Research On Target Tracking Algorithm Based On Spatial Regularization

As one core technology of computer vision,target tracking owns significant research value in many domains such as national defense and security,intelligent robotics,and gesture recognition.A large number of scientific breakthroughs have been achieved in recent years as a result of in-depth studies conducted by researchers on correlation filter tracking algorithms.However,various challenge factors in the tracking process,such as background clutter,targets out of view,and occlusion,make it difficult to further improve the performance of the correlation filter tracking algorithm.Therefore,two improved algorithms with excellent performance are proposed to enhance the tracking accuracy and robustness.The detailed work is as follows:(1)An excellent algorithm called correlation filter tracking algorithm based on deep feature target-aware and multi-index update is proposed to avoid the defect of only utilizing hand-crafted features to characterize the target,the higher complexity of training the filter,and updating the model frame by frame even when the tracking results are unreliable in spatial regularization correlation filtering tracking algorithm.After fusing the deep features of pre-training,the channel selection is performed according to the gradient information of regression loss,which enhances the characterization ability for the object.For the purpose of degrading the computational complexity and accelerate the tracking speed,the proposed algorithm utilizes the alternating direction method of multipliers to train the spatial regularization correlation filter.Furthermore,whether the model is updated or not depends on the multi-index update method,which not only improves the tracking efficiency,but also averts the model corruption that is caused by learning wrong information.The experimental results show that the success rate and the precision of the presented algorithm outperforms the compared algorithms on the OTB2015 dataset.Under the complex scenarios,the proposed algorithm has stronger tracking robustness compared with the state of the art algorithms.(2)For the correlation filtering algorithm,the traditional spatial regularization term is not adjusted with the change of the target,and the tracking model is prone to degradation by reason of occlusion and object deformation.To address the drawbacks,an excellent algorithm called adaptive correlation filter tracking algorithm based on spatial-temporal regularization and multi-feature fusion is proposed.The spatial weight matrix is adaptively adjusted according to the appearance change of the target in the tracking process,which improves the robustness of the algorithm.An adaptive temporal regularization term is introduced for the purpose of alleviating the problem of tracking model degradation.Moreover,HOG features and depth features are adaptively fused to further improve the tracking performance by the complementary characteristics of different features.The experimental results on the OTB2015 dataset show that the proposed algorithm achieves the highest tracking accuracy and the most robust tracking effect,compared with the state of the art algorithms.