Software Journal:
Theory and Applications

Send article

Entrance Registration

Our congratulations to Aleksandr Nikolaevich Sotnikov on his 70th anniversary!

[29.07.2021]

Sotnikov Aleksandr Nikolaevich - Dr.Sc. (Physics and Mathematics), Professor, an Honored Scientist of the Russian Federation, a Deputy Director of the Interdepartmental Supercomputer Center of RAS, the Editor-in-Chief of the “Software Journal: Theory and Applications”, a member of the Editorial Board of journal “Software & Systems”, an author, a permanent expert and a great friend of the Editorial Board. We wish you the best of health, prosperity, achievement of the most cherished goals!

Thank you so much for your advertency, sensitivity and help.

All ads...

Neural models of recognition components of complex structure in computer vision systems

M.V. Semina (mvsyomina@gmail.com) Moscow Aviation Institute (National Research University), the Chair of applied informatics (Student), Moscow, Russian Federation;
E.S. Ageshin (ageshin.e@mail.ru) Moscow Aviation Institute (National Research University), the Chair of production engineering of the flight-type engine (Lecturer Assistant), Moscow, Russian Federation;

This paper presents an experiment in the computer vision zone aimed at automatization the neural network training to recognize industrial objects on the example of turbo pump unit parts for the RE-120 rocket engine.
To train the neural network, the authors used both a data set consisting of photos of already existing parts
and a set of images from a CAD-program that simulates the design stage of the required product.

By comparison of test results, it confirmed the hypothesis it is possible to train computer vision systems to distinguish not yet existing objects based on screenshots of their digital counterparts (CAD-models). By collecting the required data before direct production of the product, it is possible to achieve good recognition rates even for an actual object with simple geometry.

The paper presents the results of the application of this method in comparison with the traditional teaching approach, and also considers the perspectives for using this technology in the industry.