Research On The Restoration Algorithm Of Paper Fragments Splicing Based On Deep Convolution Neural Network

The problem of splicing and restoration of paper damaged documents is that the complete documents are decomposed into various document fragments by the external force.Each document fragment needs to be spliced and restored into a complete document,and the process of obtaining the information that the document needs to express.The study of this problem has important application value in the restoration of judicial evidence,the acquisition of information and the restoration of historical documents.In this paper,we study the process of splicing and restoring electronic documents after they are printed on paper,which is called the problem of splicing and restoring regular paper documents.The goal of deep convolution neural network model learning is to build a classifier through neural network simulation,which can learn the classification standard of the data set from the large-scale marked data set and save it in the neural convolution network model,and predict which classification our data set to be predicted belongs to according to the classification standard learned from the model,which is powerful Application ability has been embodied in the field of image classification and recognition and has been applied to production and life.Clustering analysis is also a branch of machine learning theory,which has a wide range of applications.Its main goal is to divide the data set to be processed into different categories according to its characteristics.According to different category characteristics,each data can be processed pertinently,so as to reduce the degree of data confusion and the scale of data to be processed.The main idea is to abstract the characteristics of the data in the data set into "distance",and then classify the data elements with similar distance into the same category according to the distance between the data elements through the elements in the data set.In this paper,deep convolution neural network and clustering analysis are used to solve the problem of stitching and restoration of paper damaged documents with rules of crosscutting and lengthwise cutting.The main steps are as follows: first,use the deep convolution neural network to train the stitching and restoration of damaged documents in the same line,then extract the characteristics of damaged documents according to the characteristics of damaged documents,use cluster analysis to classify the damaged fragments of the same trade,and then use the trained deep convolution neural network to complete each One line of damaged documents is spliced,and then after a little manual splicing,the splicing task of each line is completed.Finally,according to the characteristics of each line of damaged docume nts,the splicing task of the whole paper damaged documents is completed.Experiments show that this method can speed up the splicing and restoration of damaged paper documents,so this method has better application value.The main contributions of this paper are as follows:·In order to reduce the scale and difficulty of the problem processing,the paper puts forward the application of cluster analysis to classify the paper damaged document fragment s into different categories according to the characteristics of the same line height and the same space between each line of text contained in the paper damaged document fragments.·In this paper,according to the characteristics of the broken paper documents,a dataset with the same characteristics as the broken paper documents is constructed,and the dataset is used to train the deep convolution neural network model and apply the trained model to the splicing and restoration of the broken paper documents.·This paper proposes the general processing steps of the stitching restoration algorithm for regular paper damaged documents and verifies its effectiveness through experiments in this paper.