An algorithm for tracking multiple trajectories in a noisy picture

✍ Scribed by Shin'ya Takahashi; Hiroaki Sakoe

• Publisher: John Wiley and Sons
• Year: 2000
• Tongue: English
• Weight: 234 KB
• Volume: 31
• Category: Article
• ISSN: 0882-1666
• DOI: 10.1002/(sici)1520-684x(200003)31:3<84::aid-scj10>3.0.co;2-u

✦ Synopsis

This paper describes an algorithm for tracking multiple trajectories observed in a noise-spoiled multilevel picture. Using a priori knowledge on the multiple trajectories to be detected, a model is formed which describes the properties of each trajectory and relationships between the trajectories. The behaviors of the trajectories and relationships between the trajectories are described by a finite state automaton. Parsing the trajectories according to the prepared model, the target trajectories in noise are detected. The trajectories in the input picture are searched for using dynamic programming technique. We prepared two noisy pictures for experimentation: a picture including two sinusoidal curves and a picture including two rectangular waves. Evaluation experiments have shown that the target trajectories can be detected with high accuracy even in the presence of high-level noise.