My main research involves using computer graphics, computer vision, and deep learning for structural health monitoring, post disaster response, and digital twins. I also conduct research in CEE engineering education. Below are the work I’ve done for different projects.
Physics-based & Heuristics-based Damage for Digital Twins
The purpose of this research is to design and develop methods to create realistic damage of digital twins. These damaged digital twins can then be used for a variety of purposes. These include generating synthetic data for a deep learning algorithm, computer vision research, and simulations. The damage generated needs to not only look realistic but also be consistent with the physics behind it.
My main contribution for this paper was the design and development of physics-based damage and heuristics-based damage creation methodologies. Physics-based damage is directly based on finite element model (FEM) damage hotspots, while heuristics-based damage is based on structural analysis results. My co-authors ran the actual analyses, while I used both of the results from their work as inputs to my work. Additionally, I created a method to turn an image of an as-built structure’s material into undamaged textures that can be applied to a digital twin.
Vision-based Displacement Measurement of Miter Gates
The purpose of this research is to measure displacements of miter gates through a series of images using computer vision. This saves time and money as well as helps automate the process.
My main contribution to this paper was as the main developer and algorithm designer for a series of MATLAB scripts that use computer vision to measure displacement of miter gates through a series of photos. The main script uses a KLT-based optical flow method to measure the displacements of a set of user-specified points. Some of the other scripts visualize the generated displacements as a heat map on top of the original image and animate the displacements in a plot for easy visualization.