Research And Application Of Video Summary Algorithm

The development of Internet,especially the development of screen recording technology,has enabled many people to learn professional knowledge by watching videos,and to monitor certain screens to understand user behavior.However,while technology brings convenience to users,it also brings many problems.Due to the large amount of video data,it takes a lot of time for all viewing,which brings a burden of viewing to the user.Therefore,it is a problem worth studying how to make the user's information on the screen recorded video convenient and quick.In this paper,based on the research of existing video summary algorithm and the characteristics of screen recording video,three different video summary algorithms are proposed for different situations.The proposed video summary algorithm can effectively reduce the time for users to watch video and improve efficiency.In order to verify the application effect of the video summary algorithm,this paper also designed a paperless test anti-cheating system,and the video summary algorithm is well applied.The main work and innovations of this paper are as follows:(1)A video summary algorithm based on block sparse representation is proposed.The algorithm mainly on multiple screens records video with similar content and is used to produce video summaries of multiple videos.The algorithm first divides the video into similarities and controls the size of the block.Multiple video blocks are then embedded into an embedded space and a representative video block is selected using a sparse representation.Finally,the key frames are selected from the representative video blocks to form the final video summary.(2)A video summary algorithm based on multi-layer dictionary learning is proposed.The algorithm mainly uses for screen records video with small changes in frequency.The algorithm uses the idea of multi-layer dictionary learning,first learns to initialize the dictionary and uses the initialization dictionary to sparsely represent subsequent frames.When the reconstruction error is greater than the threshold,indicating that the content that has not appeared before is present in the current frame,the frame is added to the video summary,and the dictionary is updated with the frame.A video frame with similar content later can be well represented by a dictionary without causing a large reconstruction error.This process continues until the end of the video.Finally,the redundant frames in the video summary are removed to form the final video summary.Experiments show that the algorithm can handle video recording on a single screen very well.(3)A video summary algorithm based on dictionary selection and k-means clustering is proposed.The video of the candidate recorded in this article takes a long time,and it takes a lot of time to find evidence of the cheating of the candidate.In order to save the time for teachers to find evidence of cheating for candidates,an algorithm for video with less variation and greater redundancy for test video is proposed.The algorithm can quickly filter out a large number of redundant frames in the video,and extract a dozen key frames according to the algorithm to generate video summary files,which effectively reduces the time for teachers to find evidence of cheating of candidates,and improves the efficiency of anti-cheating system.(4)A paperless test anti-cheating method was proposed to verify the third video summary algorithm.The system is mainly divided into three modules: examination machine,teacher machine and management machine.The test machine mainly has functions such as screen recording,recording log,and generating video summary files;the teacher machine mainly has functions such as controlling the recording status of the test machine,storing the video summary file,and searching for the video summary;the management machine mainly has log analysis,video download,and video playback.Experiments show that this method can have a better effect in obtaining cheat frames.