Structured light centerline extraction algorithm based on skeleton-guided spatial weighting

When using line structured light for 3D reconstruction, if the object being measured has continuous undulating structures and is highly reflective, the laser stripe on the object's surface can easily exhibit problems such as width changing drastically and edge burrs protruding irregularly. To address the issue that existing algorithms cannot accurately extract the laser center point in such cases, a line structured light center extraction algorithm based on skeleton-guided spatial weighting is proposed. The algorithm preprocesses to extract accurate laser stripe regions and then advances directly from the two edge endpoints to obtain the skeleton points. Next, it corrects the protruding edges of the laser stripe and performs spatial weighting based on the positional relationship of adjacent skeletons. The initial position of the center point is extracted using the grayscale centroid method based on weighted gray distribution. Finally, the Hessian matrix is calculated to correct the center point coordinates along the laser stripe's travel direction. Experimental results show that, compared with the improved unilateral tracing method, our method reduces the error by 15% and improves stability by 58%. Compared with the grayscale centroid method, our method is 1.6 times faster, indicating that the algorithm has high accuracy and good robustness. The laser stripe center extraction accuracy is less affected by image brightness, while maintaining a moderate running speed.