Method for Rrecognizing and Generating Three-Dimensional Objects to Create Digital Twins

Abstract

Logistics automation is a key component in the digitalization of transport infrastructure companies. Particularly, solutions aimed at optimizing container yard management processes at sea and river ports are being actively proposed. To achieve this, digital twin technologies, which enable real-time visualization of all processes, are being used. A separate task is the reconstruction of a company’s territorial topology, which is often based on internal data or drawings that may be outdated or difficult to interpret. This paper proposes an approach for the automatic reconstruction of the territorial topology of marine and river logistics enterprises, including the definition of container contour yards boundaries, using artificial intelligence technology. A set of computer vision algorithms, consisting of two U-Net models and a refined YOLOv8 model, was developed. The contour detection achieved an mAP@0.5 score of 0.95. Additionally, a prototype editor was implemented to restructure port topologies, transform data, and visualize the area in 3D space. Testing was conducted, revealing an average contour deviation of -0.153 from the real data. Additionally, an algorithm for GIS system data conversion and topology visualization in 3D was developed using a graphics engine.