Metrological characterization of the intel RealSense D415 stereo depth camera for additive manufacturing

Abstract

Additive manufacturing has emerged as a disruptive technology for producing complex and customized components across various industries. As the demand for high-quality parts grows, the need for robust quality inspection methods becomes critical. This master’s thesis investigates the feasibility and accuracy of utilizing a single stereo camera system, specifically the Intel RealSense D415, for quality assessment of miniature components derived from the additive manufacturing process. Unlike traditional inspection techniques that primarily focus on dimensional inaccuracies and defects, this research introduces a novel evaluation approach, termed the Stereo Accuracy Assessment in Additive Manufacturing via Plane Fitting (SAAPF), designed to verify the accuracy of stereovision cameras for uniform, low-texture parts with millimeter-scale dimensions at close-range distances of 50 cm. The methodology involves creating CAD models, producing them via additive manufacturing using the Ultimaker Connect 2+, and capturing point cloud data with the Intel RealSense D415. The data is analyzed using the Intel RealSense Viewer and CloudCompare software, with accuracy assessed through plane fitting, cloud-to-cloud distance, and Gaussian distribution techniques. The research contributes a structured process for evaluating the camera's resolution limits and accuracy in this specific application. If validated, this approach could be integrated into additive manufacturing systems, paving the way for next-generation 3D printers with built-in precision quality control capabilities.