Research On Paper Pieces Reconstruction Method For Document Security

Converting paper-documents into paper pieces is the most common way to protect document information.Since paper pieces are numerous,mixed,and indistinguishable,it is extremely challenging to recover broken documents from paper pieces by a reverse operation.Especially with the advancement of technology,the way of breaking paper documents has been changed from manual tearing to shredder cutting,which makes the recover even more difficult since the paper pieces are smaller and more similar with each other.Although there are great difficulties,it is significant to study reconstruction of paper pieces because the destroyed document usually contains important or sensitive information.If this kind of high-value information can be recovered,it would have a major impact on the country,the enterprise or the individual.Moreover,with the advent of the information age,people pay more attention to information security.Paper pieces reconstruction has become a hot issue for domestic and foreign scholars because of the particularity of its research objects and the frontier of research goals.This research synthesizes multidisciplinary knowledge such as computer vision,pattern recognition,mathematical statistics,signal processing,data mining,and cryptanalysis.Nowadays,with the increasing complexity of paper pieces,the existing reconstruction methods have many disadvantages,and there is still a long way to practical application of this technology.In order to effectively realize the recovery of broken documents,this paper focuses on several aspects of paper pieces reconstruction such as the framework of paper pieces reconstruction,clustering of paper pieces,paper pieces matching,and pieces matching optimization methods.This dissertation carries out the research as follow:(1)A common framework of paper pieces reconstruction is established.This framework is constructed according to the characteristics of paper pieces.First,in the pieces acquisition module,paper pieces are converted into digital images,and image preprocessing methods are used to obtain standardized pieces.Secondly,in the pieces clustering module,the mixed pieces are classified according to their sources.Finally,in the pieces matching module,the disordered pieces are rearranged to restore the original appearance of the broken document.The framework is reasonable and simple,and it can cope with a variety of complex reconstruction situations,effectively achieving automatic recovery of broken documents.(2)A paper pieces clustering method based on document layout is proposed.This method makes full use of the distribution characteristics of characters in paper pieces and the correlation of text lines,estimates the clustering number and starting point of clusters of the pieces accurately,and realizes the clustering of paper pieces effectively by combining the structure layout of the document itself.Because this method digs deep into the intrinsic properties of paper documents and grasps the differences and connections between pieces accurately,it achieves good clustering effects when dealing with complex homologous pieces clustering problems.(3)A paper pieces matching method based on character structure correlation is proposed.According to the structural characteristics of the characters in the pieces,this method describes the characters graphically,and uses the number of mismatched combinations and the summation of the matching probability as a measure of the pieces matching,combining the regular pattern of character reconstruction,to realize the paper pieces matching by a mutual correction algorithm.This method,which has high accuracy and great stability,can overcome the interference caused by factors such as font transformation,text skew and text defect,and it has achieved good matching effect in experimental results.(4)A piece matching optimization method based on genetic strategy is proposed.According to the property of paper fragments,this method first converts the order of the pieces by sequence coding,and then uses the novel fitness function to guide the global search of the pieces,while improving the search efficiency by genetic operator,and finally improve the computational performance of the algorithm by optimizing the operating parameters.This method has strong search ability and high matching accuracy.It can implement matching optimization of pieces effectively in a global scope.