A New Method for Target Detection Based on Analysis of MotionVector Field in Dynamic Video Scene

摘要

In this paper we propose a new method for target detection based on an analysis of the motion vector field in a dynamic video scene. Our method, which integrates motion vector field analysis and image segmentation, can be used to treat with videos acquired by mobile cameras. It can even be used to detect multiple targets in a scene simultaneously with no advance learning. Using some video materials from a YUV video sequence from Arizona State University, we successfully determined the boundaries of the moving targets by the proposed method. The algorithms detected the targets and enabled us to obtain very satisfactory results.