Vision System for Relative Motion Estimation from Optical Flow

L.A. Borkovskaya, A.V. Borkovskiy, А.А. Ilnitskaya, D.A. Tupa

This paper talks about the vision systems that allow the receipt of information on relative object motion in real time. It is shown, that the algorithms solving a wide range of practical problems by definition of relative movement can be generated on the basis of the known algorithms of an optical flow calculation. Works in the field of an optical flow calculation have been conducted for more than 30 years. Last decade these methods were used in a wide range of applied problems due to increase of computers computing capacity and the occurrence of specialized graphic processors. There are many articles that have been written on the subject of optical flow methods. There are also widely available libraries with an open code, in which the ready applications of the most popular optical flow methods could be found (for example, OpenCV, LTI-Lib, VXL). Methods of an optical flow appear to be useful for segmentation of images, and also for detection of obstacles from moving objects.

Key words: robotic vision, software package, segmentation of images, optical flow, OpenCV