Reporting Dynamic Information from a Sequence of Image Frames

摘要

In this paper, an approac to reporting dynamic information from a sequence of image frames is presented. The sequence of image frames contains the stationary information (e.g. bakcground) and the dynamic information (e.g., moving ojbects). Object extraction and trajectory extraction are two main steps in our approach. For object extraction, the stationary information is first kept in a data-preserving image; the background is then obtained by the median of the tationary information in the sequence of data-preserving images; and the moving objnects are finally addressed by comparing each frame in the sequence of images with the obtained background. For trajectory extraction, based on the sequence of moving objects addressed frames, the trajectory of each moving object is found by a matching process (in early frames) and a prediction process (in other frames). Thus multiple moving objects tracked by the proposed sppraoch is possible. Experimental resutls have shown its potential capability.