Ranorex Forum

Hi,

With the help of RANOREX tool, Could we be able to read text from images?
or is it possible to automate CAPTCHA?

Thanks

Hi,

if the text is in the picture, it is not possible to read the text - since it is a picture.
What you can do is an image based validation of your image. Maybe this helps you in any way. See Image-Based Automation.

Kind Regards,
Larissa
Ranorex Support Team

I see this was last asked back in July 2012, and Ranorex is now on 4.0.3

Has there been any advances in testing with Captcha ??

We have a requirement to test our online regristration process, which includes;
1 - Responding to a random Captcha image
2 - Responding to a verification email

Thanks in advance,

The whole point about CAPTCHA is to make sure you are a human and not a computer. It is supposed to be impossible (difficult) for a computer to read the CAPTCHA. Good luck finding any, legitimate, test automation software company to help you with this...

I worked on a project that had CAPTCHA validation for users. In the test environment the developers turned this off so we could use automation tools. The actual login process was only a very small part of the test set and the CAPTCHA validation was later tested manualy

Hello,

Unfortunately, Ranorex doesn't provide a way to get a text out of an image.
Therefore, it's not possible to automate CAPTCHA.

Regards,
Markus (T)

Thanks for the tip Ian. We had our developers turn off the Captcha and Account Verification for our testing scripts.

Larry

Larry

Pleased I could be of help.

Hi,

Not a 100% working solution but it can works if the word isn't deformed, twisted, enlarged... but we use those methods to overide some simple captcha:

method1 (used with different width characters)):
- save image to hard disk and turn it to a B/W to eliminate all noise.
- use Tesseract (google it) program to OCR it and save text to a file
- read the txt file with the scanned text and use it in your automation
accuracy % : <50%


method2 (used with numbers and same width characters):
- save image and make slices of the same size with all characters
- turn every slice to B/N
- every slice must have the same size in to order to make some control pixels
- check if control pixels are black in order to check what number is

for example to check for an '8', you must check for pixels that make 8 different from others numbers (marked as '#')
not the best solution but it works fine with an accuracy of ~95%


regards,
Mateo.