Neural Network Based Skeleton Recognition and Sudoku Solving

摘要

This paper presents the development of an Neural Network Based Skeleton Recognition and Sudoku Solving. The main objective of this work is to recognize the number and its corresponding position from a Sudoku image and also to solve any valid Sudoku. The recognition system is designed through an artificial neural network model. The neural network uses the mechanism of feedforwardbackpropogation technique where minimizing error is taken into consideration. Using the gradient of the criteria-field helps in weights modification and thus optimizes the system. The Sudoku is then solved using the backtracking algorithm which is a trial and error method. It takes into account one selection at a time from the multiple choices (1-9). This technique can solve any valid Sudoku. The final result is made to display on the original image by using the database. The database consist of template numbered images which is obtained from the segmentation from the Sudoku images. The system is well trained and effective in recognizing the number compared with the traditional template matching.