| Summary: | 碩士 === 國立交通大學 === 資訊科學系 === 90 === In the past years, video acquisition hardware has been considerably improved. The streams obtained from video acquisition hardware have good quality, so the area of developing video-surveillance systems has attracted researcher’s attention. At the same time, the video-surveillance systems have generally been commercially applied. In this paper, we will propose two surveillance systems. One is an elevator detection system, which can detect subjects’ coming in and going out. The other is a human tracking system, which can track pedestrians on the street. The surveillance systems are implemented utilizing two steps: segmentation and tracking. In the segmentation step, a segmentation algorithm proposed by Horprasert et al. [9] is modified to segment persons in mages. Color and brightness information are used to determine a pixel belonging to background or foreground. In the tracking step, based on the segmented result, we can track the moving persons. The spatial information is adopted to solve the intersection condition when two persons meet during the tracking time.
|
|---|
