Acceptance of crossed out handwritten words

I am working on handwriting recognition and related things on the visual studio platform and using openCV libraries. Log in binary scanned .tif images.

Currently, I entered the checkpoint, trying to find a way to recognize crossed out words, how to cross out (cancel) words in you using a straight / curved line. I am not going to do individual character recognition, because it will be a waste of computing power.

Is there a way to recognize such occurrences in an alternative way?

Below are two ideas that I came across, but I'm not sure - 1> use a mask like <0 0 0, 1 1 1, 0 0 0>, which will help find all the horizontal lines ... but this will be a very big guess. lines can be wavy in any orientation.

2> skeletonize the entrance and find the intersections. this will give me quite a few intersections - including those that are due to the line used to cross out the word. using some approximation, such as least squares, etc. I can get an approximate string. but there is a problem that intersections can occur in many places - for example. 2 intersections at 'b', etc.

any suggestions?

+3
source share
3 answers

Hough transform ?

hough , : handwrittingdetected lines

openCV. cvHoughLines2.

+2

? Poly (Ten-Chin) , . , /, , .

, "" , (. - Hu Moments), , , .

ML (Machine Learning) OpenCV, . + , , , - (SOM, KNN, KMeans).

0

. , . , , , , . , char , -, , , . , , , , .

, .

0
source

Source: https://habr.com/ru/post/1719079/


All Articles