02/14/15

Character Recognition

A system for character recognition was developed.  It is based on a algorithm which is able to carry out an image segmentation process.   This process  produces several objects which are then analyzed in order to separate the background from the characters in a digital image.  

A second process is carried out in order to find special particles called "Principal Points" which permit to carry out the recognition process.       

A video showing the robot AIBO reading Japanese characters by applying the method of "Principal Points" can be seen here.  

Figure 1 shows some segmentation results.

Figure 1.  Segmentation results.

After obtaining the segmented objects, they are analyzed by comparing each object with its surrounding objects.  As a result, characters can be separated from the background.  Figure 2 shows the separation process.  

Figure 2.  Separation of characters from the background.

As can be seen, this method can separate properly the background from characters.  The recognition process can be carried out even if a small amount of light intensity is present.   This method has been tested using both Hidden Markov Models and Artificial Neural Networks.  

Below, you can see the results of a character recognition system after applying this method.

This site was last updated 07/06/14