Stuady On The Method Of Video Object Tracking Based On Local Image Features

It is a wide range of application in modern intelligent transportation systems,intelligent monitoring systems, medical navigation of surgical instruments positioningand military target detection of video target tracking, but video target tracking in theactual environment is often complex and changing. These difficulties can be overcomedby target tracking based on local features of a certain extent, to achieve target detectionand tracking more accuratly. So it is of very important significance and practical valuewith analysis and in-depth study of video target tracking based on local features. Thisarticle focused on the research of target detection and tracking method based on localfeatures.The main content is as follows:(1)Undertake a thorough discussion of the development process and the keytechnology of the video target tracking, Cited the value of video tracking technology inthe field of military and civilian. Discuss the existing target tracking algorithm onclassification.(2)Some detailed analysis of the details of the SIFT(Scale invariant featuretransform) and SURF(Speed Up Robust Features) algorithm were done, and comparethe difference between them according to several experimental simulation. The resultswere analyzed according to the experimental data, which demostrate the superiorperformance of richness, uniquly and stably of the local features algorithms in imageprocessing.(3)Combine the SURF(Speed Up Robust Features) algorithm and BBF(Best BinFirst) algorithm to improve the speed of the feature matching stage, propose a trackingwindow update strategy and a template update strategy,the formers allows the trackingwindow changes as target size varies adaptively, template update strategy is even moreoptimized with the partial occlusion or target completely disappeared during tracking. Inthis paper we combine the algorithm mentioned above, implement the algorithms withOpenCV and VS2005as a development platform, accomplish the experimental targettracking system based on SURF feature.After the test of several sets of experimental data, the results prove that it can running effectively in the situation of scale transformation, affine transformation,illumination changes and partial occlusion.and, of processed in real-time.All mentionedabove indicate that it had achieved the expected goals.