Computer Graphics and Pattern Recognition

Key Area of Scientific Research

Our company is one of a few companies, which conduct constant research in the field of computer vision, pattern recognition systems, robotics management systems, multimedia data transfer systems, real time video stream recognition systems and interactive 3D scene modeling systems.

Find the Right Solution

Modern information technologies and computer vision systems have opened up new ways for processing the visual information received from video input devices like video cameras, digital photo cameras and scanners. The success in creating efficient pattern recognition systems consists in the development of algorithms for image processing and their implementation in software. This provides solutions for some very difficult problems, such as the following:

Counting the number, size and shape of microbes in water, identifying at the same time their size, color and type and then checking in a comparative database for similar objects as well as renewal of this database;
The analysis of Clinker quality with the recognition of its structure and identity of the object's form and size;
Studying the structure and make up analysis of blood by identifying and counting all immersed objects;
Vehicle Identification Number plate tracking and recognition in real time under the conditions of noise, different illuminance and local obstacles;
Faces recognition in system with restricted access;
Mobile management and control systems for robotics;
Raster image tracing and conversion into vector images.

Ready to Use Solutions

Our research results in ready to use systems of world level, which can be used by our clients in their own products or can be available as ready to use solutions described in “Own Solutions Overview” section. We constantly work improving the systems we have already developed and creating new algorithms and systems. Our solutions are implemented in the ready to use software, and we, on our part, keep on working on broadening the implementation areas of our developed solutions.

Theoretical Background and Methods

In our work we use different methods for pattern recognition and images processing. Here are the most used of them:

Images Normalization (automatic geometric transformation compensation including projection transformation). This set of methods has been developed by our team
Tracking Normalization – methods which allow tracking moving objects and taking different information for object management and recognition. This method has also been developed in our team. It was described in articles available here.
Neural Networks
Segmentation algorithms including those developed in our team
Moving detection based on images difference
Objects recognition based on neural networks, correlations, partial correlations and points of interests
Histograms analysis
Cluster analysis
Differential equations, integrals, mathematical morphology, numerical methods including own solutions used for images processing
Skeleton's building methods developed in our team