Fig. 1. Scenario of autonomous train positioning in tunnel. (a) Running train with VLC; (b) cross-sectional view of metro tunnel

Fig. 2. VLC process based on CMOS

Fig. 3. Stripe image of LED lamp

Fig. 4. Stripe recognition of LED lamp image

Fig. 5. Steps of train positioning

Fig. 6. Perspective projection model of binocular stereo vision

Fig. 7. Image processing flow of receiver

Fig. 8. Mounting angle of receiver

Fig. 9. Image of bright spot with motion blur and image after filtering. (a) Image of bright spot with motion blur; (b) image after Wiener filtering

Fig. 10. Experimental platform of autonomous train positioning using VLC and binocular stereo vision

Fig. 11. Static train positioning. (a) Train positioning result; (b) positioning error distribution

Fig. 12. Comparison of positioning results

Fig. 13. Comparison of train positioning at speed of 20 km/h before and after compensation with IMU. (a) Comparison of Phi;(b) comparison of positioning results

Fig. 14. Results of dynamic train positioning at different speeds

Fig. 15. Train positioning errors at different speeds. (a) Distribution of train positioning errors at speed of 20 km/h; (b) range of train positioning errors