The second problem is the recovery of vertical and horizontal debris.

We establish the second model and give a segmentation line grouping algorithm based on text line features. The segments in the line grouping are horizontally spliced to obtain the restored segment lines, and then vertically spliced to obtain the final restoration result. These two splicing strategies are based on the traveling salesman problem in the model 1 Among them, the characteristics of text lines are the regularity between text lines. Using the regularity of text lines, we can not only group fragments into lines, but also? High-accuracy vertical text mosaic.

According to the second model, there is no manual intervention in the debris reduction results in Annex 3; In the process of Appendix 4 fragment repair, there is manual intervention after the horizontal splicing of grouped fragments, that is, the repair results of individual fragments are adjusted manually occasionally.