Broken document splicing method based on DFS and improved center clustering method

Abstract

The invention belongs to the technical field of broken document splicing and discloses a broken document splicing method based on DFS and an improved center clustering method. The method comprises the following steps: carrying out splicing based on similarity, and carrying out error correction based on difference degree; and based on the improved center clustering method and a DFS splicing restoration algorithm, fully excavating matching degree of two pieces of fragments based on parameters of similarity degree, difference degree and black threshold and the like in the DFS splicing restoration algorithm. The method further reduces human factors and obtains a good clustering effect; splicing is carried out based on similarity, and error correction is carried out based on difference degree, thereby avoiding a genetic algorithm and improving accuracy of the algorithm; and by deducing a threshold range formula, the effect of the center clustering method is allowed to be better, and matching degree of the two pieces of fragments is fully excavated based on parameters of similarity degree, difference degree and black threshold and the like in the DFS splicing restoration algorithm, thereby further reducing human intervention.

Description

technical field [0001] The invention belongs to the technical field of broken document splicing, and in particular relates to a broken document splicing method based on DFS and an improved central clustering method. Background technique [0002] At present, the rule-cut document fragment splicing problem is mostly divided into two steps. The first step is to use the clustering method to cluster all the fragments in order to find out the fragments that originally belonged to a row, which is convenient for subsequent processing. In the second step, the shreds belonging to a row are spliced ​​separately based on the similarity algorithm or genetic algorithm, and manual intervention is performed in some special splicing situations. However, due to the complexity of cutting paper fragments, clustering methods and splicing algorithms are often unable to cope with special situations. The splicing algorithm at this stage mainly has the following disadvantages: the clustering algori...

AI Technical Summary

Problems solved by technology

However, due to the complex cutting of shredded paper, clustering methods and stitching algorithms are often unable to cope with special situations

The splicing algorithm at this stage mainly contains the following disadvantages: the clustering algorithm is prone to failure

For special situations such as blanks before paragraphs caused by paragraphs in documents, clustering algorithms at this stage often fail to cope with them

The principle of splicing algorithm based on similarity is too single, lacks robustness, and cannot cover all possibilities in the splicing process

For NP-hard problems such as fragment splicing, as the scale of fragment splicing increases, this method does not achieve very good results

On the other hand, the genetic algorithm, a commonly used optimization algorithm for solving NP-hard problems, does not perform well in practice, mainly because parameters such as population size, population mutation rate, and crossover rate in the genetic algorithm need to be set artificially. If it is unreasonable, the calculation result will converge to the local optimum and the desired effect will not be achieved.

[0003] To sum up, the problem existing in the prior art is that the splicing algorithms that exist at this stage either use clustering methods or genetic algorithms

However, the clustering features selected by the clustering method are too single, and the genetic algorithm needs to manually set parameters. The above reasons lead to the poor effect of the algorithm on the actual fragment splicing

Images