Thesis & Dissertations
You can find the advised PhD and MS thesis & dissertations below
Ph.D. Dissertations Advised (Major Professor)
1) Robiulhossain Mdrafi
Dissertation: Data-Driven Sparse Computational Imaging with Deep Learning Date of Degree: 5-13-2022 Abstract: Typically, inverse imaging problems deal with the reconstruction of images from the sensor measurements where sensors can take form of any imaging modality like camera, radar, hyperspectral or medical imaging systems. In an ideal scenario, we can reconstruct the images via applying an inversion procedure from these sensors’ measurements, but practical applications have several challenges: the measurement acquisition process is heavily corrupted by the noise, the forward model is not exactly known, and non-linearities or unknown physics of the data acquisition play roles. Hence, perfect inverse function is not exactly known for immaculate image reconstruction. To this end, in this dissertation, I propose an automatic sensing and reconstruction scheme based on deep learning within the compressive sensing (CS) framework to solve the computational imaging problems. Here, I develop a data-driven approach to learn both the measurement matrix and the inverse reconstruction scheme for a given class of signals, such as images. This approach paves the way for end-to-end learning and reconstruction of signals with the aid of cascaded fully connected and multistage convolutional layers with a weighted loss function in an adversarial learning framework. I also propose to extend our analysis to introduce data driven models to directly classify from compressed measurements through joint reconstruction and classification. I develop constrained measurement learning framework and demonstrate higher performance of the proposed approach in the field of typical image reconstruction and hyperspectral image classification tasks. Finally, I also propose a single data driven network that can take and reconstruct images at multiple rates of signal acquisition. In summary, this dissertation proposes novel methods on the data driven measurement acquisition for sparse signal reconstruction and classification, learning measurements for given constraints underlying the requirement of the hardware for different applications, and producing a common data driven platform for learning measurements to reconstruct signals at multiple rates. This dissertation opens the path to the learned sensing systems. The future research can use these proposed data driven approaches as the pivotal factors to accomplish task-specific smart sensors in several real-world applications. |
M.S. Thesis Advised (Major Professor)
|
|
My Ph.D. Dissertation
Title: Feature detection algorithms in computed images
Author: Ali Cafer Gurbuz Advisor: James H. McClellan School of Electrical and Computer Engineering Georgia Institute of Technology Abstract: The problem of sensing a medium by several sensors and retrieving interesting features is a very general one. The basic framework of the problem is generally the same for applications from MRI, tomography, Radar SAR imaging to subsurface imaging, even though the data acquisition processes, sensing geometries and sensed properties are different. In this thesis we introduced a new perspective to the problem of remote sensing and information retrieval by studying the problem of subsurface imaging using GPR and seismic sensors. We have shown that if the sensed medium is sparse in some domain then it can be imaged using many fewer measurements than required by the standard methods. This leads to much lower data acquisition times and better images representing the medium. We have used the ideas from Compressive Sensing, which show that a small number of random measurements about a signal is sufficient to completely characterize it, if the signal is sparse or compressible in some domain. Although we have applied our ideas to the subsurface imaging problem, our results are general and can be extended to other remote sensing applications. A second objective in remote sensing is information retrieval which involves searching for important features in the computed image of the medium. In this thesis we focus on detecting buried structures like pipes, and tunnels in computed GPR or seismic images. The problem of finding these structures in high clutter and noise conditions, and finding them faster than the standard shape detecting methods like the Hough transform is analyzed. One of the most important contributions of this thesis is, where the sensing and the information retrieval stages are unified in a single framework using compressive sensing. Instead of taking lots of standard measurements to compute the image of the medium and search the necessary information in the computed image, a much smaller number of measurements as random projections are taken. The data acquisition and information retrieval stages are unified by using a data model dictionary that connects the information to the sensor data. |