Number Recognition of Sudoku Grid Image with Artificial Neural Networks

Abstract

In this study it is aimed to capture a SUDOKU grid image, to process this image, to recognize the numbers in the grid image with Artificial Neural Networks and finally to constitute a 9 x 9 number array with these numbers. The reason of choosing SUDOKU game as the input material is the thought of SUDOKU game as a prototype of real world fitting problems. After this number recognition is completed successfully, a robot software who finds the right solution of a SUDOKU game automatically will be developed. The next aim of this robot software is solving real world fitting problems.