Image processing - count number of blobs

[Anonymous Monk]

I am presented with a number inside an image that I believe is generated by a captcha generator (securimage/phpcaptcha). I don't actually need to read the number represented in the image, but I do need to know when the magnitude changes. So if the current image contains "0.00005", I need to count the digits to find that it has six, so that when I get an image containing "0.09" and see that it contains only 3 digits, I can take the necessary action.

Here is an example image. I first tried using OCR (tesseract/gocr,/ocrad), but the captcha makes that too hard without first doing pre-processing, which might be a harder problem to solve (the text changes color and the background noise dots are randomly colored). So after doing some research it looks like Blob_detection is what I need to do and that it sounds like it might be an easier problem to solve. Except I can't find anybody doing anything similar in Perl. The only thing I can find is contained in much larger frameworks, like OpenCV or Matlab, etc., which would require using a different language and calling system on an external program.

Has something like this been done before in Perl?

[BrowserUk]

Whether that example is a particularly easy one, but it cleans up very easily to the point where OCR might stand a chance, without doing anything fancy.

Of course, if this is legit, you could point us at the source of the image and we could verify that what you are asking us to help you with isn't breaking any T&Cs.

We could also obtain some more examples for testing our approaches to the problem.

[Anonymous Monk]

Of course, if this is legit, you could point us at the source of the image and we could verify that what you are asking us to help you with isn't breaking any T&Cs.

Of course this is against the TOS, just like any form of web scraping. The source site thinks that only humans should have any use of the data so make it hard to access, and I want to automate a process that involves the magnitude of the data.

While everybody here was being so cynical and snooty, the Python community welcomed me and helped me convert the script to Python so I could easily interface with OpenCV. And another soul sees the light- not a better language per se, but a better community for sure.

[Anonymous Monk]

Of course this is against the TOS, just like any form of web scraping. The source site thinks that only humans should have any use of the data so make it hard to access, and I want to automate a process that involves the magnitude of the data.

While everybody here was being so cynical and snooty, the Python community welcomed me and helped me convert the script to Python so I could easily interface with OpenCV. And another soul sees the light- not a better language per se, but a better community for sure.

Got link?

[SuicideJunkie]

If this really has nothing to do with solving captchas, I suggest asking the source to provide better images that don't look like captchas.

If they expect you to comprehend the image using software, then they should provide an image that is not intentionally difficult to understand using software.

[Anonymous Monk]

which would require using a different language and calling system on an external program.

Why do you think that ? Why is that a problem?

[Anonymous Monk]

keep walking spammer :)

[Anonymous Monk]

keep walking spammer :)

Hey twit, if you read the OP I said I didn't need to actually read the number, which means this has nothing to do with solving captchas.

[Anonymous Monk]

Hey twit, if you read the OP I said I didn't need to actually read the number, which means this has nothing to do with solving captchas.

Keep walking spammer