Tracking The Target Sparse Expression And Machine Learning

Object tracking technology is one of the core research topic in computer vision.The reason is that the military object tracking technology in people’s daily life andhigh-end has an important role, and has wide application prospect. However, due to theillumination change in the scene, dynamic background noise, object deformation andthe occlusion and other factors. Let the object tracking technology in practicalapplication is difficult to ensure the accuracy and stability.In recent years, the object tracking algorithm based on sparse representation is ahot spot, and achieved good results. But there are several problems, such as: grayfeature for illumination, pose variation and occlusion is not robust;1normminimization for each particle is required. In order to solve these two problems thispaper constructed by Gabor feature dictionary, solve the light sensitivity of change andscale change on gray feature, using the candidate particle dictionary screening importantparticles, solve the problem of high computational complexity. In the sparserepresentation of object tracking algorithm based on single feature, are generally used,Single feature is not robust. This paper proposed tracking algorithm based on multifeature sparse expressed fusion machine learning based object. In order to overcome theglobal template in the treatment of local changes caused by occlusion of clumsiness andobject template and the candidate template matching problem of the great error, thispaper proposes the local representation model and robust, which is robust in dealingwith occlusion and deformation.Three algorithms in this paper are based on publicly available video library to doexperiments, some obvious changes in illumination, scale changes, the objectdeformation and severe occlusion video sequence, three experiments achieved goodresults in the sequence, they have some Robustness In the above changes.