Fig. 1. Detection and identification of LED-ID based on BP neural network (BPNN)

Fig. 2. Extraction process of LED-ROI. (a) Gray scale images; (b) processing of binary images; (c) closing operation; (d) obtaining LED-ROI

Fig. 3. Stripe images with different frequency features. (a) Modulated signal frequency is 1000 Hz; (b) modulated signal frequency is 1250 Hz

Fig. 4. Stripe images at different imaging distances. (a) 40 cm; (b) 50 cm; (c) 60 cm; (d) 80 cm; (e) 100 cm; (f) 130 cm

Fig. 5. Stripe images with different duty cycles. (a) Duty cycle is 80%; (b) duty cycle is 50%

Fig. 6. Process of image recognition and feature extraction of optical stripe code

Fig. 7. Principle of area feature extraction

Fig. 8. Principle of feature extraction of frequency and duty cycle for optical stripes code

Fig. 9. BP neural network (BPNN) structure of classification recognition for LED-ID

Fig. 10. Train positioning model based on optical camera communication

Fig. 11. Relationship between different coordinate systems

Fig. 12. Rotation relationship between camera coordinate system and world coordinate system

Fig. 13. Experimental platform for train positioning

Fig. 14. Two-dimensional positioning resultsat Z=0 m

Fig. 15. Two-dimensional positioning results at Z=0.2 m

Fig. 16. Two-dimensional positioning results at Z=0.4 m

Fig. 17. Two-dimensional positioning results at Z=0.6 m

Fig. 18. Comparison of mean positioning error at different vertical distances

Fig. 19. Cumulative distribution function (CDF) of positioning error

Fig. 20. Dynamic positioning trajectory