Video copy detection method based on deep learning and graph theory

Abstract

The invention discloses a video copy detection method based on the deep learning and graph theory, and aims to improve the speed and accuracy of image frame similarity matching. The technical scheme is to extract an image frame in a reference library video, extracting features of the image frame, and obtaining a feature matrix of the image frame in the reference library video; extracting an image frame in the video to be detected, extracting features of the image frame, and obtaining a feature matrix of the image frame in the video to be detected; performing similarity matching between the feature matrix of the image frame in the video to be detected and the feature matrix of the image frame in the reference library video by using an approximate nearest neighbor search algorithm, and obtaining a similarity matching list; and judging and locating video copy fragments by means of the graph theory according to the similarity matching list. The method of the invention can be used to cope with various video editing transformations and an arbitrary matching form between the video to be detected and the reference library video, solve the technical problem that the "noise" has a great influence on the matching effect and improve the speed and accuracy of the similarity matching of the image frames.

Description

technical field [0001] The invention relates to a method for copy video detection in the technical field of multimedia information processing, and its essence is an image frame similarity matching method and a video subsequence similarity matching method, and is a video detection method that can adapt to various editing transformations. Background technique [0002] With the development of information technology, video capture equipment and video editing software, video data shows a trend of massive growth. At the same time, the development of video-related services such as video sharing and video recommendation enables users to participate in video-related activities more conveniently. A large number of videos are uploaded and downloaded on the Internet every day. Taking the YouTube video website as an example, more than 100 hours of videos are downloaded every minute. On some social networking sites, a large number of videos are also downloaded and shared every day. This ...

AI Technical Summary

Problems solved by technology

The video subsequence similarity matching method based on frame fusion can cope with video frame rate changes, frame insertion, frame loss and other editing transformations, and can also solve the situation that there are only some copied segments in the video, but the video subsequence similarity based on frame fusion The matching method is computationally complex, and the algorithm implementation is also difficult

[0006] Therefore, this series of research faces some typical limitations in copy video detection: First, most of the current research uses local features with high robustness for similarity matching of image frames, which results in the similarity matching speed of image frames being still low. Slow, especially for large video databases, the speed limit is more obvious

There are two main reasons affecting the speed: (1) the number of image frames in the video library is huge, and it is a time-consuming task to perform image frame similarity matching in the entire library; (2) the number of local key points in the image frame is huge, The key point description vector has a high dimensionality, and it takes a lot of time to perform one-to-one matching of local key points between image frames

Second, most of the current copy video detection research focuses on the case where the video to be detected is equal to the length of the reference video and the video to be detected is a subset of the reference video.

Although these methods have achieved relatively good results in solving their respective application problems, when the video to be detected is also spliced ​​with copied segments and non-copyed segments and the copied segments are only a subset of the reference video, the current research method is become difficult to deal with, and the accuracy of detection will be greatly reduced

Moreover, in the process of image frame similarity matching, there are often some "noises" in the matching results, and the existing methods cannot remove these "noises", resulting in the inability to find the optimal matching subsequence from the unordered image frame matching results

[0007] Existing copy video detection research cannot effectively meet the needs of video retrieval, copyright protection, and video content supervision. It is particularly important to study a fast and effective copy video detection method

Images