Muhammad Rzi Abbas Department of Mechatronics and Control Engineering
muhammadrziabbas@uet.edu.pk Lecturer, Mechatronics Dept. University of Engineering and Technology, Lahore • Simple subtraction of images acquired at different instants in time makes motion detection possible, assuming a stationary camera position and constant illumination. • A difference image d(i, j) is a binary image where non-zero values represent image areas with motion, that is, areas where there was a substantial difference between gray-levels in consecutive images f1 and f2 • Noise can be suppressed by thresholding the amount of difference, but this may prevent the detection of slow motion and small object motion. • Result’s of this approach are highly dependent on an object- background contrast. • On the other hand, we can be sure that all the resulting regions in the difference images result from motion. • Trajectories detected using differential image motion analysis may not reveal the direction of the motion. • If direction is needed, construction of a cumulative difference image can solve this problem. • Cumulative difference images contain information about motion direction and other time-related motion properties, and about slow motion and small objects motion as well. • The cumulative difference image is constructed from a sequence of n images, with the first image (f1) being considered a reference image. • A static image is taken as a reference and all subsequent images are subtracted from it to get the difference image. • A problem with this approach may be the impossibility of getting an image of a static reference scene if the motion never ends; then a learning stage must construct the reference image • Subsequent analysis usually determines motion trajectories; often only the center of gravity trajectory is needed • A practical problem is the prediction of the motion trajectory if the object position in several previous images is known • Optical flow reflects the image changes due to motion during a time interval dt, and the optical flow field is the velocity field that represents the three-dimensional motion of object points across a two-dimensional image • It should represent only those motion-related intensity changes in the image that are required in further processing, and all other image changes reflected in the optical flow should be considered errors of flow detection. • For example, optical flow should not be sensitive to illumination changes and motion of unimportant objects (e.g., shadows). • Optical flow computation is based on two assumptions: 1. The observed brightness of any object point is constant over time 2. Nearby points in the image plane move in a similar manner (the velocity smoothness constraint). The partial spatial derivative in the x direction and the partial spatial derivative in the y direction are shown here the partial derivative in time direction is shown here Horn and Schunck Lucas and Kanade • Motion, as it appears in dynamic images, is usually some combination of four basic elements: • Translation at constant distance from the observer. • Translation in depth relative to the observer. • Rotation at constant distance about the view axis. • Rotation of a planar object perpendicular to the view axis. • Motion form recognition is based on the following facts • Translation at constant distance is represented as a set of parallel motion vectors. • Translation in depth forms a set of vectors having a common focus of expansion. • Rotation at constant distance results in a set of concentric motion vectors. • Rotation perpendicular to the view axis forms one or more sets of vectors starting from straight line segments. • Image Processing, Analysis and Machine Vision by Milan Sonka, Vaclav Hlavac and Roger Boyle, 3rd Edition, 2008. • Chapter 16 (Section: 16.1, 16.2, 16.2.1 and 16.2.4)