Research of Tracking Models Based on SURF

Abstract

Filter tracking models based on SURF (Speed-Up Robust Features) points are presents in this paper. They are used for robots to track the objects. The SURF points in the frames are matched by RANSAC (RANdom SAmple Consistent) with objective template in tracking process. It presents several filter tracking models, like SURF+PF (Particle Filter), SURF+KF (Kalman Filter), SURF+EKF (Extended Kalman Filter) and SURF+UKF (Unscented Kalman Filter). Some of the experiments are designed to verify the robustness and the real-timeliness of the models. Experiments are used to compare models with each other. Results of these experiments show application fields of the tracking models. These results may be useful for robots to track the objects that perform such kinds of movements.