Research On Object Tracking Algorithm Of Discriminant Correlation Filtering Based On Tensor Learning

Object Tracking is a foundation in the field of vision,with powerful application capabilities.Discriminant-related tracking uses the technology of diagonalization of the circulant matrix to further improve the speed and accuracy of tracking.Kenernal based tracking shows the powerful discrimination of nonlinear kernels,then further breakthrough based on multi-kernal tracking.The deep learning improves the performence of image classification,and the trackers based on the transmission network also obtain a better performance.However,the current multi-kernal correlation filters ignore the interaction between the feature channel and the multi-kernal,and only use summation and superimposition of the multi-kernal.This method would destroy the structural information of the high dimension feature map and reduce the performance of the tracker;in the deep convolutional neural network based tracking,there is no clear explanation for the uncertainty of the target labeling.,Relying heavily on manual labeling and the selection of loss function.In response to the above two types of problems,we respectively proposed High-Order Multiple Kernelized Correlation Filter in Tensor for Visual Tracking(HOMKCF)and alpha divergence-based Siamese network Tracking(alphaTK).In HOMKCF,we represented multiple kernals in tensor form to protect the structure of high-order feature maps,and simplify the tensor regression model based on the framework of tensor convolution sparse coding.In the test of benchmark data sets such as OTB and UAV,our tracker surpasses many excellent tracking algorithms in terms of accuracy and success rate.In AlphaTK,we minimized the alpha divergence between the conditional probability density output by the network and the conditional probability of annotations of the samples,void to choose the loss function.We modeled the noise generated by the label from the probabilistic perspective,and interpreting the uncertainty of the annothions.The Tracker is trained on a large number of data sets,and achieved excellent results on the OTB and UAV benchmark data sets.