The usage of the markers is world-wide in object following. There are several variables existing, from the simple crosses, through the ones seen in infra-red light, to the complex pattern, which can be difficult to detect in the environment.
The existing procedures aren’t enough efficient for the embedded systems with low performance. The ARToolkit, or the robust SIFT and SURF need far too much energy for detecting a simple marker. By the way at the color-based detecting we needed a environment-independent solution.
The most expedient denouement was to create an own detecting algorithm. The method neglects all of the complicated or too long functions. Of course there are computationally intensive functions in my solution too, but the algorithm tightens the conglomeration of the potential markers before them. Filters the too small and too big objects, also the non-square shaped polygons. At the end of the filtering the solution detects the marker and its inner structure with a low resource-intensive sampling method.
The OpenCV library has lots of functions that helps the detection of the objects and also following the founded markers. In the last algorithm of my thesis I use the 2.4.2 version of OpenCV, with the C++ interface, which contains several pre-implemented functions.