Moving Target Detection and Labeling in Video Sequence Based on Spatial-Temporal Information Fusion

摘要

A novel method of moving target detection based on spatial-temporal information fusion was proposed. It combines the temporal properties within multiple frames and the spatial information in single frame of a moving target in video image sequence by using time-domain recursive filtering for background recovering and spatial image segmentation for separation of target from background. Together with follow-up morphological filtering and blob analysis, the area of an interested moving target can be detected and labeled by bounding box. Its application to swimmer tracking in underwater swimming video image sequence manifested the effectiveness of the method, which is expected to be applied to the implementation of underwater automatic tracking swimming video system.