Focusing on the issue that the traditional kernel correlation filter (KCF) algorithm has poor performance in handling scaling invariance and model drifting phenomena, an improved anti-occlusion and scale adaptive tracking method was proposed. The average peak-to correlation energy (APCE) and the peak value of correlation filtering response were used as confidence indexes to determine whether the target is blocked. In the case of non-occlusion, the target was scaled, and the corresponding response value was calculated after filtering. The response values of different scales were compared, and the maximum is the best scale value. The OTB-2013 was used as an evaluation standard and the obtained results were compared with those by the traditional KCF. The experiments demonstrate that the method improves the tracking success rate and the accuracy in occlusion cases, and is also robust to the change of the target scale.