Deep-Learning-Based Detection of Obstacles in Transit on Trams

  • Yiming Li
  • Guoqiang CaiEmail author
Conference paper
Part of the Lecture Notes in Electrical Engineering book series (LNEE, volume 617)


Due to the pervasive employment of trams, the measurements to keep the transit of trams safe are necessary and emergent. In this sense, we put forward a Neural Network based on Convolutional Neural Network, in which we made some modifications to make it more flexible. Such as passthrough layer to ensure some small objects detectable, anchor boxes to ensure a high-speed detection, and batch-normalization layers to make the network be malleable for objects with different distributions. With this network, we can efficiently detect possible obstacles, such as pedestrians, cars and some other objections that may endanger the trams. We test the network among several databases with 5000 samples, and the average accuracy rate is 94.12%, the average detecting speed is 30 FPS, the smallest detectable object’s size is 20 × 20 pixels, these all show remarkable result.


Deep learning Tram Real-time detection Batch normalization 



This work was supported by the National Key Research and Development Plan (No. 2018YFB1201601-07).


  1. 1.
    Currie G, Reynolds J (2011) Managing trams and traffic at intersections with hook turns. Transp Res Record: J Transp Res Board 2219:10–19CrossRefGoogle Scholar
  2. 2.
    Samala RK, Chan HP, Hadjiiski LM et al (2017) Multi-task transfer learning deep convolutional neural network: application to computer-aided diagnosis of breast cancer on mammograms. Phys Med Biol 62(23)Google Scholar
  3. 3.
    Hertel L, Barth E, Käster T et al (2017) Deep convolutional neural networks as generic feature extractorsGoogle Scholar
  4. 4.
    Shelhamer E, Long J, Darrell T (2014) Fully convolutional networks for semantic segmentation. IEEE Trans Pattern Anal Mach Intell 39(4):640–651CrossRefGoogle Scholar
  5. 5.
    Ren S, He K, Girshick R et al (2015) Faster R-CNN: towards real-time object detection with region proposal networks. In: International conference on neural information processing systems. MIT Press, Cambridge, pp 91–99Google Scholar
  6. 6.
    He K, Zhang X, Ren S et al (2014) Spatial pyramid pooling in deep convolutional networks for visual recognition. IEEE Trans Pattern Anal Mach Intell 37(9):1904–1916CrossRefGoogle Scholar
  7. 7.
    Girshick R (2015) Fast R-CNN. Computer ScienceGoogle Scholar
  8. 8.
    Ioffe S, Szegedy C (2015) Batch normalization: accelerating deep network training by reducing internal covariate shift. In: International conference on international conference on machine learningGoogle Scholar
  9. 9.
    Szegedy C, Liu W, Jia Y et al (2015) Going deeper with convolutions. In: Computer vision and pattern recognition. IEEE, New York, pp 1–9Google Scholar
  10. 10.
    Szegedy C, Vanhoucke V, Ioffe S et al (2015) Rethinking the inception architecture for computer vision. Comput Sci 2818–2826Google Scholar
  11. 11.
    Giusti A, Dan C C, Masci J et al (2013) Fast image scanning with deep max-pooling convolutional neural networksGoogle Scholar
  12. 12.
    Schmidt-Hieber J (2018) Nonparametric regression using deep neural networks with ReLU activation functionGoogle Scholar
  13. 13.
    Olkkonen H, Pesola P (1996) Gaussian pyramid wavelet transform for multiresolution analysis of images. Graphical Models Image Process 58(4):394–398CrossRefGoogle Scholar
  14. 14.
    Neubeck A, Gool LV (2006) Efficient non-maximum suppression. In: International conference on pattern recognition. IEEE, New York, pp 850–855Google Scholar
  15. 15.
    Redmon J, Farhadi A (2016) YOLO9000: Better, Faster, Stronger, 6517–6525Google Scholar

Copyright information

© Springer Nature Singapore Pte Ltd. 2020

Authors and Affiliations

  1. 1.Beijing Jiaotong UniversityBeijingChina
  2. 2.State Key Lab of Rail Traffic Control and SafetyBeijingChina

Personalised recommendations