Particle Filter Tracking

Ball Auto Tracking: A Particle Filter with Colour Segmentation-based Detectors

This page contains my past work of soccer ball tracking with detectors in the project "Object Highlighting for Mobile Video" at Thomson

Abstract:

An efficient method of ball localization in soccer game video integrating conventional detection and tracking is proposed. The ball detectors are based on color segmentation. We first use the likelihood ratio approach to detect playfield areas. Then, a strong ball detector is created based on shape analysis on foreground blobs in these areas and used to trigger a ball tracker. Besides, a weak ball detector is built with outputs from the playfield likelihood and the ball’s color, which is integrated into the observation likelihood in the tracker. In the ball tracker, motion estimation is embedded to generate a better proposal distribution for the particle filter and a mixture model is tailored to handle ambiguity due to the cluttered background. In addition, both occlusion and template drifting are coped with explicitly. By counting the duration of continuous frames in occlusion, the detector can be rebooted to lead the recovery from tracking failure. Promising results of automatic ball detection and tracking in soccer game videos are presented to illustrate that the proposed scheme handles heavy clutter, complete occlusion and motion blur effectively.

Figure 1: Playfield segmentation and foreground blob extraction.

Figure 2:Diagram of "Detect-then-Detect-and-Track(DtDaT)".

Figure 3:Tracker's recovery from failure due to motion blur.