Fig. 1. Light source layout

Fig. 2. Reflection models. (a) Node reflection model; (b) traditional reflection model

Fig. 3. Reflection node positions. (a) Reflection point positions; (b) calculation principle of reflection points

Fig. 4. Fully connected neural network model

Fig. 5. Parallel fully connected neural network model

Fig. 6. Convolution-Dilated Convolution model

Fig. 7. Flow chart of Mot-PSO-K-Means++

Fig. 8. Flow chart of improvement of K-Means++

Fig. 9. Loss value iteration curves. (a) Iteration curves of 4-LED; (b) iteration curves of 12-LED

Fig. 10. Line graph of relation of MAE and number of kernels

Fig. 11. Loss value iteration curves

Fig. 12. Loss value iteration curves. (a) Convolution neural network iteration curves; (b) PFCNN iteration curves

Fig. 13. Initial population optimization. (a) Iteration curves of convolution neural network; (b) clustering centers of improved K-Means++

Fig. 14. Algorithm iteration curves

Fig. 15. Flatness parameter iteration curves based on variance parameters. (a) 4-LED iteration curve; (b) 12-LED iteration curve

Fig. 16. Optimal layouts of 4-LED based on variance parameters. (a) 4-LED optimal location; (b) intensity distribution of 4-LED; (c) SNR of 4-LED

Fig. 17. Optimal layouts of 12-LED based on variance parameters. (a) 12-LED optimal location; (b) intensity distribution of 12-LED; (c) SNR of 12-LED

Fig. 18. Flatness iteration curves under Monte Carlo parameters. (a) 4-LED iteration curve; (b) 12-LED iteration curve

Fig. 19. Optimal layouts of 4-LED under Monte Carlo parameters. (a) 4-LED optimal location; (b) intensity distribution of 4-LED;(c) SNR of 4-LED

Fig. 20. Optimal layouts of 12-LED under Monte Carlo parameters. (a) 12-LED optimal location; (b) intensity distribution of 12-LED; (c) SNR of 12-LED