03/15/2022
By Sokny Long
The Francis College of Engineering, Department of Mechanical Engineering, will hold a Doctoral Dissertation defense by Daniel Rohe on “Image Processing Techniques for Structural Dynamics Testing using Radiographic Images.”
Ph.D. Candidate Name: Dan Rohe
Defense Date: Tuesday, March 29, 2022
Time: 10 a.m. to noon EST
Location: This will be a virtual defense via Zoom. Those interested in attending should email Zhu_Mao@uml.edu at least 24 hours prior to the defense to request access to the meeting.
Committee Chair (Advisor): Zhu Mao, Ph.D., Associate Professor, Department of Mechanical Engineering, University of Massachusetts Lowell
Committee Members
- Peter Avitabile, Ph.D., Professor Emeritus, Department of Mechanical Engineering, University of Massachusetts Lowell
- Alessandro Sabato, Ph.D., Assistant Professor, Department of Mechanical Engineering, University of Massachusetts Lowell
- D. Gregory Tipton, Ph.D., Senior Scientist, Sandia National Laboratories
Brief Abstract: Photogrammetric testing methods for structural dynamics applications have become widespread in recent years due to the reduced cost and increased resolution of high speed cameras, as well as the increase in computational capabilities required to store and process the vast quantities of data that can be obtained from a single test. While feature tracking and DIC techniques remain the most popular, these rely on a contrast patterns applied to the test article which can be tracked by the respective algorithms. An alternative technique, phase-based image processing, uses the part's own local contrast pattern to extract motions, which are derived by filtering images using a set of complex spatial filters which isolate spatial frequencies and orientations in the image.
While photogrammetric testing techniques can be fielded more quickly and acquire more data than, for example, a traditional accelerometer or roving hammer test, there remains the limitation that these techniques can only measure the visible surfaces of test articles; if there is an article with significant dynamic responses deriving from the motion of internal components, these techniques may not achieve a characterization of the entire test article from only surface measurements. One option to overcome this limitation is to use radiographic imaging. However, radiographic images introduce their own difficulties for image processing. Applying a desired local contrast pattern may be difficult or impossible for a given radiographic test, so any displacement extraction will need to rely on local contrast in the images themselves. Radiographic images also tend to be noisier than visible light images. Finally, because each pixel now represents a line through the part rather than a point measurement on the surface, there may be multiple displacements present at each pixel.
This work develops the phase-based processing approach to extract motions from radiographic images. The dissertation will develop an approach to generating synthetic images, which can be used to develop and test image processing algorithms while providing a “truth” result against which the algorithm's result can be compared. The ability to perform experimental modal analysis using the degrees of freedom from the phase-based processing approach will be demonstrated. The dissertation will develop algorithms for extracting 3D motions from a set of images using finite element expansion techniques, which can extrapolate motion from a single set of images, or interpolate motion in regions of the image with low contrast. Finally, the developed techniques will be applied to synthetic radiographic images to demonstrate the suitability of the approaches to the stated application.
All interested students and faculty members are invited to attend virtually.