A study of soil modelling methods based on line-structured light—Preparing for the subsoiling digital twin

Abstract

The virtual model forms the foundation for building a digital twin system; however, methods for modelling dynamically changing soil in subsoiling have not yet been studied. To provide technical guidance for constructing such a system, this study employs a line structured light method for soil model construction. After conducting field and indoor trials, the extreme value method, grayscale centroid method, and Steger algorithm are used to extract the laser centreline. Results indicate that the extreme value method and grayscale centroid method require relatively little processing time—approximately 1.9 ms and 16 ms, respectively—with processing times being nearly the same in different environments. In contrast, the Steger algorithm requires over 300 ms. Regarding memory usage, the three methods demonstrate similar memory consumption when processing images of different environmental conditions: the extreme value method stabilizes at 86.48 MB, the grayscale centroid method at 105.72 MB, and the Steger algorithm fluctuates around 110 MB. The grayscale centroid method exhibits the best stability, making it most suitable for centreline extraction in the digital twin system. During 3D reconstruction, camera capture frequency is positively correlated with reconstruction quality, while movement speed negatively correlates. Each image’s processing time is under 1 ms, showing that the line laser 3D reconstruction method meets the real-time requirements of the digital twin system for subsoiling.