FULL TEXT IN RUSSIAN


Mekhatronika, Avtomatizatsiya, Upravlenie, 2018, vol. 19, no. 2, pp. 104110
DOI: 10.17587/mau.19.104-110


Features of Pattern Recognition Methods for the Turn Control System of Mobile Robot

T. M. Volosatova, tamaravol@gmail.com, A. V. Kozov, alexey.kozov@gmail.com, Bauman Moscow State Technical University, Moscow 105005, Russian Federation

Corresponding author: Volosatova T. M., Ph. D., Associate Professor, Bauman Moscow State Technical University, Moscow, 105005, Russian Federation,
e-mail: tamaravol@gmail.com

The paper considers the problem of designing the automatic control system of robot based on computer vision. The paper contains the problem formulation of designing the computer vision system to control mobile robot and describes restrictions and basic requirements for control system. The choice of the operating system for mobile robot is described. The paper also contains overview and comparison of recognition methods of the specified object in the image using modern program tools and considering restrictions of mobile computing platforms. The following methods have been analyzed: detection using a marker, pattern matching, image detection and contour matching, feature detections, object classification, method of statistical image analysis, application of artificial neural networks. The paper describes possible ways to implement these methods using OpenCV library. Experiments with OpenCV implementations have been made, and also main advantages and disadvantages of compared methods have been determined. As the result of comparison the method of image segmentation and contour matching has been chosen to be implemented in control system prototype. The paper also describes the implemented prototype of the mobile robot angular velocity control system, which is based on pattern recognition system. The implemented system has been tested in a robotic simulator Gazebo. Research of possibility of distance measurement to a recognized object, as well as research of the effectiveness of different methods combinations can be further directions of development of this work.
Keywords: computer vision, pattern recognition, mobile robot, control system, recognition efficiency

For citation:
Volosatova T. M., Kozov A. V. Features of Pattern Recognition Methods for the Turn Control System of Mobile Robot, Mekhatronika, Avtomatizatsiya, Upravlenie, 2018, vol. 19, no. 2, pp. 104110.

DOI: 10.17587/mau.19.104-110

 

To the contents