Abstract:To address the issues of low accuracy and inefficiency in traditional sand vessel volume measurement methods,which primarily rely on manual operations,we propose a deep learning-based sand vessel measurement algorithm.Integrating 3D point cloud technology with the PointNet++ deep learning network achieves automatic classification and precise segmentation of sand body point clouds inside the vessel.An artificial intelligence (AI)-based sand vessel volume measurement system is developed,and a PointNet++ model is built to perform semantic segmentation on the collected point cloud data and calculate the sand volume.The results show that in an overseas reclamation project,the system reduces the internal volume processing time for a single vessel from 60 min (manual) to under 10 min,improving efficiency significantly.The accuracy of automatic sand compartment classification increases to over 95%,and the measurement error rate drops to below 2%.The proposed AI-based sand vessel volume measurement approach significantly enhances the efficiency and accuracy of sand load calculations,reduces on-site operational risks,and meets the intelligent measurement demands of modern marine engineering,offering strong practical value for engineering applications.
