identify letters and differentitate it from from irrelivant lines and spaces

I have a numberplate image input which i need to automatically crop the characters of the numberplate only.

i asked in a previous post on how to do it but no response and the post is not ordered properly. i decided to take a different approach.

i used region props and i looped through the coordinates cropping and displaying, i see that some of them displayed each letter or number individually.

this is where the problem comes in, is there a way of recognizing whether the boundary coordinate is for a letter and not for irrelevant shapes or lines.

This is the approach i took

loop through all the boundarybox values i.e. to length(reg)

if the size of the width and height is greaterthan 110 and 140 than the boundary is for the wholenumberplate

continue loop

if size of height is between 50 to 90 and width is between 14 and 70 than the boundary coordinates is for a letter

add all the boundary values for all the letters in an array

get min value for xmin and ymin from the array for height=max ymin value + height - ymin width= max xmin values + width - xmin

we have the xmin,ymin,width,height to crop

this works for some and for some it does not work

so i want to know, how to differentiate the boundarybox coordinate for a letter or a number from boundingbox coordinate for useless lines.

e.g. boundingbox = [115.5,128.5,50,66] a coordinate for a letters another one = [110,130,40,69] a coordinate for a useless lines and shapes

i want to find a way to differentiate between the both so i don't get an array including coordinates for useless liens and shapes.

if i know this then it will be easier from there.

thanks

Answer 1

[Walter Roberson]

What you are asking for is not generally possible.

On the below, everything in green is significant.