Data and code for "Data-Driven Extrinsic Volumetric Calibration of 3D Galvanometric Lasers and Off-Axis Vision Systems for Free-Form Processing"

This repository contains the datasets and code associated with the article “Data-Driven Extrinsic Volumetric Calibration of 3D Galvanometric Lasers and Off-Axis Vision Systems for Free-Form Processing.” The repository includes the experimental 3D data acquired during the volumetric calibration and validation campaigns, the processed calibration datasets derived from these acquisitions, and the scripts used to construct and evaluate the camera-to-laser mappings. The material covers the main stages of the proposed black-box extrinsic volumetric calibration workflow, including point-cloud processing, calibration-mark extraction, geometric reference-coordinate generation, calibration-data preprocessing, polynomial regression, artificial neural network training, offline model evaluation, and physical positioning validation. The Revision folder contains the final code corresponding to the revised version of the manuscript and should be considered the primary implementation associated with the published study. This code incorporates the methodological and experimental changes introduced during peer review, including the revised geometric formulations, random/depth-based/spatial holdout strategies, regression baselines, polynomial degree and regularization selection, ANN architecture selection across repeated initializations, measurement-repeatability analysis, statistical comparisons, calibration-support classification, temporal-stability analyses, and the additional quantitative validation experiments. The repository also retains the original MATLAB implementation in the Matlab directory outside the Revision folder. These files correspond to the methodology and code used for the initial manuscript submission and are preserved for traceability. This earlier implementation also constitutes the code base used with the scanning laser profilometer in the qualitative free-form processing experiments. Users interested in reproducing the final quantitative results and methodological formulation reported in the revised manuscript should therefore use the code provided in Revision. The experimental material includes calibration and validation data obtained with the Photoneo XS structured-light scanner, together with data associated with the scanning laser profilometer and the application-oriented free-form processing demonstrations. Quantitative positioning validation in the revised study is performed with the Photoneo configuration and includes inclined planar targets, discrete and continuous processing on a curved surface, repeated-scan measurement assessment, intra-day and inter-day stability experiments, direction and speed diagnostics, and an independent microscopic comparison. The profilometer experiments are retained as qualitative demonstrations of the calibration and processing workflow rather than as quantitative positioning validation. The regression code includes complete multivariate polynomial models fitted through ordinary least squares or Ridge regression, as applicable, and feedforward artificial neural networks trained using the Levenberg–Marquardt algorithm. The revised implementation also contains the procedures used for validation-based hyperparameter selection and structured generalization assessment. The repository is intended to support independent verification and reproducibility of the proposed volumetric registration between off-axis 3D measurements and the operational Cartesian commands of a commercial 3D galvanometric laser system with dynamic focus. Repository structure. The repository is distributed across several ZIP archives only to facilitate upload and download of the complete dataset. All archives correspond to complementary parts of a single directory tree rooted at ArticleEnvironment, which is the directory structure expected by the provided scripts. To reconstruct the original environment, extract all required ZIP archives into the same destination so that their ArticleEnvironment folders are merged while preserving the original subdirectory names and relative paths. The final revised MATLAB implementation is located in ArticleEnvironment/Matlab/Revision; MATLAB code outside this folder corresponds to the original submission workflow and is retained for traceability and for the qualitative profilometer experiments.