Modeling, Analysis, and Stochastic Simulation of Images (MASSI)

Our projects are involved with Material Science systems.

Active Projects

Fiber Reinforced Materials Feature Extraction

The application of compression / injection molded fiber reinforced polymer composites in automotive and aerospace applications has the potential to decrease weight while maintaining high mechanical properties.

Uncertainties about how the complex microstructure and defects influence the mechanical response and integrity of these materials have prevented widespread adoption of material models for structural applications, as a means of accelerated certification and lightweighting.

We apply the Marked Point Process Combined with Image Processing techniques to model all the objects in the image system.


Voids created during injection molding process	3D reconstruction of Fibers and voids

AFRL Project:

Rare Event Simulation:

For many different types of systems, in areas as varied as science, medicine, engineering, and finance, characterizing the occurrence of rare events is of great importance. In many cases, rare events represent failure modes of a system. Characterization of such events can be incredibly difficult, because it cannot be achieved using typical statitistical methods, which require many repeated observations of a phenomenon.

The key idea is to tailor the importance sampling distribution to preferentially sample from specific parts of the tails of a Gibbs distribution. This means that the question of how the occurrence of a rare event affects an image of a system depends on what factors drive the rare event.

Given an energy function representing the behavior of a system and an energy function representing the factors that lead to a rare event.

$q(x) = frac{1}{Z} exp(-frac{1}{k_BT}(E(x)-gG(x)))$

Where E(x) is an energy function that describes the system, k_B is Boltmann’s constant, represents temperature, G(x) is an importance sampling energy function, is a scalar-valued importance sampling parameter that influences how rare the event to be simulated, and is the partition function.

Completed Projects

Channel Modeling in Dendritic Mixtures:
Hybrid MRF/MPP Model:
3D Ellipse MPP