Algorithm of Character Recognition
Moscow Scientific Research Television Institute, Moscow, Russia
March 2013 – June 2014
Overview
Designed and implemented a character recognition algorithm for extracting telemetry data from satellite TV images through the detection of binary character patterns.
The algorithm segments characters using contour analysis and determines similarity between input and template images via cross-correlation with FFT. The system, written primarily in C++ with OpenCV, processes real-time video streams and stores recognition results in an SQL database.
Implemented multithreaded processing for concurrent image handling, improving throughput and responsiveness. The algorithm’s complexity and performance were benchmarked against other image recognition methods.
The application was developed within the Qt Creator IDE for Windows, featuring a graphical interface built with Qt and Qwt for data visualization.
Key Achievements
- Developed a C++ character recognition algorithm for satellite TV telemetry data extraction.
- Implemented contour-based segmentation and FFT-based cross-correlation for pattern matching.
- Utilized OpenCV for image preprocessing and analysis steps.
- Integrated SQL database for storing recognition results and metadata.
- Optimized algorithm performance with multithreaded processing of video streams.
- Designed visualization and configuration UI using Qt and Qwt frameworks.