Investigation on Photogrammetry and LiDAR Models for Caves/Mines 3D Reconstruction

The study and 3D reconstruction of surfaces and objects can be performed using different types of sensors, each one with a particular reconstruction method. In the present work, we analyze and define the best combination of sensors and algorithms to perform 3D reconstruction of unstructured indoor environments with unnatural lightings, such as caves and underground mines. Reproducibility, fidelity, and cost-effectiveness are the criteria used for selecting an algorithm. As the environments analyzed are large, over 250 m³, methods based on volume and surface measurements were chosen, using readings obtained by monocular camera and LiDAR. LiDAR is used in conjunction with cameras to allow the reconstruction of realistic textures. We propose to perform a two-step analysis: (i) photogrammetry with a monocular camera, and (ii) obtain and merge the point cloud using a LiDAR. At the first stage of the work, we tested the most popular algorithm for features detection and extraction accuracy: SIFT, BRISK, ORB, AKAZE, and SURF, at the environments defined in the scope of this work. Then, the second reading of the environment was performed to obtain the point cloud through LiDAR. Finally, the complete reconstruction of the site is performed by merging the obtained data. The validation test of the methodology were performed in a deactivated gold mine in Ouro Preto - Brazil, known as "Mina du Veloso" (latitude: -20.376360, longitude: -43.516036), which allowed to define the most accurate extraction algorithm features, as well as showing that the combination of approaches enabled faithful 3D reconstruction. The resulting map of this touristic place enabled people with mobility problems to get to know it through virtual reality immersions, as well as providing data for speleological studies. The methodology developed in this work can also be applied to other areas, such as the industrial sector, academic and governmental for accurate 3D reconstructions of the environment.