Detection of Swimmers in Swimming Video Sequences

摘要

We can record swimmers' real-time training actions when underwater trolleys carrying cameras track swimmers. Swimmers' (moving object) location in images can be detected automatically from the acquired image sequences with image processing, video analysis and computer processing techniques. We can predict the target's location in next frame on previous locations then we can track and shoot swimmers automatically using trolleys carrying cameras. We present the method of moving object detection from video sequences based on temporal-spatial information fusion after we researched on common moving object detection methods. This method combines single-frame's spatial domain features with multi-frame's time domain features and acquires moving target's location in video sequences through spatial detection. We can realize effective detection of moving objects with further image processing techniques such as image filtering and connective components marking. The simulation result of swimmers' detection in swimming video sequences and the comparison with that of other detection methods shows the effectiveness of this method.