MULTI-OBJECT TRACKING USING ST-MRF, GMM, MODIFIED RUNNING AVERAGE AND CAMSHIFT - A COMPARATIVE STUDY
Divya Divya, Padmanayana · IJITR International Journal of Innovative Technology and Research - IJITR International Journal of Innovative Technology and Research · 2015
V ideo-based object tracking in static or in dynamic scenes is one of the challenging problems with vast variety of applications, is currently one of the most active research topics in computer vision. This paper mainly focuses on performing survey on tracking moving objects in video scenes in both pixel-domain and compressed-domain with detailed descriptions of tracking strategies and examining their pros and cons. Survey of tracking methodologies in both pixel and compressed domain for object recognition and tracking includes modified running average, Gaussian Mixture Model, Spatial-temporal MRF and Camshift. Experimental result has been evaluated for different video sequences with different conditions such as noise; illumination changes, shadow, scale change in the objects etc. estimate the performance of these algorithms. Result obtained has better accuracy, good performances and with the consumption of less processing time according to the evaluation criteria.