关键词: 运动目标跟踪, 线性支持向量机, 在线更新, 支持向量, Adaboost

Abstract: Considering the tracker as a binary classification problem，a novel moving object tracking algorithm is presented，which is based on on-line ensemble SVMs.First of all，linear SVMs is choosed as the classifiers to distinguish the target from the background.A simple yet effective way is used for on-line updating linear SVMs，where useful “Key Frames” of target are automatically selected as support vectors.Then，each linear SVM is separately given different weight through the Adaboost algorithm and is on-line ensembled to get a strong classifier.Experimental results show the robustness of the proposed algorithm，especially under large appearance change during tracking.

Key words: moving object tracking, linear Support Vector Machine, on-line updating, support vector, Adaboost