Optical Character Recognition (OCR) dengan Menggunakan Neural Network dan Feature Extraction

Abstract

Optical character recognition (OCR) is used to translate the character (character images) into a text format. This approach has many criteria for the introduction of the character image pattern detection, one approach taken is to use feature extraction method by using a method of supporting, supporting one of the methods used is the method of neural network, with the purpose of introduction of running with maximum filtering. Feature extraction is one way to recognize an object by looking at specific traits that mapped patterns based on the characteristics possessed by an image of the character. The use of image processing filtering of total number of 10000 Epoch utilize backpropagation network (BPN) to save the iterative process of learning. Target pixel that has been divided into 9 blocks converted into the target pixel. The training process is done just once for the type and size of the same letter that starts with arial font type, size 8, size 12 pixels. Image of character letters A.bmp learning has 99.75%. learning the numbers show the image of the character image 1.bmp 100%.