This research theme comprises two research projects, Reconstruction & Monitoring of Time-dependent Biological Systems, and Composite Structure Analysis using Tomography & Volume Correlation.
Reconstruction & Monitoring of Time-dependent Biological Systems
This project aims to develop automated tools for 3D reconstruction of time varying objects from scientific images, and automated tools for statistical analysis of objects and/or their attributes. Such statistical models and analysis tools are required to perform inferences of the object in the future, and thus would aid in scientific conclusions. These tools, although to be constructed to be general and applicable to a wide number of applications, are motivated by a growing number of biological applications. Such applications produce large amounts of indirect measurements of the object (images), and to measure attributes of the object to construct statistical models and perform inferences, 3D reconstruction of the object over time is required. The reconstruction methods we wish to construct will be both for images from ordinary cameras as well as tomographic images used in many scientific studies. This project will be motivated and focus on two applications, one on plant reconstruction and analysis for crop phenotyping, and the other on biofilm reconstruction and analysis for understanding biofouling in membrane filter applications.
Composite Structure Analysis using Tomography & Volume Correlation
The past few decades have seen rapid incorporation of composite structures into all industrial sectors. Experimental characterization tools are key for accurately understanding the behavior of such materials on different scales, and achieving better design for demanding, long-term applications, in which aging (involving heterogeneous mechanisms such as diffusion/reaction) has to be understood in the volume. Digital volume correlation (DIC/DVC) (using optical and X-ray imaging) has been gaining popularity in allowing for a unique observation of material behaviors on different scales. But the core of correlation methods relies on dated image processing algorithms that have since been replaced by more advanced ones in the image processing community.
This project is combines the KAUST CoHMAS lab's expertise of the requirements in the field of composites engineering with KAUST Visual Computing Center's expertise in processing raw image data and in developing optical testing facilities to make a series of advances in the field of non-contact composite characterization. More generally, this project contributes not only to the composite material community, but the whole mechanical engineering / material science community interested in 3D characterization of heterogeneous materials. We develop 3D tomographic reconstruction techniques that take into account the underlying microstructure in their regularizations. For example, in CFRPs (carbon fiber reinforced polymers), the shapes of the fibers will be used to improve the reconstruction technique. By moving away from classical reconstruction algorithms which are linear, through incorporation of more advanced regularization techniques (for example the L1 norm which allows for distinct phases in the structure), and through use of our knowledge of the X-ray opacity of different phases, the reconstruction process can be made more resilient. This means reducing common unwanted features such as noise, beam hardening, and geometric distortions in the final reconstructions. Experimental tests by the CoHMAS lab are used to validate our joint numerical and algorithmic work.