Research On Moving Target Tracking Method Based On Visual Saliency

Moving target tracking is in continuous change of the state and find out the location of the object in image sequence process,and the stability of target tracking and robustness is to deal with the main problems.Due to the complex diversity of realistic environment,the general algorithm of moving target size and shape change adaptability is poorer,unable to effectively solve the complex situation and mutation of movement target tracking problem,when the target occluded,target moving too fast and lost mutation motion cannot be automatically restored to cause the failure of tracking,it is difficult to achieve the target tracking accuracy and stability.To solve target tracking problem under the mutation movement,this thesis proposes a moving target tracking algorithm based on visual significance,the algorithm to apply visual attention mechanism movement target tracking framework,using the space-time significant algorithm to test the video sequence,generate visual figure,from the visual figure corresponding significant areas to establish the characteristics of the said model to realize the moving target tracking.In this thesis made the following work:(1)Moving target tracking algorithm based on the theoretical framework and visual saliency technology were introduced,and the current commonly used moving target tracking algorithm with mean shift method,particle filtering,Calman filtering method and visual saliency detection algorithm has Itti,CA,SR and LC are analyzed and experimental.(2)Through the analysis of the current mainstream space-time saliency detection algorithm,PQFT and SEG algorithm,and the introduction of the moving object tracking algorithm framework,and then the algorithm design to track the moving target.(3)Uses the international public video sequence for moving target tracking screening test,aimed at moving targets such as missing,keep out mutation movement condition and can be accurate and stable tracking target under complex environment,and compared with the current mainstream target tracking algorithm for experiments and quantitative analysis.In this thesis,experimental results show that the method in dynamic scene such as camera shake down can accurately detect the target in space and time were significantly,effectively overcome the loss in moving target and occlusion and other complex and mutation cases unstable problem,has strong robustness,so as to realize the tracking of the target accurately under complex scene.