Intelligent Multi-Resolution 3D Modeling, Compression, Registration, Fusion and Recognition

NASA missions are being designed for multi-sensor data collection and synthesis using diverse temporal, spatial and spectral resolutions for use by multiple teams of mission specialists and scientists. This project offers two closely related technologies of 3D surface and volume (a) meshing, modeling and compression, and (b) registration and matching that will benefit the distribution, sharing and use of fused datasets. Various machine intelligence and data-driven methods are used to generate a novel 3D meshing, modeling and compression methods based on hierarchical tetrahedral and binary triangular decomposition schemes applicable to surfaces and volumes, providing high compression performance. Meshing is progressive and adaptive to any resolution and for any selected region in the scene/data, thus making it possible to model certain regions with high resolution while leaving the remaining parts coarse. (2) Registration is performed iteratively to search for optimal alignment of overlapping 3D models by minimizing an error objective function subject to an error tolerance. Following alignment, an integration algorithm fuses overlapping models. The anticipated results are effective 3D surface and volume meshing, modeling, compression and registration technologies that significantly benefit NASA's missions.