Vision-based Estimation of Crane Spreader Rotation for Automated Container Handling

Abstract

With the rapid expansion of global trade and the increasing reliance on automated systems in port operations, accurately estimating crane rotation angles during cargo stacking and un- stacking is critical for enhancing container placement precision. This study presents a computer vision-based method that leverages predefined guiding marker objects to infer the crane’s rota- tion angle. Since these markers remain fixed in position, their detected rotation in crane-mounted camera images serves as a proxy for crane rotation. Our proposed two-stage pipeline consists of the following: marker object detection to localize the location of markers and 2D keypoint regression to estimate their rotation angles. By computing the average markers’ rotation angle, we reliably infer the crane’s orientation. Experimental validation using real-world images demonstrates the accuracy and robustness of our approach, making it a reliable real-time solution in cargo positioning. This system significantly improves operational efficiency and safety in container terminal operations, enabling more precise and automated cargo handling.