News anchor lens detection method based on space-time strip pattern analysis

Abstract

The invention discloses a news anchor detection method based on spatio-temporal strip pattern analysis, comprising: intercepting consecutive N frames from an edited news video as a group, and extracting horizontal spatio-temporal strips and vertical spatio-temporal strips; extracting horizontal The corresponding image features in the space-time strips and vertical space-time strips are used to obtain the corresponding feature vectors; the feature vectors are clustered separately by the clustering method, and the time-continuous horizontal or vertical space-time strips in the same class are merged separately, as new elements in the class to obtain the final horizontal clustering result and vertical clustering result; the class containing the most elements in the horizontal clustering result and the class containing the most elements in the vertical clustering result are fused, and detected according to the fusion result News anchor shot. The invention has the advantages of high detection accuracy for various news video anchors, strong versatility, and low computational complexity. It avoids the shortcomings of existing methods that rely too much on accurate shot segmentation and other modality information.

Description

technical field [0001] The invention relates to the field of video analysis and detection, in particular to a news anchor detection method based on spatio-temporal strip pattern analysis. Background technique [0002] With the rapid development of the Internet, people can access a large amount of visual data such as images and videos. According to people's growing demand for visual data, how to effectively analyze, manage, query and retrieve these massive information has become an urgent problem to be solved. [0003] For content-based video analysis, researchers regard news videos as an important research object due to their special structural features. A news video can be seen as a combination of anchor shots, news story shots, and possibly advertising shots. Among them, the anchor lens and the subsequent continuous news story units constitute a complete news event. Because the content of the anchor shot is often an introduction to the next story unit, the detection of ...

AI Technical Summary

Problems solved by technology

[0008] The purpose of the present invention is to overcome the defects of poor robustness, low accuracy, high computational complexity, and limited versatility of the existing news anchor detection method, thereby providing a method with wide versatility, high accuracy, and low computational complexity. News anchor detection method for

Images