29 April 2022 in Vestnik of Astrakhan State Technical University. Series: Management, computer science and informatics
The article focuses on using artificial neural networks for the recognition of commercial hydrobionts in conditions of increased fluctuation. Research has not been conducted in the field of recognition of aquaculture objects for reasons of solving the applied problems and achieving the required indicators. There is described the development of software for recognition of aquaculture objects based on the YOLOv5 model, as well as the proper methodology for the semi-automatic formation of effective visual training samples. There has been analyzed the model operation scheme and substantiated the importance of the training sample and its influence on the quality of pattern recognition by a neural network. Effective identification of aquatic organisms requires the methods for dividing it into zones according to various distinctive features and separating the required objects not only from the background, but also taking into account fluctuations above the water surface. There has been developed a method for eliminating fluctuations in order to increase the efficiency of recognition of aquatic organisms by a neural network. There is described the process of testing software for the detection of commercial hydrobionts trained on a sample by the proposed method. It has been inferred that the recognition rate directly depends not only on the quality of the training sample and the volume of images, but also on the technical characteristics of the electronic computer
1 January 2014 in Communications in Computer and Information Science
1 October 2013 in Applied Mechanics and Materials
The problem of control of systems with structural and parametric uncertainty for the case when derivatives of input and output parameters cannot be measured is solved. The operability of the designed control systems is modeled for anthropomorphic robots manipulator. Selected example demonstrates the compensation of effect of bounded structure and parametric disturbances. As in the control action organization the effect on subsystem is not allowed, i.e. obtained control system is completely decentralized.
30 April 2021 in Vestnik of Astrakhan State Technical University. Series: Management, computer science and informatics
28 July 2023 in Vestnik of Astrakhan State Technical University. Series: Management, computer science and informatics
1 January 2014 in Communications in Computer and Information Science