Hierarchical DSfM for structural condition assessment

  • Developed a novel method for image-based 3D reconstruction of large infrastructure systems capable of resolving 0.1 mm details, an order of magnitude improvement over existing methods

  • Through designing a multi-scale imaging network and hierarchical image-matching technique, the developed method is capable of registering images at a much higher rate in comparison with conventional methods

  • By using a memory efficient global pixel-wise image matching method, it is possible to generate highly dense 3D point clouds on a single computer

  • This method performs robustly in generating accurate and complete 3D models of large-scale structural models with the ability to control the local point density at critical locations 

Related Publications

GPU Grant Program