Horrible video scene recognition method based on multi-view and multi-instance learning

Abstract

The invention discloses a horrible video scene recognition method based on multi-view and multi-instance learning. The horrible video scene recognition method comprises the steps of: extracting video shots from videos in a training video set, and selecting an emotion representative frame and an emotion abrupt frame for each video shot; extracting audio and vision characteristics of each video shot in the training video set, wherein the vision characteristic is extracted on the basis of the extracted emotion representative frame and the extracted emotion abrupt frame; extracting four view characteristic vectors for each video to form a multi-view characteristic set of the training video set; carrying out sparse reconstruction on the obtained multi-view characteristic set corresponding to the training video set and the multi-view characteristic vectors of the videos to be recognized to obtain a sparse reconstruction coefficient; calculating reconstruction errors of the multi-view characteristic vectors of the videos to be recognized and multi-view characteristic sets respectively corresponding to a horrible video set and a non-horrible video set in the training video set according to the sparse reconstruction coefficient, and further determining whether the videos to be recognized are horrible videos.

Description

technical field [0001] The invention relates to the field of pattern recognition and computer network content security, in particular to a horror video scene recognition method based on multi-view and multi-example learning. Background technique [0002] With the rapid development of Internet technology and applications, people's understanding and use of the Internet has become more and more in-depth. At the same time, the Internet has brought a lot of convenience to people's lives, and even changed people's way of life. On the basis of the rapid development of the Internet, the spread of harmful information such as pornography, violence, and terror has become easier and easier. Psychological and physiological studies have shown that terrorist information on the Internet is no less harmful to the physical and mental health of young people than pornographic information. Excessive terrorist information may cause people to be in extreme anxiety and fear for a long time, and ev...

AI Technical Summary

Problems solved by technology

Since not all frames of horror videos contain horror information, extracting the average features of the entire horror video will inevitably dilute the feature information of horror video frames, resulting in a decline in recognition rate

Images