To solve the problem of the deviation between the maximum position of the multi-scale correlation response map and the true position of the target, a polygon centroid based position correction method for correlation filter tracking is proposed. First, we proposed a correlation response map evaluation index to evaluate the quality of the response map, and make a credible judgment of the tracking result of the current frame. Then, we use the centroid of the polygon vertex to correct the tracking result that is judged to be unreliable to reduce the tracking result and target deviation from the correct position. Finally, the performance evaluation experiment of the algorithm in this paper is carried out on the three benchmark video sets of OTB50, OTB- 2015 and UAV20L. The Area Under Curve of this algorithm in OTB50, OTB- 2015 and UAV20L reached 0. 625, 0. 668 and 0. 429 respectively, and the tracking accuracy reached 0. 844, 0. 885 and 0. 578 respectively. The results show that compared with the mainstream tracking algorithms in recent years, our algorithm has achieved better success rate and tracking precision in a variety of complex scenarios.