作者: M Rothaar , Rehrmann
DOI:
关键词: Motion vector 、 Process (computing) 、 Segmentation 、 Pixel 、 Matching (statistics) 、 Noise (video) 、 Geography 、 Image processing 、 Artificial intelligence 、 Computer vision 、 Tracking (particle physics)
摘要: In this paper, the authors present a close to real-time system, based on concept of symbolic matching, for detecting and tracking moving objects in natural color scenes. The system uses complete segments matching rather than single pixels or features. paper describes segmentation process phase. Some advantages proposed approach has over traditional differential techniques feature point algorithms include: better performance presence noise, detection large motion between frames, dense vector fields, accurate estimation boundaries.