Research Assistant

October 2002 - January 2006, Optoelektronika (since October 2003 - Yansar), Belarus, Minsk

Software Engineer / Senior engineer 

Team size: 7 persons in Video analysis department. 

This job was related to theodolites. Theodolite is a device containing optical channel (video camera). Video camera with the help of engines follow the target. The electronic device (coordinator) must process the video from camera and give the control commands to camera engines. Our department designed video cameras and coordinators. Department consisted of 1 Circuit Design Engineer, 3 PCB Design Engineers, 2 FPGA Programmer, 1 Algorithm/Software Engineer (me).

At the beginning I had a task to do a target-tracking program. The program would be not a real-time but precise.  The program had to operate on personal computer and process video written from video camera. First of all I chose feature-based methods, Harrison corner detection and corner based optical flow calculation, but they were not very stable (robust).  Than I rewrote the program using correlation based methods. The program was realized using Visual C++ 6.0. I used OpenGL for video output, and also Intel Integrated Primitives for speed optimization. 

The goal of the next task was to design a video camera. I developed algorithms of automatic and manual contrast, brightness and histogramm enhancement and they were programmed in camera. I also calibrated the camera for radial lens distortion compensation. During this and the next task I teamed up with a FPGA programmer. The programmer gave me algorithms restrictions. These restrictions were because of using of standards libraries by the programmer.

The goal of the next task was to design a coordinator. This device is PCI board with which video camera and high-speed data storage device are connected. Coordinator had to process 1024x1024 120 fps grayscale video and detected, tracked and recognized up to 4 targets with size from 12x12 to 512x512 pixels moved with high speed. At the beginning I analyzed the current analogs. Then I did a Matlab model of current analogs realized algorithms. They are: Centroid Tracking, Rotating Edge Tracking, SSD/MAD Correlation. This model worked with video files.  Using this model we estimated computational complexity of the task. So to realize this task we chose Virtex II Pro chip. During the coordinator designing I changed model algorithms for their implementation in FPGA. I also realized and tested other more robust target tracking algorithms. They are: Ravela Correlation (based on articles by S.Ravela), RBF neural network detector (based on ZISC processor algorithm). For this I wrote a computer program working on PC and processing real-time video from video camera. This program was written on Visual C++ 6.0. I used DirctX SDK for video capture. Than I realized a target recognition algorithms. There I used Principal Component Analysis and Support Vector Machines.

At the beginning of 2005 I was promoted and I had a computer programmer as my assistant.

Software: C/C++, MS Visual C++ 6.0; Win32API; DirectX SDK; Intel Integrated Primitives; Matlab; OpenGL.

Operating systems: Windows 2000, XP.

March 2001 – September 2002, Belstats, Belarus, Minsk

Software Engineer 

Team size: 5 computer programmers

The firm did offshore programming. The customer firm was American startup. I reengineered an alien code. Customer firm gave the code. This code was a source of a financial analysis library. I am as a specialist of DSP rewrote algorithms of financial data processing with comments about their behavior (time lags, signal/noise ratio etc). Also I designed original time series adaptive smoother. 

Software: C/C++, MS Visual C++ 6.0; Matlab.

Operating systems: Windows 2000.

March 1999 – March 2001, United Institute of Informatics Problems of National Academy of Science of Belarus, Belarus, Minsk.

Junior Research Assistant

Team size: 3 researchers

The device, which we developed, consisted of digitizer and flatbed scanner. The scanner contains two digitizer’s pointers. The document is placing on a digitizer and than is scanning by a scanner with positioning by the digitizer. After that we have images and their coordinates in digitizer reference system. My task was to write a program that would automatically stitch images in one big raster using the digitizer information and without this information. I realized several methods: Fourier-Mellin transform, Harris corner based stitching. Also scanner warps images. So I designed algorithms for restoration of the images. I realized Beier-Neely (the algorithms like http://www.hammerhead.com/thad/morph.html), Free Form Deformation, affine (triangulation based) transform. Theoretical part of my PhD was connected with image stitching too.

I have been awarded with the diploma for high achievements in a science work by United Institute of Informatics Problems of National Academy of Science of Belarus in 2001 and 2002 years. 

Software: C/C++, MS Visual C++ 6.0; Win32API; Matlab.

Operating systems: Windows 98.

EDUCATION:

October 1997 – October 2000, 

United Institute of Informatics Problems, National Academy of Science of Belarus, Belarus, Minsk. 

Postgraduate student.

Theme: Automatic fuzzy edge image stitching.

In my science work I realized automatic fuzzy edge image stitching algorithm using edge enhancement and corner features. I have 3 publications in magazines and I took participations in 5 conferences.

Scientific Papers:

1. Combined input of sizeable images in a computer. Proceedings of National Academy of Science of Belarus. Series of Physical-Engineering Sciences (in Russian).

2. Geometrical and brightness distortions at optical input of graphic information and their correction. Informatics, United Institute of Informatics Problems NASB (in Russian).

3. Combined input of sizeable images in a computer. New solutions. (in Russian).

September 1989 – June 1996  

Belarussian State University Informatics and Radioelectronics, Faculty of Telecommunication, Belarus, Minsk.

Student, Master’s degree in Telecommunication Engineering.

SKILLS:

Software Languages/ Engineering Tools

· C/C++;

· MS Visual C++;

· Win32API; 

· Matlab;

· OpenGL;

· DirectX.

Image Processing/Recognition methods

· Neural Networks;

· Support Vector Machines;

· Principal Component Analysis;

· Statistical Learning;

· Pattern Recognition;

· Wavelets decomposition;

· FFT, DCT;

· Image Enhancement, Histogram Equalization;

· Segmentation;

· Morphological and Set Operators;

· Optical Flow;

· Fourier Mellin transform;

· Image Warping;

· Camera Calibration.

English: well

PERSONAL:

· Strong analytical skills with the ability to take a problem analyze it and develop a computerized solution;

· Work well under stressful situations;

· Abilities to learn quickly, improve professional skills;

· MENSA member (http://www.mensa.org). 

HOBBIES:

· Puzzles, logic games;

· Outdoors: hiking.