Multiple camera calibration and stereo depth estimation for automated quay crane

• Main Author: Tu, Weifeng
• Other Authors: Wang Dan Wei
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering; Electrical & electronic engineering
• Online Access: https://hdl.handle.net/10356/176710
• Institution: Nanyang Technological University

The seaport environment significantly depends on the quay crane system for effectively handling containers from container ships. Recently, multiple cameras have been built into the spreader, providing RGB and depth pictures to improve intelligence, safety, and operational efficiency. However, it becomes crucial to handle two key issues: (1) How to correctly calibrate the intrinsic and extrinsic calibration parameters of the cameras and (2) How to determine the depth distance of the container to the spreader cameras. This report presents a comprehensive pipeline to tackle these challenges. Firstly, the cameras mounted on the spreader are calibrated offline using the Kalibr toolbox. Secondly, the stereo camera pairs will be used to estimate the depth of the container by processing the images through the stereo match algorithm (Libsgm). Finally, the experiment suggests that the outcome for the multiple camera calibration provides a reprojection error of about 0.3 pixels, and the outcomes for the depth estimation give an error between 2% and 8%. The two algorithms are used to determine how well this method works for precisely analysing object depth, showcasing improved efficiency and safety in seaport operations involving quay crane systems.