Vinay Pandit

From Msplacements

IIIT –Hyderabad, Gachibowli, Hyderabad – 500 032, 09959730897(cell), email: vinaygpandit@gmail.com

Educational Background:

  • MS by Research in Computer Science from Lab for Spatial Informatics, IIIT – Hyderabad (http://iiit.ac.in), during 2007-2009 (with CGPA of 8.4).
  • BE in Electronics and Communications from MMEC Belgaum, under Visvesvaraya Technological University (http://www.vtu.ac.in), during 1999-2003 (Grade: First Class).

 

Work Experience Summary:

  • Four years of industry work experience in debugger design and development.
  • Worked for Tensilica Technologies India Pvt Ltd, Pune (http://tensilica.com), as software engineer from December 2004 till July 2007. Joined the BOT unit for Tensilica Inc. under eInfochips Ltd. in December 2004 and transferred to Tensilica in January 2006.
  • Worked for GalieoSoft Pvt. Ltd., Bangalore (http://www.galieosoft.com), as software engineer from September 2003 till December 2004.

 

Research Interests:

  • Image Processing (Image Texture Analysis, Image Segmentation)
  • System Software Design

 

Primary Research and MS Thesis:

Research Topic: Road detection and extraction from satellite images

Advisor: Dr. K. S. Rajan

  • Application of image texture analysis to road extraction is explored.
  • Framework for road extraction based on texture matching is proposed.
  • A scheme to detect road seed points using multi-temporal images is proposed.

Short Description:

The Goal of the thesis is to develop a road extraction system which 1) functions with no or minimal manual intervention 2) functions in scenes with various road surface characteristics (texture) 3) functions for images of various spatial resolutions 4) functions in scenes with varying road widths and curvatures and 5) functions even in the presence of obstacles like vehicles. In order to develop road extraction technique independent of spectral and special resolution of the input image the road templates and other road model parameter need to be derived from same input image. Hence, we have developed a framework where road texture can be analyzed and compared in terms of arbitrary texture descriptors. Along with utilizing various texture descriptors present in the literature we have experimented a simple k-means based descriptors in our road extraction framework. In the thesis, we have also explored the utility of multi-temporal images for road extraction. We have observed that there exists a useful relationship between locations where a typical road tracking methods based on region growing fails (shadow, vehicles) and the locations of maxima in the difference in a pair of closely dated multi-temporal images of a same geographic area. Based on this observation, we hypothesised that vehicles can be detected and can be eliminated by analyzing difference in such pair. Also the algorithm was successful in detecting road seed points and to automate the road extraction system. We have demonstrated the robustness of our algorithm by experimenting on different input image types including IKONOS, CARTOSAT-2 and Google Earth images. Our algorithm not only extracted road with higher accuracy but also extracted road regions of different widths and curvature. A demonstration of the working of the algorithm can be found in http://research.iiit.ac.in/~vinaypandit/Demo.rar and the thesis(under review) can be found in http://research.iiit.ac.in/~vinaypandit/Sample_Results.pdf.

 Road Network Extraction Using Multi-temporal Satellite Images:

The revisit capabilities of modern remote sensing satellites provide temporal data that can convey valuable information. The difference between two such consecutive revisits can be a major cue of the location of non-stationary objects like vehicles on the road. In this work such difference image is processed through series of morphological operations to detect the vehicles. Road seeds are identified in the neighborhood of detected vehicles and are used further in Fourier descriptor based road tracking algorithm (existing) to extract the road network. A pair of 0.8 meters panchromatic CARTOSAT-2 images was used for the experiments.

This work(http://research.iiit.ac.in/~vinaypandit/IGARSS.pdf) has been submitted to IEEE IGARSS – 2009 (http://www.igarss09.org) and is accepted for oral presentation and publication in the symposium proceedings.

