Automated Bundy Tube Metrology with Deep Learning

The precise measurement of angles and thickness in Bundy tubes is of paramount importance in various industries, such as automotive and refrigeration, where these tubes serve as vital components in heat exchangers and fluid transportation systems. This project tackles this task with an innovative approach of using a multi-step process that includes a deep learning model like U-Net, Mask R-CNN, YOLO segmentation, and Segment Anything models, to generate segmentation masks and the OpenCV library to extract critical geometric features like angle and thickness. The utilization of OpenCV ensures robustness and efficiency in obtaining precise measurements, even in the presence of noise and occlusions. Finally, the obtained measurements are evaluated against predetermined quality thresholds to determine the conformity of the Bundy tube. Combining these methodologies enables rapid and accurate measurement of Bundy tube angles and thickness, contributing to improved quality control and manufacturing processes.