The Study on Multiple Moving Targets Detection Method for Image Sequences Captured by a Moving Monocular Camera

• Main Authors: Zong-Che Wu, 吳宗哲
• Other Authors: Thou-Ho Chen
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/43071525220532910164

碩士 === 國立高雄應用科技大學 === 電子工程系碩士班 === 101 === This paper proposed a multiple moving targets detection method for video captured by a moving monocular camera. Such a method detect independently moving targets (like passengers and mobile vehicle, etc.) by image sequence in real-time from a monocular camera and bound it. In point of comparison to that of the fixed camera, it is relatively difficult due to the fact that both foreground and background are simultaneously moving but in different ways. Therefore, many conventional segmentation methods can’t effectively detect and extract the moving objects in the image sequences captured by a moving monocular camera. The method proposed in this study can be divided into three parts. It is first to search feature points in an image and then such feature points are classified into foreground and background based on view geometry. The feature points of foreground are employed to construct the multiple moving targets and the perspective transform matrix is used to estimate the motion of background. Based on this background motion, the contour of multiple moving targets is generated by difference between the concurrent image and an image established by perspective transform matrix. Finally, extracting and tracking the multiple moving targets' orbits by their temporal contours.