A human-centered object-oriented framework for semantic video analysis

Abstract

Semantic video analysis and retrieval is an important way to deal with massive video data. To solve the shortcomings of video analysis directly based on low-level features, this paper presents a video analysis framework based on human subjective affections, the framework according to the impact on object human subjective affection. First, object-oriented model is established, based on object-oriented thinking, we analysis the basic characteristics of the object, and apply it to the video analysis process. Second, we form a video object library, and sort the objects in the library according to their effect on human visual system, and analysis video using it. Lastly we design an interface for retrieving the video analysis. The frame address shortcomings the method directly based on the low-level features effectively, to avoid video noise and other minor objects on the video analysis results. To make the results more in line with the subjective human affections, so that Search more accurately, to better meet human needs.