Object Segmentation and Tracking for Surveillance Video

• Main Authors: Da-Jing Zhang, 張簡大敬
• Other Authors: Wei-Chang Du
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/75194413022004450519

碩士 === 義守大學 === 資訊工程學系碩士班 === 93 === Due to great progress in digital electronic and optical technologies, the acquirement and use of digital video is more and more popular especially in recent few years. From the viewpoint of human vision, digital video just takes down actual images. Because it lacks semantic information, it is necessary to understand the digital content with human brain. In this thesis, the goal is to develop an effective method of segmenting and detecting the moving objects by a fixed digital camera, and then track the moving objects to obtain high-level semantics further. This study firstly uses the spatial and temporal properties of digital video to detect all possible moving objects. In order to reduce noise, an automatic background updating method and some noise-removal filters are proposed. Based on the segmentation of separate objects from images, tracking multiple objects can be finished by matching the objects between successive frames so as to obtain individual trajectories for all moving objects. Finally, some empirical tests are given to demonstrate the effectiveness.