Realtime Video Tracking With Modified Lucas Canade With Optical Flow

Abstract

Tracking the objects which are moving is very challenging and interesting area of research. Detection of moving objects and tracking can be done efficiently by optical flow algorithm. Optical flow is superior to all energy difference method. This is a robust algorithm. In this particular work, two different algorithms modified from traditional lucas canade is done. One is the conventional method, in which the flow vectors for each frame has to be found out and thus tracking the moving objects in illumination changes, in shadows and in complex background is possible. To improve the accuracy, a modification to get the highest correlation in Eigen values is added to the conventional method and is implemented. This helps to reduce the computation cost and supports to track on-time videos. The comparison of both methods is done. The tracking outputs are shown with the standard state of art method which is the ground truth result for this tracking. More information about the moving target is additionally found in the modified approach and the betterness of the algorithm is proved in MATLAB outputs and tested in Standard PETS dataset.