Robust Multi-object Tracker Based on Feature Point Particle Filter Algorithm
Ming-Shou An, Jong-Dae Park, Hong Jong-Sun, Dae-Seong Kang · Proceedings of KIIT Summer Conference · 2011
This paper presents a robust tracker within a particle filter framework. The framework relies on feature point and works well for multiple people tracking in video surveillance applications. The particle filter is a solution to predict the particle's state and generate the proposal distribution, increasing the robustness of tracking abrupt object movement. A first step segments foreground and background using background mixture models, and gets the region of interest(ROI). A second step tracks the ROI by particle filter based on the feature points. The feature points extracted by scale invariant feature transform(SIFT) algorithm. The SIFT approach, for image feature generation, transforms the ROI into a large collection of local feature vectors. Each of these feature vectors is invariant to any scaling, rotation or translation of the image. Results on surveillance videos are reported, using in-door and out-door produced videos.