Zernike 모멘트와 LAT(Local Affine Transformatin)를 이용한 필기체 숫자 인식

Abstract

본 논문에서는 오프 라인 필기체 숫자 패턴에 대하여 Zernike 모멘트와 LAT(Local affine transformation)을 이용한 패턴 정합에 관하여 연구하였다. LAT 정합은 다른 정합 방법들보다 다소 시간이 많이 소요되기 때문에 LAT 정합에서 소요되는 시간을 줄이기 위해서 Zernike 모멘트을 이용하여서 미리 LAT 정합에 사용되어질 마스크 패턴을 선별하였다. LAT 정합에서는 인식의 신뢰도를 향상시키기 위하여 구조적 특징인 교차수와 주축을 첨가하여 LAT 정합 조건식의 가중치로 적용하였다. 실험은 필기체 숫자 패턴에 대해서 수행하고, Axil Workstation에서 시뮬레이션을 하였고, 인식률은 91.6%였다.

An off-line pattern matching method of handwritten numerals using Zernike moment and the local affine transformation(LAT) is proposed. Pattern matching is one of the character recognition methods in which mask patterns are needed.

The method using Zernike moment and LAT is suitable to recognize handwritten numeric characters, since it is invariant in scale, rotation, shearing, and translation. To improve the performance, the principal angles and crossing counts are combined to LAT as the weight of the objective function.

Several handwritten characters are used and simulated on Axil workstation. The recognition rate of 91.6% is achieved.

An off-line pattern matching method of handwritten numerals using Zernike moment and the local affine transformation(LAT) is proposed. Pattern matching is one of the character recognition methods in which mask patterns are needed.

The method using Zernike moment and LAT is suitable to recognize handwritten numeric characters, since it is invariant in scale, rotation, shearing, and translation. To improve the performance, the principal angles and crossing counts are combined to LAT as the weight of the objective function.

Several handwritten characters are used and simulated on Axil workstation. The recognition rate of 91.6% is achieved.