An adaptive distributed approach for the real-time vision-based navigation system

Real time navigation has remained as a major challenge for guidance and control of robots in indoor applications. In this paper, an adaptive image processing method is suggested to determine the 3D position and velocity of moving objects by using a distributed camera array. In this method objects are detected and 2D localized by independently and asynchronously processing in each camera. Afterwards, the extracted location data of all cameras are fed into a distributed adaptive extended Kalman filter to calculate the 3D position of multiple targets. Improving some specifications such as accuracy, area coverage, robustness against camera noise and camera failure, sampling time and the total price of the system is the main design objectives. The presented data fusion technique enables real-time localization of moving objects in 3D space by using a low-cost camera array. An experimental setup was used to illustrate the performance of the suggested localization system. (C) 2019 Elsevier Ltd. All rights reserved.