Research On Key Frame Extraction Technology Of Underwater Video Based On Python Platform

Key frame extraction is the basis of video processing technology.Whether in video fusion or video analysis,video key frame extraction plays an important role.Video key frame extraction is generally divided into two parts which are shot segmentation and key frame extraction for the segmented shot.There is a great quantity of calculation hidden in each simple process.Firstly,the theoretical system used in the whole algorithm is introduced in detail,including the hierarchical structure of video,the classification of shot segmentation algorithm,the status of clustering algorithm in machine learning and the comparing method of feature vectors.This paper reproduces the optimized algorithm in the field of common video key frame extraction.For shot segmentation,the method used by original algorithm is histogram,and the clustering algorithm uses adaptive threshold to upgrade K-means algorithm.However,underwater video has the characteristics of high noise,excessive shot disorder frames and high image ambiguity.Therefore,in the view of the special situation of underwater video,this paper proposes an improved algorithm.The improved algorithm introduces the texture features of video frames which can effectively reduce the probability of gradual change and shear frame error detection caused by the interference of brightness and chroma.At the same time,by optimizing the initial clustering center,the problem of clustering directivity caused by dense continuous frames is effectively avoided.Finally,the addition of wavelet transform can effectively suppress the noise of underwater video itself.On the premise of guaranteeing that the extracted key frames can express the main content of video smoothly,the improved algorithm can further reduce redundant frames compared with the original algorithm.