Object Tracking with Self-updating Tracking Window

A Novel Approach Coloured Object Tracker with Adaptive Model and Bandwidth Using Mean Shift Algorithm

“…We use size adaptation method proposed in [58], in this method with using a new multi-scale information measurement that named MSIIM, automatically bandwidth updated. Qian et al [58] proved that the distance between the observer and the object was reflected by the MSIIM and this feature has good performance to adjustment of bandwidth. The updating process is described as follows [58]:…”

“…Qian et al [58] proved that the distance between the observer and the object was reflected by the MSIIM and this feature has good performance to adjustment of bandwidth. The updating process is described as follows [58]:…”

“…In proposed method in [58], periodically, MSIIMs of target area compared with previous computed MSIIMs, and according to the change of the MSIIMs, size of bandwidth updated.…”

“…Although, the proposed method in [58] has good results for scale updating but due to using the classical colour histogram based mean shift tracking framework when clutters and occlusions in the appearance model accrued, it has a poor tracking results.…”

“…In scenarios that the size of target changes slowly and the background has same colour, if α has been variable, due to use procedure introduced in [58] and because of similarity between target and background, we have a miscalculation in the tracked object's MSIIM feature that affected by incorrect window scale.…”