Automatic Seal Verification by Calculating Distance Between 2D and 3D Patterns

This paper describes an efficient matching approach for automatic seal verification. Most of the conventional approaches to this problem used a ref:% @ @ (& @ erence seal impression registered as a 2D image. . L , However, it was difficult to absorb various input seal f i e seal impressions. patterns by only one reference pattern. In our previous study, one of the authors has been proposed a new algorithm in which the reference seal was measured as a 3D range image. The verification rate was improved from that by using the 2D reference image. But, in the previous algorithm, as the seal e @ @ @ @ @ verification was performed in 2D pattern space by '. A generating the optimum 2D reference seal impresForgery seal impressions for the above true them. sion for the input pattern from the 3D reference image, this way still cannot verify random dot-patterns Figure 1: Examples of seal impressions. of the seal-ink. In this paper, our approach uses 3D shape information directly in the matching process. The input seal is verified by calculating Euclidean distance from the 3D reference seal. The excellent recognition rate achieved in all the performed experiments indicate that the proposed matching is well-suited for seal verification. Moreover, the verification speed was improved remarkably.