LiDAR (Laser Imaging Detection and Ranging) technology in all its variants (e.g., mobile, terrestrial, unmanned aerial LiDAR) constitutes the primary data acquisition system in the form of point clouds. This type of data presents a computational challenge due to its massive and unstructured nature.
Our company specializes in the comprehensive processing of such datasets, aiming to extract the maximum amount of useful information through the combination of traditional algorithms and artificial intelligence models, both combined with high-performance computing techniques to bring competitive solutions to the market that enable the automation of workflows involving point clouds.
The VirtuaLearn3D (VL3D) framework for point cloud processing based on artificial intelligence methods was developed by our CTO as part of their doctoral thesis at the University of Santiago and their work as a researcher in the 3DGeo group at the University of Heidelberg. This software constitutes one of our fundamental technological pillars, allowing us to develop efficient and specific solutions tailored to each use case in a short period.
At Catallactical, we have a team specialized in advanced computing methods. This allows us to offer our services in a variety of flavors, from delivering processed clouds on our own infrastructure to deploying intelligent cloud systems for automatic point cloud processing, using supercomputers to process massive datasets quickly, or solving tasks more complicated than usual.
Cloud computing is one of the paradigms that has gained the most weight in recent years. Its success is explained by combining pay-as-you-go systems with great scalability in case of increased demand. Thus, deploying intelligent systems in the cloud constitutes a competitive alternative that allows cost control. Therefore, at Catallactical, we offer our clients the possibility of deploying intelligent systems for automatic point cloud processing as cloud services.
Sometimes, point cloud processing requires high computational demand solutions, such as deep learning-based models. These use cases are typically necessary in works of great scientific interest or highly innovative industrial projects. At Catallactical, we are experts in using supercomputers to exploit the most demanding models available in the state of the art, such as hierarchical autoencoder-based neural networks for automatic point-to-point classification.
As part of our innovation strategy, we actively participate in research projects in which we make extensive use of supercomputers, such as the FinisTerrae-III at the Galician Supercomputing Center (CESGA). Specifically, we use this supercomputer in projects such as PrehistorIA, for the automatic detection of prehistoric tumuli (burial grounds) from LiDAR data in collaboration with the University of Santiago de Compostela. We are also using these resources to research, together with the Computer Architecture Group, massive end-to-end processing on a large geographical scale using neural networks from LiDAR data from the second edition of the National Plan of Aerial Orthophotography (PNOA-II).