An Automatic Visual Detecting Method for Semantic Object in Video

In this paper, we propose an approach to automatic detection of semantic object. The method provides an effective content expression pattern for semantic analysis and retrieval of video. In the moving semantic object detection model, motion contrast is computed based on the planar motion (homography) between frames, which is estimated by applying RANSAC algorithm on point correspondences in the scene. In the semantic object detection model of static frame, the three features used are intensity, color and texture. Then a dynamic fusion technique is applied to combine these models. The automatic detection method can greatly decrease computation and be used in pervasive computing environment conveniently. Experimental results verify efficiency of proposed approach.