To solve the problem of poor performance of the traditional Kernel Correlation Filter (KCF) under complex situations such as target scale variation, occlusion and deformation, a KCF tracking algorithm with adaptive model update based on Radon transform is proposed.By using the motion information, the optical flow method and the inter-frame difference method are adopted to predict the possible region of the target, which significantly reduces the search range and improves the speed of the algorithm.The Radon transform is insensitive to noise and invariant to moment translation and scale change.The optimal scale is determined by using the peak value of the matching degree of the moment features, which improves the accuracy of the algorithm while reducing the calculation amount.According to the nonlinear relationship between the learning rate and the peak value of the response graph in the model updating strategy, a parabolic learning rate curve is constructed to adaptively update the model.The tracking accuracy of the algorithm is guaranteed even when the target is temporarily lost or is false.The experimental results show that the proposed algorithm has good real-time performance, high success rate and high tracking accuracy.