Cross-section extraction and model reconstruction of lava tube based on L1-medial skeleton

Abstract

Lunar lava tube caves, due to their unique geographical structure and potential as shelters, have repeatedly emerged as prime candidates or priority areas during discussions on lunar base site selection. However, owing to current technological limitations and the complexities of lunar exploration, obtaining detailed data directly from the interiors of lunar lava tubes poses a significant challenge. Hence, the study of terrestrial lava tubes assumes critical importance. By delving into Earth’s lava tubes, we can gain insights into crucial aspects such as their formation mechanisms, geological characteristics, and stability, which can then inform and guide research on lava tubes found on extraterrestrial bodies like the Moon and Mars. During the data acquisition process for lava tubes, given their intricate spatial configurations and dim lighting conditions, we have opted to employ LiDAR technology to collect point cloud data, revealing the internal structural information of the lava tubes. Consequently, this paper thoroughly explores a method for extracting lava tube parameters and constructing models using this point cloud data. This approach encompasses four key steps: data preprocessing, axis construction of the lava tube, cross-sectional feature extraction, and 3D model reconstruction. Initially, raw data is refined through the creation of point cloud voxels to mitigate noise interference. Subsequently, an iterative algorithm is utilized to obtain the L1-medial skeleton, enabling precise capture of the lava tube’s axis information. Using this extracted axis, we proceed with sampling at fixed intervals to derive cross-sectional points of the lava tube. These points are then utilized to extract fundamental geometric parameters such as cross-sectional area, length, and width. Ultimately, by integrating all these cross-sectional point cloud data, a comprehensive 3D model of the lava tube is constructed. Remarkably, this method achieves a high degree of accuracy, with 99 % of the extracted cross-sectional points exhibiting an error of less than 0.12 m, and 99 % of the model’s positional errors falling within 0.15 m. This not only enhances the precision of parameter extraction but also provides significant technical support for in-depth studies of extraterrestrial volcanic activities and lava tube formation mechanisms. Furthermore, it offers a valuable tool for parameter extraction and research on other irregular natural tunnels.