Research On Feature Extraction Technology In Video-copyright Blockchain

In the Internet age,not only intellectual property copyright,but also the protection of originality of derivatives such as original articles,web photography,small video and so on.However,the rich promotion channels and the growing number of users directly or indirectly lead to the wanton dissemination of information.With the improvement of copyright awareness,Blockchain technology has been expanded in the field of digital copyright management.Blockchains are open,transparent,and use high redundancy to ensure data security.Digital copyright management based on Blockchain technology relies on new concepts such as the weak-centralized distributed architecture of Blockchain,which provides an effective way for digital copyright management.However,the high redundancy of storage makes it easier to redundant data if the large-footprint digital copyright is fully stored in the Blockchain.Based on the above,considering the significance of storage on large files such as videos,this paper introduces a way of extracting key information using deep learning algorithm.With the continuous expansion of comprehensive in various fields,deep learning technology is widely used in video content recognition,content detection,target recognition and so on.In terms of content recognition,deep learning has different ways of dealing with the global or local characteristics of video.In this paper,deep learning algorithm Open Pose based on spatial feature extraction and the deep learning algorithm LRCN based on spatial-temporal feature extraction are analyzed respectively.In terms of device performance,Open Pose algorithm can run on CPU and GPU with lower performance,and is relatively more efficient than LRCN.In this paper,we use Open Pose to output key points of BODY?18 for the posture of characters in human video,and use the double-SHA256 algorithm to generate Hash values as transaction data stored in blocks.Store it asfeature information in the Blockchain.This method not only ensures the privacy of large file data information,but also ensures the efficient storage of information,and improves the practicability of the system.This paper takes in-depth learning technology as the core of feature extraction and Blockchain technology as the data storage architecture,designs and implements a simulation system for video copyright key information data.This reduces the cost of copyright storage and ensures that critical information cannot be modified.