PROJECT TITLE :

Distribution Driven Extraction and Tracking of Features for Time-varying Data Analysis

ABSTRACT:

Effective analysis of options in time-varying data is essential in varied scientific applications. Feature extraction and tracking are 2 necessary tasks scientists rely upon to induce insights regarding the dynamic nature of the big scale time-varying information. However, usually the complexity of the scientific phenomena solely allows scientists to vaguely define their feature of interest. Furthermore, such options will have varying motion patterns and dynamic evolution over time. Therefore, automatic extraction and tracking of options becomes a non-trivial task. During this work, we have a tendency to investigate these problems and propose a distribution driven approach which permits us to construct novel algorithms for reliable feature extraction and tracking with high confidence within the absence of accurate feature definition. We have a tendency to exploit 2 key properties of an object, motion and similarity to the target feature, and fuse the knowledge gained from them to come up with a sturdy feature-aware classification field at every time step. Tracking of options is done using such classified fields that enhances the accuracy and robustness of the proposed algorithm. The efficacy of our method is demonstrated by successfully applying it on many scientific information sets containing a wide range of dynamic time-varying features.