Investigating the Three-Dimensional Reconstruction Accuracy of Additively Manufactured Parts using Structure from Motion

Abstract

Additive manufacturing (AM) adoption in production environments requires the ability to verify geometry and tolerances. Inspection technologies such as laser and structured-light scanners offer good performance but face limitations in cost, portability, and flexibility. There is a need for complementary low-cost approaches for routine inspection tasks. This paper develops a workflow for part inspection using Structure-from-Motion (SfM), a method that creates three-dimensional (3D) objects from photographs. The workflow is tailored for AM inspection, systematically isolating the influence of surface appearance (texture/ reflectivity), sensor modality, image count, and imaging parameters (downsampling and magnification) on reconstruction fidelity. The paper studies 3D reconstruction of an AM engine bracket fabricated with six controlled surface finishes, with both polymer and metal parts. The bracket is imaged with monochrome and color machine-vision cameras. We demonstrate calibrated capture, incremental reconstruction, and joint bundle adjustment. The reconstruction quality is quantified through photometric consistency and dimensional accuracy as a function of the number and type of images used for SfM. The reconstructed 3D objects are compared against laser scan and caliper measurements. We find that surface appearance strongly dictates reconstruction accuracy, and that monochrome imaging provides more stable feature detection and lower error than color imaging. Increasing the number of viewpoints improves accuracy up to approximately 360 images, beyond which additional viewpoints yield little benefit. Reasonably good 3D reconstruction is possible with 180 images, depending on the part finish and other parameters. We find that SfM can achieve dimensional accuracy smaller than 0.1 mm, with residual errors concentrated in occluded or high-curvature regions. The paper establishes practical guidelines for applying SfM as a low-cost metrology tool for AM part inspection and reverse engineering.