Research On Kernel Correlation Filtering Tracking Algorithm Combining Multi Feature And Scale Estimation

Object tracking technology(Object Tracking Technology)refers to a series of operations such as detecting a video image sequence,extracting a feature model,and designing a filtering template for response recognition,so as to obtain information such as the position of the target.With the development of computer intelligence,target tracking technology is gradually applied to daily life.It has broad application prospects and important research value in military security,intelligent transportation,video surveillance and other fields.However,traditional target tracking algorithms cannot be effectively filtered for occluded scenes due to the single selection of features,and there are inadequacy in scale estimation.When dealing with complex scenes,the tracking results are not good,and the accuracy and success rate are low.Therefore,in view of the above problems,the research content of this article is as follows:(1)A kernel correlation filtering target tracking algorithm based on multi-feature fusion is proposed.This method first establishes a kernel ridge regression model,uses a two-dimensional kernelization related position filter,fuses the directional gradient histogram(HOG)feature and the color histogram(CN)feature,and weights the fused tracking coordinates according to the response size,accurately.Determine the center position of the target;then,use the level of the peak sidelobe ratio of the filter response to determine whether occlusion occurs.When the sidelobe ratio of the characteristic response is lower than the set threshold,the filter template is temporarily updated to make the tracking algorithm more robust.The greatness improves the tracking accuracy and success rate in complex scenarios.(2)A kernel correlation filtering target tracking algorithm based on scale estimation is proposed.This method uses the response value of the feature point to set the corresponding weight,combines the displacement difference of the optical flow method to perform the scale estimation,and uses the average dichotomy to set the scale pool to adapt to the target size.The displacement of key corner points between video frames is calculated using the optical flow method to estimate the deformation ratio and size of the tracked target,and the tracking frame is scaled by combining the scale set.Improve the accuracy and avoid introducing too much clutter to increase the error.The algorithm is more robust to scale-changing scenarios,improves the accuracy and success rate of target tracking,and achieves better tracking results.