Penjejakan Wajah Dengan Menggunakan Improved Mean Shift Tracking

Abstract

Visual face tracking is methods tracking that face as a target. In the previous research, an object tracking system was carried out using the haar cascade method, but in the object tracking system it was not resistant to occlusion and background noise interference. So in this paper, face tracking system that is resistant to occlusion and background noise is developed. In this paper, the faces traced will be detected using the haarcascade viola jones method, then an Improved Mean Shift Algorithm will be used, which is by combining the Mean Shift method and Corrected Background Weigthed Histogram. Mean Shift is used to detect coordinates on targets that are resistant to occlusion interference. The Corrected Background Weighted Histogram method is added to the Mean Shift method to eliminate noise background that have a same fitur with object target. The results of this methods, the target can be resistant to occlusion and background noise. Improved Mean Shift tracking has average error 3.87 pixel, with standart deviation 1.54 pixel