Senzorický subsystém mobilního robotu

Abstract

The thesis deals with a sensor subsystem of mobile robots. The sensor subsystem described in this thesis is in the category of systems providing visual perception of the environment around a mobile robot to its operator. It is a subsystem which makes stereovision perception to the operator. The thesis is divided into two main parts. The first part deals with stereovision principles and algorithms and the second part deals with video transmission from the robot to the operator. The first part starts with description of camera calibration. Here is explained everything from elimination of camera distortion, to stereocalibration and rectification. In this thesis was developed an innovated method of depth focusing without mechanical turning of the cameras – the need of mechanical turning was substituted by a mathematical image transformation. Then there is described a designed system of automatic focusing to a specific depth determined based on the region of interest of operator’s view. This system of automatic depth focusing is especially useful for stereovision. Further there are presented some tests of filters of the automatically determined depth. The goal of these filters is to remove noise in the focusing depth determination. The next main part of the thesis presents a system for image transmission from the robot to the operator. Here are analyzed existing commercial devices for video transmission. Then the principle of a developed system of image transmission is described. The priority of this system is to provide as actual images as possible. The developed software solves this problem by capturing a new image only after the last image has been already displayed. Each image is compressed to increase the transmission speed. The last part contains some tests of this system and presents the results. The test were done on Atom 1.6 GHz processor on the transmitting side and Pentium 4 3.2 GHz processor on the receiving side, both equipped with 2 GB RAM. The transmission was realized via LAN, which was wireless for our purpose.