cross-Connected Convolution Neural Networks Based on Compressive Sensing Domain
    SIMUTOOLS
    Springer
    DOI: 10.1007/978-3-030-32216-8_57
Jiping Xiong1,*, Lingfeng Ye1, Fei Wang1, Tong Ye1
  • 1: Zhejiang Normal University
*Contact email: xjping@zjnu.cn

Abstract

This paper proposes an algorithm of simulated traffic sign recognition based on compressive sensing domain and convolution neural networks for simulated traffic sign recognition. And the algorithm can extract the discriminative non-linear features directly from the compressive sensing domain. The image is transformed into compressive sensing domain by measurements matrix without reconstruction. This paper proposes a cross-connected convolution neural networks (CCNN) with an input layer, 6 six hidden layers (i.e., three convolution layers alternating with three pooling layers), a fully-connected layer and an output layer, where the second pooling layer is allowed to directly connect to the fully-connected layer across two layers. Experimental results show that the algorithm improves the accuracy of simulated traffic sign recognition. The recognition of the algorithm is possible even at low measurement rates.