基于内容特征的碎片拟合算法研究

Research on Reconstruction Algorithm of Document Fragments Based on Context Features

针对碎片拟合过程中存在大量由碎纸机形成的相似、相近甚至相同的碎片边缘,再好的边缘拟合算法也难以正确选择唯一候选碎片边缘的问题。提出了基于碎片中文字、图、表等内容信息在碎片边缘留下的内容特征,判定与目标碎片匹配候选碎片的思路。界定了特征点、特征向量等碎片内容特征的概念,给出特征点、特征向量的提取算法以及基于内容的碎片拟合算法。实验结果表明该算法正确、有效,为计算机自动合成碎片奠定了基础。

It is very difficult and time-consuming to choose the best candidate using edge contours only,because of the similar and same edge found among document pieces during reconstructing shredded documents.Therefore,an operator is needed to instruct computers to join all the broken pieces together,which would undoubtedly slow down the process. An algorithm to choose the best matching piece with the target is proposed based on distinguishing features of characters,tables,and figures intersecting the outer contour. The connotations of feature points and feature vectors are firstly presented. Then,the algorithm to extract feature points and feature vectors are explained in detail. Finally,an algorithm to reconstruct all the shredded pieces are developed. The experiment results show that the proposed approach is robust and efficient for document pieces to be matched automatically.