Color Based Urban Scene Classification:

Proposed algorithm in this work, classifies the given urban scene into buildings, roads, vehicles and natural objects like vegetation and water bodies. Classified building region is further processed for delineating individual buildings. The algorithm starts with approximating the image in Hue-Saturation-Value (HSV) color space which is further used in region growing based segmentation. Segmentation accuracy for buildings and road is refined using their geometrical features. Spectral information is used to detect vegetation, shadow and water. Experiments were carried out on a Quickbird image comprising of 61cm panchromatic image and 2.44 meter multispectral (4 bands) images.

This work(http://research.iiit.ac.in/~vinaypandit/IPCV.pdf) has been submitted to IPCV-2009 (http://www.world-academy-of-science.org/worldcomp09/ws/conferences/ipcv09) and is accepted for oral presentation and publication in the conference proceedings (Not registered).

 

Publications:

  • Vinay Pandit, Sudhir Gupta and K.S.Rajan, "Automatic road network extraction using high resolution multi-temporal satellite images", Proceedings of IEEE IGARSS 2009.
  • Sudhir Gupta, Vinay Pandit, and K.S.Rajan, "Remote sensing based season calendar for Indian districts using MODIS data", Proceedings of IEEE IGARSS 2009.
  • Vinay Pandit, Sudhir Gupta and K.S.Rajan, "Color based urban scene classification using high resolution satellite Imagery". Accepted for IPCV-2009 conference proceedings but not presented.
  • Sudhir Gupta, Vinay Pandit, and K.S.Rajan, "Suitability mapping for locating a Special Economic Zone”, Proceedings of Remote Sensing and Photogrammetry Conference, 2008.

  

Related Course work:

  • Digital image processing
  • Medical image processing
  • Spatial informatics
  • Computer vision
  • Pattern recognition

  

Additional Academic Activities:

  • Teaching assistance for Digital Image Processing course conducted by Dr Jayanthi Sivaswami. Class consisted of both BTech and MS students.
  • Participated in Cansat competition (building can sized satellite) organized jointly by American Astronautical Society and NASA. Our team was the first Indian team to participate and stood at 8th position. Developed ground control station user interface.

  

Work Experience Details:

 Tensilica Technologies India Pvt. Ltd:

 Xtensa Xplorer: Tensilica’s Xtensa technology provides SOC designers with the configurable and extendible processor core supported by automatic hardware and software generation.  Xtensa Xplorer, an Eclipse based IDE with CDT plug-ins (C/C++ Development Tooling), integrates software development, processor optimization and multi-processor SOC architecture tools into one common design environment. It also integrates SOC simulation and analysis tools. Xtensa Xplorer helps in development of TIE (Tensilica Instruction Extension) instructions that maximize performance for a particular application.

 Important Contributions:

  • Designing and developing a source locator, which picks appropriate source files during debugging.
  • Designing and developing multi-processor launch, this helps in launching user developed simulator.
  • Enhancing variable view to persist previously set configurations (submitted to CDT, https://bugs.eclipse.org/bugs/show_bug.cgi?id=172132).
  • Enhancing memory view to support load/restore memory contents.

 

GalieoSoft Pvt. Ltd:

Galieo Development Suite: The Galieo Development Suite (GDS), developed using C and Java (Swings), provides complete solution for embedded system engineers working on ARM processor based systems, by providing a wide range of tools like project manager, fully featured Editor, powerful ARM compiler and feature packed Debugger. The GalieoSD debugger integrated in GDS, helps the embedded system engineers in doing source level debugging of applications written in high level languages like C and C++. It supports JTAG emulators like MAJIC, Multi-ICE, ACE and Abatron.

 Important Contributions:

  • Interfacing ACE emulator to GalieoSD.
  • Extracting debugging information from ELF image.
  • Enhancing Debugging Features.
  • Interfacing Serial Debug Monitor to GDS.
  • Designing and developing user interface for GDS.

 

Skill Sets:

Programming Languages Known:

  • C
  • Java (Eclipse Plug-in Development, Swings)
  • Matlab

 

 

 

Retrieved from "http://www.iiit.ac.in/msplacements/index.php/Vinay_Pandit"
Personal tools