Feature Point Extraction Algorithm of Compound Groove Based on Linear Structured Light

When using structured light vision system to extract composite groove feature points, it is difficult to extract feature points due to the surface reflection of the groove and the interference of ambient light. A novel algorithm based on linear structured light was proposed to adapt to the environment under different noises. Firstly, the linear laser fringe image obtained by the structured light vision sensor was preprocessed, adaptive threshold segmentation was used to reduce the influence of different brightness and contrast, and the center point of the weld was obtained by skeleton extraction. Secondly, Gaussian density function was constructed to fit the weld center line by integral method and least square method. Finally, the function was adjusted according to the modified L-M algorithm, and the residual difference between the logarithmic data point and the model prediction was minimized. The results show that the root-mean-square error of the proposed method is less than 0.20 mm, which can meet the requirements in the actual welding process.