Robust adaptive sub-pixel advancement extraction algorithm for line laser centers with curvature variation

To address the issues of missing center points, positioning deviations, and poor continuity in existing line-structured light center extraction algorithms under complex curvature variations, stripe ends, and noise interference, this paper proposes a robust adaptive sub-pixel extraction algorithm for line laser centers. First, edge padding is applied to the image. For light stripes with varying curvature, a continuous and smooth laser stripe orientation field is generated based on structure tensor estimation. Second, this orientation field guides an internal advancement strategy to effectively compensate for the loss of center points at stripe ends caused by skeleton thinning. Finally, bilinear interpolation is performed along the normal direction, combined with the gray-scale centroid method to achieve high-precision sub-pixel center localization. Experimental results show that in noise resistance tests, the proposed algorithm completely extracts 401 center points even at a noise variance of 0.14, with a mean absolute error (MAE) below 0.23, significantly outperforming classical algorithms such as Steger and the directional template method (DTM). In real-image tests, the algorithm achieves continuous, complete, smooth, and redundant-free center extraction for light stripes with different curvatures while maintaining high computational efficiency.