Accurate 3D mapping and immersive visualization for Search and Rescue

P. Musialik,K. Majek,Piotr Majek,M. Pełka,J. Będkowski,A. Maslowski,A. Typiak
DOI: https://doi.org/10.1109/RoMoCo.2015.7219728
2015-07-06
Abstract:This paper concentrates on the topic of gathering, processing and presenting 3D data for use in Search and Rescue operations. The data are gathered by unmanned ground platforms, in form of 3D point clouds. The clouds are matched and transformed into a consistent, highly accurate 3D model. The paper describes the pipeline for such matching based on Iterative Closest Point algorithm supported by loop closing done with LUM method. The pipeline was implemented for parallel computation with Nvidia CUDA, which leads to higher matching accuracy and lower computation time. An analysis of performance for multiple GPUs is presented. The second problem discussed in the paper is immersive visualization of 3d data for search and rescue personnel. Five strategies are discussed: plain 3D point cloud, hypsometry, normal vectors, space descriptors and an approach based on light simulation through the use of NVIDIA OptiX Ray Tracing Engine. The results from each strategy were shown to end users for validation. The paper discusses the feedback given. The results of the research are used in the development of a support module for ICARUS project.
What problem does this paper attempt to address?