Research on variable scale target tracking method based on multi feature fusion

Illumination, rotation, deformation and other problems have always been the main problems affecting the target tracking effect, especially when the scale changes in the process of target movement, the whole target frame will drift. In order to solve the problem of real-time accurate tracking when the size of the target fluctuates greatly, many problems exist at the same time. Based on KCF algorithm, LBP features are fused adaptively by using feature weighting method, and variable scale updating strategy is added. A variable scale tracking method under multi feature fusion is proposed. In this paper, video sequences from obt-50 and obt-100 datasets are selected for simulation experiments, and the proposed algorithm is compared with KCF algorithm. The position error is improved by 1%–2%, and the overlap accuracy is improved by 2%–3%, which verifies the robustness of the algorithm. When complex background, illumination, rotation, deformation and scale change appear in the process of target motion, the algorithm can solve the above problems.