Method for outputting size-fixed sequence DPP (determinant point process) video summary based on dynamic planning

Abstract

The invention discloses a method for outputting a size-fixed sequence DPP (determinant point process) video summary based on dynamic planning. The method comprises the following steps: 1, converting an input video into a video frame set; 2, uniformly sampling the video frames in the video frame set obtained in the step 1 at a speed of 1 frame per second to obtain a new video frame set; 3, performing image processing on each video frame in the video frame set obtained in the step 2, and extracting a feature vector of each video frame; 4, performing machine learning by using the feature vector obtained in the step 3, and establishing an initial DPP model; 5, on the DPP model obtained in the step 4, outputting the number of required key frames to the user by using a k-SDPP method. According to the method, a user can adjust the number of the output key frames according to own needs while the sequence between the video frames is considered, and the defects of a seqDPP algorithm and a k-DPPalgorithm are overcome.

Description

technical field [0001] The invention belongs to the technical field of computers, and in particular relates to a video summarization method. Background technique [0002] With the rapid development of science and technology, especially the rise of smart phones, almost every mobile phone has a camera function. In other words, in today's era, each of us is a potential photographer, and video data has entered An era of explosive growth. According to statistics, 144,000 hours of video data are uploaded to YouTube every day. In addition, in China, live broadcast platforms such as Douyu, Huya, and Douyin are also "prolific producers" of video data. We are accumulating a large amount of video data every day. Video data has grown explosively, especially the rise of Internet products such as Weibo and live broadcast, which has driven the rapid growth of video data. According to statistics, in 2018, video data accounted for 60% of mobile phone data every month, with a total of 28EB....

AI Technical Summary

Problems solved by technology

The seqDPP algorithm proposed by Gong et al. can overcome this shortcoming. The Bayesian structure used by the seqDPP algorithm fully considers the sequence between video frames, but it also has a problem that the number of output key frames cannot be fixed.

That is to say, before the result of the algorithm is run, the user does not know how many key frames it will output, and the number of output key frames is not adjustable, that is, the user cannot specify the number of key frames before the algorithm runs, and cannot Output the number of keyframes according to the user's needs

Another algorithm called k-DPP can fix the number of output key frames, but it ignores the sequence between video frames, so the output result of k-DPP is far from that of seqDPP, and the user satisfaction looks a bit unsatisfactory

Images