Application of Template Matching for Optical Character Recognition
Authors:
- 1 January 2019
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 2213-2217
Abstract
Optical character recognition (OCR) is one of the common research problems in the computer vision which is used for industrial processes automation. In this work the algorithm of OCR based on template matching recognition is presented. Algorithm is adopted for character recognition with consideration of low contrast conditions and texture on the background. The main feature of this method is interest area detection and adaptive binarization before pattern matching. The proposed method does not require large set of training samples, which may be critical in some applications. Also algorithm is easy to implement and has low complexity. Testing was conducted on custom dataset under the various conditions of illumination and pattern readability. The quality of the algorithm was estimated using standard binary classifier metrics and the distribution of correlation function in comparison with the template.Keywords
This publication has 5 references indexed in Scilit:
- Optimizing Expected Intersection-Over-Union with Candidate-Constrained CRFs2015
- The relationship between Recall and PrecisionJournal of the American Society for Information Science, 1994
- A Computational Approach to Edge DetectionIEEE Transactions on Pattern Analysis and Machine Intelligence, 1986
- A fast two-dimensional median filtering algorithmIEEE Transactions on Acoustics, Speech, and Signal Processing, 1979
- A Threshold Selection Method from Gray-Level HistogramsIEEE Transactions on Systems, Man, and Cybernetics, 1979