Multi-model high-precision 3D measurement method dominated by weighted dimensionality reduction

Abstract

In order to address the problems of long processing time, low accuracy, and insufficient stability in traditional Sheet-of-Light centerline extraction algorithms, a multi-model high-precision 3D measurement method dominated by weighted dimensionality reduction is proposed. This method transforms the problem of extracting the 2D coordinates of the Sheet-of-Light centerline into the problem of calculating vector centroids and matching vector indices within the Grayscale Vector Pool. The Grayscale-Multi-State Weighted Dimensionality Reduction Model is employed to accurately compute the vector centroids in the Grayscale Vector Pool. Subsequently, a traversal computation model is constructed to perform a single traversal of the Grayscale Vector Pool, yielding the matching results between vector indices and their corresponding grayscale centroids.The effectiveness of the proposed algorithm is validated through comparative and 3D measurement experiments. Compared with the Steger algorithm, the proposed method reduces the root-mean-square error (RMSE) of the extracted Sheet-of-Light centerline by 0.11965 pixels and shortens the computation time by a factor of 1.31. Furthermore, the repeatability accuracy error is reduced by 0.21667 pixels compared with the Steger algorithm, while the computation time is shortened by a factor of 5.50. These results demonstrate that the proposed method offers high accuracy and significant application value.