A Study On The Correlation Filter Based Object Tracking Algorithm In A Complex Background

Compared with other methods,the tracking algorithm based on correlation filter model has obvious speed advantage.However,in some complex scenarios,such as fast motion,occlusion,deformation,etc.,the accuracy of this algorithm will decrease significantly.To this end,this article has carried out the following two aspects of research.(1)Proposing a tracking algorithm based on feature fusion and occlusion detection.Under the relevant filtering model,the article adopt the texture features HOG(Histogran of Oriented Gradient)and color features CN(Colour Name)fusion complementary advantages,build characteristic response mechanism of fusion,using every dimension characteristics of the channel response to calculate weight,and then automatically adjust to fusion characteristics.According to the response function of the fusion feature,build shelter judgment mechanism,according to the target response value of the parameter's value and the APCE values change condition,finally complete filter model update,design the effective selection strategy,determine the final location of the target.The proposed method enhances the characteristic description of the sample and increases the tracking accuracy of the occluded data set.(2)Proposing a tracking algorithm based on adaptive model updating.adopting the way of integration calculation to implement characteristics sharing calculation and constructing of gaussian filter model by calculating the sample template feature and histogram feature.Build sample parameter model,according to the sample parameters change,build the model update mechanism.In order to solve the problem of model drift caused by fast motion,this article abandons the traditional model updating strategy,and obtains the updating parameters of the model to update the model parameters in real time.The proposed method speeds up the calculation process of fusion characteristics and improves the robustness of tracking performance.The two aspects of research that this article has proposed effectively enhance the tracking algorithm in the shade and rapid movement tracking performance under complex scene.The experimental results verify the stability and robustness of the proposed algorithm for a variety of scenarios of video sequences.