Semi-supervised neighbor propagation learning and multi-visual dictionary model-based intelligent video analysis method

Abstract

The invention relates to a semi-supervised neighbor propagation learning and multi-visual dictionary model-based intelligent video analysis method. The method includes the following steps that: key video frames are extracted based on a sample-and-hold strategy; the key frames are converted into a grayscale image, and the grayscale image is averagely segmented into a plurality of image blocks, and OM feature vectors are generated based on the sequence of the average brightness of the image blocks; a video sample is divided based on AP clustering, so that mutually exclusive video sub clusters can be formed; the category labels corresponding to each sub cluster are determined according to the video sub clusters obtained through the division of the AP clustering, and a multi-visual dictionary is constructed; the category label of a video to be detected is judged online according to the minimum distance judgment criterion; and closed-loop feedback adaptive reconfiguration learning is adopted to reconstruct a multi-visual dictionary adaptive to a new environment, so that the category label of the video to be detected can be re-judged. With the method of the invention adopted, the problem that efficiency and precision in key frame extraction cannot be taken into account simultaneously can be solved, and enhanced video features have very high robustness for approximate global linear change; and accurate and intelligent detection adaptive to video mode change can be realized.

Description

technical field [0001] The invention relates to the technical field of computer graphics and image processing, in particular to an intelligent video analysis method based on semi-supervised neighbor propagation learning and a multi-visual dictionary model. Background technique [0002] With the popularization of smart phones, digital cameras and other terminal devices and the Internet, the production, storage and transmission of videos are more convenient. More and more people watch videos through the Internet, and make videos of their own life experiences and upload them to the Internet for sharing. Video services are no longer limited to traditional radio and television, entertainment and film and other life service industries, but are widely used in many fields such as politics, military, science and education, medical care, transportation, and security. Facing the "explosive" growth of the number of videos, how to quickly and accurately retrieve, match, and classify vide...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the deficiencies in the prior art, the present invention provides an intelligent video analysis method based on semi-supervised neighbor propagation learning and a multi-visual dictionary model, which solves the problem that the key video frames in the prior art cannot take into account both efficiency and accuracy, and enhances video features to Robustness to approximate global linear changes, enabling accurate intelligent detection that can adapt to video mode changes

Images