Research on Real-Time Object Tracking by Improved Camshift

A new scheme is put forward to realize robust and real-time object tracking by CamShift combining color information and improved LBP. Continuous adaptive mean shift (CamShift) algorithm is a good choice for object tracking with its high speed and insensitiveness to the rotation and size of the target, but it is influenced by environmental lights and color information. Local binary patterns (LBP) is a satisfactory texture descriptor invariant to lights but sensitive to rotation, so improved LBP was put forward by introducing principle of permutation group to overcome the influence rotation of the target to LBP. Then probability density distribution functions were constructed based on the improved LBP and color histogram separately. At last, CamShift is used to realize real-time and robust object tracking. Experimental results show that the scheme can acquire good tracking performance under complex background.