Keypoint-based object tracking using modified median flow

This paper proposed robust object tracking using error filter. Proposed method consists of four steps: i) execution of tracking algorithm based on key point, ii) estimation of key point using error filter, iii) selection of key point as filtering outliers using Random Sample Consensus (RANSAC), and determination of object movement. The proposed method can track when object is occluded, abruptly change of appearance. As a result, the proposed methods can be applied to various application such as advanced driver assistance system, surveillance system.