KEY POINTS OPTIMIZATION USING SIFT ALGORITHM FOR SIGNATURE IDENTIFICATION

Abstract

SIFT Algorithm is a popular method for finding features of an object. It uses the principle of repetitive method for finding key points within an object. The number of rounds for analyzing key points for each picture is not equal leading to the doubt as to whether or not the identified key points are appropriate. This research proposed a method for determining the number of rounds to be analyzed for obtaining the optimized key points for signature identification by comparing calculated Euclidean distances of signature key points in each round with the assigned value of  as 0.2, and using the signature from optimized round to designate the key points of stored original signature for comparison with a test signature by the SIFT Algorithm with the assigned value of  as 0.25, by the SIFT Algorithm. In addition, the comparison of signatures was also performed by 10 individual persons. In this research, samples of 200 signatures from 20 persons, each providing 10 signatures, were tested and found that the optimized key points were found in the sixth round with the distance between the stored original signatures and the signatures used for testing, using SIFT Algorithm, being less than or equal to the assigned  value. The distance from the comparisons of 20 sample signatures, between 20 genuine signatures and their 20 test pairs, using SIFT Algorithm, was less than the assigned  value. The test results showed that the comparison success percentage using the SIFT Algorithm was 82.50 and the comparison success percentage by 10 individual persons was 77.25.