Research On The Application Of Reinforcement Learning In The Ceramic Industry And The Field Of Reversible Data Hiding

In recent years,reinforcement learning has been widely applied in many fields,and its decision methods are often used to solve sequential decision problems due to their strong universality.As an important development achievement of artificial intelligence,reinforcement learning endows agents with the ability of self supervised learning,which can continuously learn according to the rewards and punishments obtained in training,and ultimately make high-level decisions based on learning experience.Therefore,reinforcement learning has become an important means of understanding decision control and optimization problems that cannot be ignored.This article focuses on the application of reinforcement learning algorithms in the ceramic industry and reversible data hiding field,in order to solve decision control and optimization problems in these two fields.The main research content is as follows:(1)Application research in the ceramic industry: The modern ceramic production process involves the regulation of many process parameters,some of which decision-making behaviors can affect the quality of the final ceramic product.The process parameters of traditional ceramic production processes are determined through multiple experiments based on the experience and behavior of engineers.However,empirical behavior often makes it difficult to accurately determine process parameters,and process parameters can dynamically change with actual operating conditions(such as external atmosphere).Under the influence of changing working conditions,it is difficult to ensure the stability of product quality.In order to solve manual decision-making problems dominated by experience and achieve dynamic updates of parameters in actual processes(operating conditions).This article proposes for the first time an intelligent decision-making framework for ceramic production based on deep reinforcement learning algorithm.The framework constructed includes: prediction model and decision model(i.e.environment module and intelligent agent module).Based on data mining technology,simulate and update various working conditions in the production process.The random forest algorithm is used to build a prediction model based on the production data of multiple production links(including spray drying links)collected on site to predict the corresponding product quality in a timely manner.The decision-making model can quickly and adaptively adjust the process parameters according to the predicted product quality.In order to verify the effectiveness of the decision-making model,the spray drying process is taken as an example to make the ceramic tile products achieve the expected product quality.The experimental results show that the accuracy of the prediction model is better than that of other similar methods for product quality prediction,with an average improvement rate of 2%.At the same time,after several iterations,the intelligent decision-making algorithm of spray drying can improve the qualification of production process parameters to 95%,with good control effect.Prove that this decision-making framework can be applied to parameter decision-making control in other stages of ceramic production,or to parameter optimization problems in multiple stages.(2)Application research in the field of reversible data hiding: With the rapid development of information technology and internet technology,people can publicly or obtain digital information from various channels,and multimedia information security issues are increasingly receiving attention.Reversible data hiding technology,as an effective method to protect information security,has attracted widespread attention due to its ability to extract secret information from the encrypted carrier of hidden information and accurately recover the original carrier.The rate distortion(capacity and distortion)performance of reversible data hiding algorithms largely depends on the steepness of the histogram and the selection of edge information(peak and zero points).Existing algorithms use genetic algorithms to select edge information and minimize distortion in dense images.This article improves existing reversible data hiding algorithms and proposes an improved reversible data hiding algorithm based on reinforcement learning.Utilizing reinforcement learning(Q-Learning)algorithm to adaptively adjust crossover and mutation parameters to improve the algorithm’s search ability and optimize the rate distortion performance of images.The experimental results show that the convergence speed of our method is faster than that of genetic algorithm.At the same time,under the same embedding capacity(such as 0.7bpp),the peak signal-to-noise ratio reaches 39.72 dB.