Recognizing Receivers Of Official Letters With Template Matching Method And Learning Vector Quantization
By : Gede Ngurah Joy Indra P
Faculties : Fakultas Matematika dan Ilmu Pengetahuan Alam
Department : S1 Ilmu Komputer
An official letter is a letter containing official matters or specific business addressed to departments or individuals. Today nearly all institutions have a computerized system to facilitate their administration and it is very possible that the system can automatically send a letter that has been digitalized to the recipient. In the existing system, in its aplication, the systems made still using manual input in the data details of the letter entry. In this study a system is developed that can recognize automatically details of the letter, especially in the recipientís address. Two major problems in this study, namely how the Template Matching method obtain the position of mailing address or the recipient of letters in the image and how Learning Vector Quantization perform recognition of mail address or recipients. In the Template Matching test methods in the search for position the recipients it is done in normal letter test images (without noise) and test images letter with the noise on the recipients. On testing Learning Vector Quantization the best value of learning rate is sought in recognition of the recipientís name letter. The research looked at the accuracy of the methods used in a search recognition of address or recipients. From the results of experiments performed, the method of Template Matching in the application of position search of recipients had 100% accuracy in normal letters or without noise. While on the noise the line had a 100% accuracy, noise with spot 78,26% and wet noise 91,30%. And methods of Learning Vector Quantization in the recognition of the recipients had an accuracy of 81,29%.
Keyword : Learning Vector Quantization, Official Letter, Template Matching.