Improved Mean Shift Tracking Algorithm by Multi Scale Motion Information
Sallama Adhab Resen · IOSR Journal of Engineering · 2013
Tracking object algorithms in moving platform environment are a challenging mission because they treat variety of speeds for moving object problems.Common method is Mean Shift Algorithm (MSA) due to its simple and efficient procedure.However, the lack of dynamic window size scale in its target model makes it unsuitable for tracking objects in real applications system.Dual static cameras placed in front and rear moving platform gathered information surrounding moving platform.Since the scale of the targets diverse irregularly there are needs to adjusted bandwidth scale when the appearances of tracking object changes.In this paper an automatic window-size updating method based mean shift with motion information obtained from Optical Flow (OF).Objects move with variety of speeds and directions surrounding moving platform .The proposed module speed and direction of moving object join together as factor updating scale according to how much moving object near or far from camera.Proposed framework initialized by MSA followed by Multi Scale Motion Measurement (MSMM).Experimental results show that scale update could select the proper window size of tracking region depending on speed of moving object.Evaluate performance of MSMM compare with traditional MSA by mismatch ratio and time complexity.