3D Surface Reconstruction with Polynomial Regression for Multistage Stamping Die using Depth Camera

At present, many advancements have been made in 3D reconstruction, and is still expected to grow for at least another decade. We applied a single depth camera to capture an image of a stamping die and store the value of depth points for creating the corresponding point cloud for surface reconstruction. The processing steps involved defining and extracting the field of interest (FoI) of the point cloud with additional filtering to remove unwanted noise. To reduce processing time during the B-spline surface reconstruction, clusters were formed within the point cloud to retrieve only the peak point of each punch in the die. The average error in millimeters between the real object’s dimension and the resultant point cloud was below 10% and the average time taken for the surface reconstruction was 2 to 3 min.