A predictive, size-invariant convex shape detector for high-speed and accurate image processing

In this paper a novel convex shape detector is presented suitable for high-speed image processing, regardless of the object´s size. Its primary application is visual servoing in robotics, where an industrial robot is guided using visual information. In addition, it was found that this algorithm is also suitable for high-accurate applications, which was confirmed by experiment and comparison to existing algorithms. The focus of this paper is put on processing the image information rather than preparing the image using a number of morphological operators. The two main objectives are: achieving a high-accurate and high-speed convex shape detector. It is shown that both of these goals were achieved, by position accuracy within 1 pixel and a processing speed as low as 0.1ms.