A robust target tracking algorithm based on spatial regularization and adaptive updating model

The correlation filtering-based target tracking method has impressive tracking performance and computational efficiency. Nevertheless, a few issues limit the accuracy of the correlation filter-based tracking methods including the object deformation, boundary effects, scale variations, and the target occlusion. This article proposes a robust target tracking algorithm to solve these issues. First, a feature fusion method is used to enhance feature response discrimination between the target and others. Second, a spatial weight function is introduced to penalize the magnitude of filter coefficients and an ADMM algorithm is employed to reduce the iteration of filter coefficients when tracking. Third, an adaptive scale filter is designed to make the algorithm adaptable to the scale variations. Finally, the correlation peak average difference ratio is applied to realize the adaptive updating and improve the stability. The experiment’s result demonstrates the proposed algorithm improved tracking results compared to the state-of-the-art correlation filtering-based target tracking method.