Research On Digital Image Transaction Management Based On Watermark And Blockchain Technology

A large number of digital image transaction websites have appeared at home and abroad to meet the demands for digital images in most social fields.But in general,there are some deficiencies in those websites.For example,the service for users is insufficient in image copyright protection and digital image copyright authentication,the service for users is insufficient in the preservation of transaction evidence which has legal effect,the personalized services for users are insufficient,and few types of business are provided.There are also some problems in the theoretical research of digital image transaction management,such as the lack of systematic research,and the shortage of research on the copyright protection and content authentication of digital images,on the management of smart contracts which are used by transactions in blockchain,and on the personalized recommendation of digital images.The study is done in order to solve those existing problems.In the aspect of theoretical research,some problems of digital image transaction management before,during and after transaction are studied.In this paper,some digital image multipurpose zero watermarking algorithms are proposed for the copyright protection and copyright authentication of digital images before transaction by using digital watermarking technology.The transaction information generated in the process of digital image transaction is written into the blockchain for storage.The smart contract classification algorithms of blockchain are studied.According to the user's historical data stored after data image transaction,a personalized digital image recommendation algorithm based on intelligent recommendation technology is studied.In the aspect of practical research,a digital image transaction management system is designed.The proposed algorithms are applied to the system,and the corresponding management models are designed to solve the problem of lacking types of services provided by digital image transaction website.The contributions of the study are summarized as follows:(1)Two kinds of digital image multipurpose zero watermarking algorithms are proposed which are based on singular value decomposition and deep learning.Before the digital image transaction,research on digital image copyright protection and content authentication is insufficient.The study of digital image copyright protection and content authentication is conducted to solve this problem.The watermark algorithm in the transform domain algorithm is more robust than the watermark algorithm in the spatial algorithm.Some shortcomings of discrete Fourier transform and discrete cosine transform are overcome by discrete wavelet transform(DWT).The singular value obtained by singular value decomposition(SVD)can represent the inherent algebraic characteristics of the image and has good stability.The key features of the image can be obtained by the deep neural network.Therefore,a digital image multipurpose zero watermarking algorithm based on SVD and deep convolutional neural network(DCNN),and a digital image multipurpose zero watermarking algorithm based on SVD and deep belief network(DBN)are proposed through the combination of DWT,SVD with DCNN and DBN respectively.The zero robust watermark image and zero semi fragile watermark image are both been constructed in the two kinds of zero watermarking algorithms.The simulation results show that the robust watermarks of the two algorithms have good resistance to various strong attacks,and the extracted semi fragile watermark image can also locate the tampered position of the original image.(2)Two smart contract classification algorithms are proposed.The blockchain technology is introduced into digital image trading process management in order to overcome the lack of legal effect in preservation of transaction evidence for transaction in many digital image transaction websites.In the process of transactions on the blockchain,smart contracts need to be used.Therefore,the smart contract classification algorithms are studied in order to manage smart contracts effectively.Smart contract belongs to text information.The numbers of different types of smart contract are often of great difference.Therefore,the smart contract classification belongs to unbalanced text classification.The first step for smart contract classification is converting the smart contract into data that can be recognized by computer.Since there is no corpus for smart contracts now,Word2 Vec is used to build a smart contract corpus firstly.Then all smart contracts are transformed into digital vectors of equal length using Word2 Vec and smart contract corpus.The second step for smart contract classification is studying how to classify digital smart contracts.Because the smart contract classification belongs to unbalanced text classification,random weight to extreme learning machine(ELM)and the weighted categorical entropy loss function are used to overcome the defect of traditional classification methods.The auto-encoder has the characteristic of reducing the dimension of data.The bi-directional long short-term memory(Bi LSTM)has the ability to memorize context.Therefore,a smart contract classification algorithm based on auto-encoder random weight ELM network and a bi-directional long short-term memory smart contract classification algorithm based on weighted categorical entropy loss function are proposed.The results of the simulation experiments of two smart contract classification algorithms show that the two algorithms are effective.(3)A personalized digital image intelligent recommendation algorithm based on weighted Text Rank and self-organizing feature map(SOM)is proposed.In order to solve the problem of insufficient research on the personalized recommendation service provided for users in the management after the completion of digital image transaction,the personalized digital image recommendation is studied.The images of interest to users are obtained through two sources.The first source is the images in the order of the similar users of the current user.The second source is the same type of images in the database as those placed in the order by the current user in the last time.Some images that users are interested in are selected from the candidate image sets which are collected from similar users and databases,and are recommended to the current user.The influence degree is different in reflecting the interest of user such as the user's final search words in the website,the images that are put into order and car at different time,the key words which can represent the types of images in the user's historical data.Therefore,utilizing the characteristics that Text Rank algorithm is suitable for extracting short text keywords,this study designs the weighted Text Rank algorithm to extract the keywords of the user's historical data.The SOM is used to find similar users of the current user because of the characteristics of SOM which can cluster data through competition.Simulation results show that the proposed algorithm can find the similar users of the current user effectively,and can recommend digital images to the current user which he is interested in.(4)A digital image transaction management system is designed.The digital image transaction management system is designed to deal with the problem that digital image processing,digital image copyright protection,digital image authentication,transaction storage,personalized recommendation service,and smart contract classification are not provided in many digital image transaction websites.This study designs the framework and functions of digital transaction management system,and some models of digital image transaction management,data security protection management,copyright protection management,smart contract management,and personalized recommendation management,which can be applied in the digital image transaction management system.To a certain extent,the above methods can solve such problems as the insufficient service in digital image copyright protection and copyright authentication,preservation of transaction evidence which has legal effect,smart contract classification in the application of blockchain technology,and less types of website business.The research method of applying the technology of digital watermark,blockchain and intelligent recommendation to digital image transaction management can provide some ideas and methods for the research of management science.If the methods are applied,the businesses profits can be increased because the increase of business types and personalized digital image recommendation for users can attract more users.