Fig. 1. Design of the laser alignment system. (a) Alignment system structure; (b) Principle of coordinate transformation

Fig. 2. Improved SSD model

Fig. 3. Diagram of the depthwise separable convolution

Fig. 4. Spatial attention module

Fig. 5. Diagram of the feature fusion module

Fig. 6. Two-dimensional strength distribution of three channel pixel values in the RoI area of a captured image. (a) Red channel; (b) Green channel; (c) Blue channel

Fig. 7. Comparation on histograms after grayscale processing using two weighted average methods. (a) Using conventional weights; (b) Using improved weighted average method

Fig. 8. Images obtained using different threshold processing methods. (a) Original Image; (b) Grayscale image; (c) Using iterative method; (d) Using Otsu method; (e) Using proposed method with the threshold is 133

Fig. 9. The evaluated curve during training. (a) Training set loss curves; (b) Validation set loss curves; (c) Validation set RoI AP curves

Fig. 10. Predict results of different network models. (a) VGG16-SSD; (b) MobileNet-SSD; (c) Improved MobileNet-SSD

Fig. 11. Experimental device diagram

Fig. 12. Comparison before and after alignment. (a) Coordinate of center point of light spot before alignment is (324.5262.0); (b) Coordinate of the center point of the light spot after alignment is (319.35278.97)