Semantic events detection method and system in video

Abstract

The present invention discloses a method for detecting semantic event in video and a system thereof. The method of the invention comprises the following steps: normalizing movement vector and filtering noise movement vector; estimating camera movement; dividing movement object; and detecting semantic event. The system of the invention comprises the following components: a movement vector preprocessing module which is used for normalizing the movement vector and filtering the noise movement vector with an experience regulation; a camera movement estimating module which is used for determining the movement mode and movement parameter of camera; a movement object dividing module which is used for dividing movement object with the movement vector after movement compensation of camera; and a semantic event detecting module which is used for detecting the semantic event in the video lens. The method and system for detecting semantic event in video can detect the semantic event conception in the video more accurately and high-efficiently.

Description

technical field [0001] The invention relates to the technical field of video content analysis, in particular to a method and system for detecting semantic events in videos. Background technique [0002] With the development of digital video technology, video content analysis technology becomes more and more important. Semantic concepts in videos include various types such as objects, scenes, and events. Detection requires the use of information from different modalities, such as keyframe images, audio, etc., which can be used to detect semantic concepts contained in videos. The most basic work of using motion information is to extract motion features in video, including camera motion estimation and segmentation of moving objects in video. This works well for detecting semantic concepts of events in videos, such as walking, violence, and marches. Most existing works on semantic event detection are based on image feature analysis of video keyframes. This means that extract...

AI Technical Summary

Problems solved by technology

However, there are still many problems in these methods: firstly, for inter-coded macroblocks, the DCT coefficients are not calculated based on real pixel values, but are obtained based on the difference between the current macroblock and its reference frame macroblock; Secondly, there are many intra-coded macroblocks in the video code stream, these macroblocks cannot provide motion information, especially the I frame in the code stream, the macroblocks in the entire frame are intra-frame coded; the last and most important Yes, many macroblocks contain a lot of noise. The motion vector in the MPEG video stream is calculated according to the fast macroblock matching algorithm in the encoding process. The calculation error is relatively large, and the calculation result may not truly reflect the macro Motion information for blocks, especially for areas where texture is not very obvious

[0004] Most motion analysis methods use a 6-parameter affine model or an 8-parameter projective model to estimate the camera motion, and these methods are computationally complex

Images