Machine Learning for Object Recognition from High Volume Radio Frequency Data

Share

Much research efforts have been devoted to applying machine learning algorithms in video imagery for object recognition. However, very limited open literatures can be found on machine learning in radio frequency data. Hence, this research explores application of deep learning algorithms to the task of Automatic Target Recognition (ATR) in synthetic aperture radar (SAR) imagery. Radar enables imaging ground objects at far greater standoff distances. However, the false-alarm rate of both human and machine-based radar image recognition is unacceptably high. Existing ATR algorithms also require impractically large computing resources for airborne applications. The goal of our research is to advance the state-of-the-art ATR capability by developing a more accurate, real-time, and low-power object recognition system. We implemented Convolution Neural Network (CNN) based SAR object recognition algorithms in GPU and energy efficient computing systems. We received acceptable classification accuracy on relevant SAR data. We will discuss technical challenges and future research on radio frequency object recognition.



  Date and Time

  Location

  Contact

  Registration



  • 152 Lamb Memorial Dr.
  • Rochester, New York
  • United States 14623
  • Building: Golisano Hall (GOL)
  • Room Number: 1435

Staticmap?size=250x200&sensor=false&zoom=14&markers=43.0860201%2c 77
  • Starts 06 March 2017 12:00 AM
  • Ends 05 May 2017 12:00 AM
  • All times are US/Eastern
  • No Admission Charge
  • Register


  Speakers

Dr. Uttam Majumder

Dr. Uttam Majumder

Topic:

Machine Learning for Object Recognition from High Volume Radio Frequency Data

Much research efforts have been devoted to applying machine learning algorithms in video imagery for object recognition. However, very limited open literatures can be found on machine learning in radio frequency data. Hence, this research explores application of deep learning algorithms to the task of Automatic Target Recognition (ATR) in synthetic aperture radar (SAR) imagery. Radar enables imaging ground objects at far greater standoff distances. However, the false-alarm rate of both human and machine-based radar image recognition is unacceptably high. Existing ATR algorithms also require impractically large computing resources for airborne applications. The goal of our research is to advance the state-of-the-art ATR capability by developing a more accurate, real-time, and low-power object recognition system. We implemented Convolution Neural Network (CNN) based SAR object recognition algorithms in GPU and energy efficient computing systems. We received acceptable classification accuracy on relevant SAR data. We will discuss technical challenges and future research on radio frequency object recognition.

Biography:

Dr. Uttam K. Majumder is a senior electronics engineer at Air Force Research Laboratory (AFRL) and have been working since August 2003. His research interest includes Machine Learning for object recognition, High Performance Computing, Synthetic Aperture Radar (SAR) algorithm development for surveillance applications, Radar Waveforms Design and Digital Image Processing. Among various awards, Dr. Majumder received Air Force Distinguished Civilian Award and AFRL Science and Technology achievement award for radar systems development.


Dr. Majumder received Bachelor of Science (BS) degree from the Department of Computer Science, The City College of New York (CCNY) in June 2003 graduating in Summa Cum Laude. He earned an MS degree from Air Force Institute of Technology (2007) and a Ph.D. degree (in Electrical Engineering) from Purdue University (2014), West Lafayette, Indiana. He also earned an MBA degree from the Wright State University (2009). He was an adjunct faculty at Wright State University, Dayton, Ohio. He developed a graduate level course on SAR Signal and Image Processing. He lectured a short course on “SAR Signal Processing” at IEEE International Radar Conference 2010, Washington D.C. Dr. Majumder published more than 30 technical papers (journal and conference) in various technical proceedings and submitted a patent application on “Orthogonal Waveforms Design”. Dr. Majumder is a senior member of IEEE.

Dr. Uttam Majumder

Topic:

Machine Learning for Object Recognition from High Volume Radio Frequency Data

Biography:





Agenda

11:30 AM Social/Pizza

12:00 PM Presentation