Search

In this paper a problem of tracking multiple robots is addressed. A central vision system is used to find possible robots in the environment using identical colour marks mounted on each of the robots. To distinguish robots a robust tracking algorithm considering robot kinematics is implemented. However, swaps of robots identifications can still happen and are usually unrecoverable. Therefore swaps of robots identifications and initial initialisation problems are tackled using a supervision, which is based on correlation of command, and estimated robots velocity is implemented.

In the paper a global vision scheme applied to a fast dynamic game - robot soccer is presented. The process of robots positions and orientations estimation is divided into two steps. In the first step, the Bayer format image is acquired from camera, then the RGB image is interpolated and pixels are classified into a finite number of classes. At the same time, a segmentation algorithm is used to find corresponding regions belonging to one of the classes. In the second step, all the regions are examined. Selection of the ones that are parts of the observed object is made by means of simple logic procedures. A data filtering is used to improve identified noisy data. The novelty is focused on the optimization of the image acquisition algorithm as well as the processing time needed to finish the estimation of possible object positions.