Research And Implementation Of Visual Tracking Algorithm Based On Sparse Representation

Visual tracking is an important subject in the field of computer vision.It is widely used in intelligent monitoring,visual navigation and human-computer interaction.In view of the problem of visual tracking,many researchers at home and abroad have done a lot of researches,and present many effective tracking algorithms.However,there are still many difficulties in achieving robust tracking of apparent changes in complex and natural scenes,including changes in scene lighting,target rotation,local occlusion,morphological changes in the target,and changes in scale.In recent years,sparse representation theory has been successfully applied to the visual tracking algorithm,and has achieved good tracking effect,has become the concern of the majority of scholars.In this thesis,we focuses on the difficulty of the above-mentioned visual tracking and deeply studies the application of the sparse representation in the visual tracking algorithm,aiming at further improving the robustness and accuracy of visual tracking algorithm.Firstly,the thesis introduces the background and significance of the visual tracking and analyzes the current research situation and the main challenges.Secondly,the relevant theoretical methods and the application prospects of sparse representation in the field of visual tracking are expounded.Thirdly,a sparse-based visual tracking algorithm based on importance weighting is proposed.By introducing the affine transformation,the tracking result can better describe the state of the target.The structure sparse representation is used to model the target,the local image information and the structural information are used to reduce the degradation of the model.According to the role of the expression object,the localized image is weighted and the robustness of the target model is improved.In the stage of template updating,the occlusion detection mechanism effectively reduces the interference of local occlusion to the tracking result.Finally,the proposed algorithm is implemented on Matlab,and the tracking performance is compared with several state-of-the-art methods on classical and challenging Benchmark image sequences.The results of qualitative and quantitative experiments show that our algorithm not only can track the target accurately,but also has better performance than other advanced tracking algorithms in the complex occlusion,rotation,scale change,fast motion and illumination change.This thesis focuses on the difficult problem in visual tracking,and researches the sparse representation visual tracking algorithm to improve the tracking performance of the visual tracking algorithm in complex scenes.