IMAGE SEGMENTATION AND OBJECT RECOGNITION BASED ON CONVOLUTIONAL NEURAL NETWORK TECHNOLOGY 

Anatoly I. Godunov, Doctor of technical sciences, professor, professor of sub-department of automatics and telemechanics, Penza State University (40 Krasnaya street, Penza, Russia), E-mail: Godunov@pnzgu.ru
Sergey T. Balanyan, Candidate of technical sciences, associate professor, doctoral candidate of sub-department of aviation armament and effectiveness of combat use, Air Force Academy named after professor N. E. Zhukovsky and Yu. A. Gagarin (54A Starykh Bol'shevikov street, Voronezh, Russia), E-mail: bst76@yandex.ru
Pavel S. Egorov, Adjunct of sub-department of aviation armament and effectiveness of combat use, Air Force Academy named after professor N. E. Zhukovsky and Yu. A. Gagarin (54A Starykh Bol'shevikov street, Voronezh, Russia), E-mail: ahtuba.egor@mail.ru 

623.465.7 

10.21685/2307-4205-2021-3-8 

Background. An analysis of the processes of image segmentation is being carried out. An original method of image segmentation using a convolutional neural network is proposed. Materials and methods. A comparative assessment of existing segmentation methods such as threshold segmentation methods: Otsu, Niblack, Bernsen, Savola, as well as the method of image segmentation using a convolutional neural network is carried out. Their advantages and disadvantages are evaluated. Examples of image segmentation by various methods are given. Algorithmic descriptions of segmentation methods are presented. Experiments were carried out to study the segmentation of frames (images) from a given video sequence. Results and conclusions. The results of the experiment, showing the operation of one or another segmentation method, are presented. 

adaptive methods, threshold methods, segmentation, Otsu's method, Niblack's method, Bernsen's method, Savol's method, convolutional neural network