Video summarization for surveillance system using clustering algorithm and deep learning based classification

Abstract

Video surveillance plays a key role in ensuring safety and security newlinewith the enhancement of Digital Video (DV) technology. Owing to the newlineincreasing number of real-world applications, the Surveillance System (SS) is newlinean active research area in Computer Vision. Analyzing the video always newlineremains a challenge, increasing the burden of the Video Analyst for predicting newlinethe interactions and activities pertaining to the particular event. For newlineinvestigating along with evaluating the activities, the SSs are employed in newlinemany applications. Thus, there is an urgent requirement for an Intelligent newlineSystem that would offer the required solution for the current scenario by newlineevaluating the video data. Enormous data is produced, stored together with newlineprocessed for security purposes as of the single or multi-surveillance camera. newlineIt is challenging for an analyst to go through the entire content owing to time newlineconstraints. The Video Summarization (VS) is presented to manage the Video newlineContent (VC). Current VS methodologies have attempted to offer the VS but the newlineschemes have Execution Time (ET) along with condense the content of a newlinevideo in a domain-specific manner. For indicating the similarity betwixt and#8215;2 newlineimages, many prevailing techniques utilize color spaces like RGB or HSV in newlinethe estimation of Mutual Information (MI) for Key-Frame (KF) extraction. newlineNevertheless, every current technique is time-consuming along with newlinefrequently ineffectual. Most prevailing systems relays on shot detection that newlinebecame inaccurate owing to the availability of various kinds of transitions like newlinefade in, fade out, abrupt cut, etc betwixt and#8215;2 adjacent Video Frames (VFs). newlineBesides, in the video s final step, it failed to offer the order of frames. newlineBy employing normalized K-Means (KM) and Quick Sort (QS), newlineefficient VS for SS was proposed for surpassing those disadvantages. newlinePresenting the effective VS for SS deploying optimization-centric newlineclassification algorithms is the key goal.(1) Split video into frames (2) pre-sampling, (3) provide ID number, (4) Feature Extraction (FE), (5) Feature Selection newline(FS), (6) clustering, (7) extract frames, (8) video summary are newlinethe eight steps of 1st approach. From the YouTube videos, the newlineinput video is gathered; in addition, it is segmented into newlineframes. newline newline