Similar or repeated video retrieval method and device

Abstract

The invention relates to a similar or repeated video retrieval method and device, the device comprises an application program module, a task scheduling module, a gene extraction module and a gene database module, the method comprises the following steps: firstly extracting video genes from all videos in a video database, and inserting the video genes into a Milvus vector database; during retrieval, and the gene of the video to be retrieved is also extracted, a retrieval request is initiated, and the gene database returns a retrieval result which specifically comprises video gene storage and similar video retrieval. The invention has the advantages that 1) compared with the prior art, the retrieval accuracy is higher, and the situation of missing detection can be reduced; (2) retrieval results are sorted according to the number of similar pictures, and the retrieval requirements of users are better met; 3) the retrieval result contains the overlap ratio of the to-be-retrieved video and the target video, so that the user can judge whether the target video is a repeated video or not; and 4) the invention does not depend on video text description information, and only needs to provide a video file.

Description

technical field [0001] The invention relates to a similar or repeated video retrieval method and device, belonging to the technical field of computer multimedia (image and video) information processing. Background technique [0002] Nowadays, the Internet is full of a large number of repetitive or similar videos. How to quickly retrieve duplicate videos from massive videos has become an increasingly urgent need. [0003] The solutions in the prior art, one is to set up a database based on text information such as video title or introduction and then perform text retrieval, and the other is to use a vector database for retrieval (Li Zhaoxing "Similar Video Search—Opera's Milvus Practice"). Extract the video image features and record the vector database, then calculate the average similarity between the video to be retrieved and the hit video image set in the retrieval results, sort according to the average similarity of the image set, and the average similarity is lower than ...

AI Technical Summary

Problems solved by technology

[0004] However, in the absence of video information, the text retrieval scheme is not applicable. For example, sometimes there is only a batch of video files without any video description information. If you want to find duplicate videos in this batch of videos, you cannot use text retrieval. method

On the other hand, the text description information is often not comprehensive enough, and sometimes the recall rate of retrieval results is low due to lack of information

[0005] Although the method based on the average similarity of the vector database combined with the image set does not depend on the text description information, the similarity between the image set and the image set is defined as "the average value of the similarity between the image and the image set", resulting in sometimes The problem that videos with low overlap (duration of similar clips) rank high

For example, the hit video and the video to be retrieved contain the same advertisement, and the content of the advertisement is exactly the same. At this time, the hit video will be ranked very high in the retrieval results, while the ranking results of the video with a high degree of overlap will be lower, which is obviously impossible. Meet the needs of use

Images