Optical Flow
Motion detection using machine vision has been a popular research field in different application areas such as automated video surveillance, medical imaging, and robotics for a long time. Thus, quite different methods have been developed to tackle aspects of the basic problem. A main approach in this research field is the development of methods for the computation of optical flow (OFC). Optical flow is defined as the representation of the projections of 3D motion on a sequence of 2D images. Its computation yields a velocity vector field. The value and the direction of the 2D vectors in this optical flow field represent an approximation of the observed relative motions caused by a possibly moving camera and/or by independently moving objects in the scene. The approach produced efficient solutions for a variety of tasks related to mobile robots, e.g. real-time time-to-contact estimation for obstacle avoidance, multiple motion segmentation, and more.