Research On Video Lens Edge Detection And Key Frame Extraction Algorithm

With the rapid development of information technology,the Internet and multimedia technologies are widely used,and a large amount of video information is produced in people's daily lives and work.They bring great convenience to people's lives.However,in the face of complex and diverse video data content,how to store and retrieve video in a large amount of video data has become a hot topic in video research and needs to be solved.Based on this,this paper mainly studies some issues in video retrieval,mainly including shot detection and key frame extraction,and introduces the lens edge detection algorithm and key frame extraction algorithm in detail,and analyzes the characteristics of each algorithm.In the feature extraction technology of video frames,both of the two aspects of lens detection and key frame extraction are first extracted by features,and then normalized methods are used for feature fusion to obtain significant features.The former adopts an improved traditional method to extract features of video frame images,and the latter uses a deep convolutional neural network method to extract features.Experiments show that in the application of this article showed good results.In the lens segmentation technique,this paper proposes a lens detection method based on an improved dual detection model.In this method,in the lens edge detection algorithm,the concept of non-uniform block color histogram discontinuity is first used,and the sliding window adaptive binary search algorithm is used to perform initial detection of the shot boundary,so that the initial shot detection set is obtained.Then the multi-feature fusion adaptive dual threshold method is used to re-examine the edge of the video lens,and the texture and tone product characteristics of the n frames before and after the scene where the shot change may occur in the initial inspection are used,and cross-scale merge and normalization is adopted.The algorithm obtains a comprehensive feature saliency map,and on this basis,an adaptive dual-valve value is used to further determine whether it is a gradual change lens,an abrupt lens,and an interference-caused lens misdetection,thereby accurately and efficiently detecting the edge of the video lens.Experiments show that the algorithm has a good adaptability in the video lens edge detection.In key frame extraction technology,this paper proposes a key frame extraction algorithm based on deep convolutional neural.That is,deep learning techniques are used to extract key frames.Firstly,the convolutional neural network is used to extract the advantages of video frame image feature.The features of video frame images extracted by convolution layer,downsampling layer and full link layer are used to fuse the features of saliency maps.Then the similarity measure algorithm is used to perform the similarity measure between frames.Extraction of keyframes.Experiments show that the key frames extracted by this method can relatively express the main content of the video.