Type of Culminating Activity

Graduate Student Project

Graduation Date

4-2015

Degree Title

Master of Science in Computer Science

Department

Computer Science

Major Advisor

Amit Jain

Comments

Geophysics research involves the study of geophysical events with the help of large amounts of data collected from instruments such as radars, deployed on the ground, aircraft and satellites. Radar signal processing applications analyze radar sensor data to estimate material properties of the subsurface. The amount of data generated is increasing exponentially as hardware improves and more observations are being captured to make accurate predictions. However, the speed of signal processing algorithms has not kept pace with increase in measurement speed making the processing techniques less efficient and time consuming.

In this project a parallel radar signal processing algorithm for snow depth estimation was implemented using the Hadoop framework. This will enable researchers to effectively execute the processing algorithm on existing massive datasets using Hadoop in the near future, and also use Hadoop as a reliable and stable storage system for the large files generated by the algorithm.

To find the efficiency and accuracy of the parallel processing algorithm, experiments were executed on datasets ranging from 0.5GB to 4GB on a Hadoop cluster. The results demonstrated that Hadoop can process the radar signals on large data sets much faster than the currently used sequential signal processing algorithm implemented in MATLAB and can store result files generated by the algorithm more efficiently. The Hadoop algorithm was also deployed on a High Performance Computing cluster to identify the challenges in running in a cluster environment that uses traditional batch job systems.

Download

Share

COinS