An Online Real-Time Multiple Object Tracking System

• Main Authors: Danny Chen, 陳世傑
• Other Authors: Jer-Min Tsai
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/v582sx

碩士 === 崑山科技大學 === 資訊管理研究所 === 96 === The contents of most existing on-line fish navigation systems are composed of text, static pictures and pre-recorded video. This fact decreases the interest of the users. A live broadcast of the interested fishes is a better alternative since it offers the real scene. In this thesis, we describe a fish navigation system which uses the live broadcast video instead of the pre-recorded one. For advanced users and researchers, the behaviors of fishes are important information. Our system can track the fishes in the real-time video based on multi-object tracking techniques and record the loci of fishes. Since the ability of the computer at the client site and the network bandwidth are uncertain, we implement the multi-object tracking algorithm at the server site. Thus, our system composes of two parts: At the client side, a java applet is implemented for observing the scene and the relevant information. At the server side, a multi-tread java program is implemented to track the fishes and provide the video and loci to the clients. With this kind of architecture, when the network suffers congestion, the video quality in the client side would be poor however the loci are still correct since tracking is executed at the server side. In addition to the applications of navigation and biological researches, we hope our system can also benefit the object teaching.