Mobile OCR using centroid to boundary and Backpropagation Neural Network
DIGIT RECOGNITION SYSTEM USING BACK PROPAGATION NEURAL NETWORK Devinder Singh1 and Baljit Singh Khehra2 1Department of Computer Science, Mata Gujri College, Fatehgarh Sahib, INDIA, E-mail: firstname.lastname@example.org 2Department of Computer Science and Engineering, B.B.S.B. Engineering College, Fatehgarh Sahib, INDIA, E-mail: email@example.com Digits are used in different types of data like vehicle number plate, numeric data, written data, meter reading etc. Digit recognition plays an important role in many user authentication applications in the modern world. In the proposed work, back propagation neural network based digit recognition system has been developed. Such system has four phases: preprocessing, segmentation, feature extraction and classification. When digit image is scanned, quality of image is degraded and some noise is added into this image. So it is necessary to reduce the noise and improve the quality of the digit image for OCR system. For this purpose, image enhancement is necessary. For this, frequency domain based Gausian filter is used to improve the quality and denoise the digit image. The goal of image segmentation is to separate the clear digit print area from the non-digit area. After segmentation, binary digit image is skeleton to reduce the width of digit into just a single line. In proposed system, coordinates bounding box, area, centroid, eccentricity, equiv-diameter features are used by back propagation neural network as a classifier to recognize digits. Experiments have been performed on standard dataset of digits. Results of this study are quite promising and 96.6% accuracy has been achieved by the proposed system.