Object Tracking Based on Machine Vision and Improved Svdd Algorithm

Object tracking is an important research topic in the applications of machine vision, and has made great progress in the past decades, among which the technique based on classification is a very efficient way to solve the tracking problem. The classifier classifies the objects and background into two different classes, where the tracking drift caused by noisy background can be effectively handled by one-class SVM. But the time and space complexities of traditional one-class SVM methods tend to be high, which makes it do not scale well with the number of training sample, and limits its wide applications. Based on the idea proposed by Support Vector Data Description, we present an improved SVDD algorithm to handle object tracking efficiently. The experimental results on synthetic data, tracking results on car and plane demonstrate the validity of the proposed algorithm.