OCR of license plates

[Gunsuka]

I wanted to build something that could open a photo and extract the text, OCR seems like the perfect task.

The photos can come from various sources, but are photos of license plates.  I'm attaching a couple of examples.

So far I can't get Tesseract to identify anything, not even a single character.

I know Tesseract is working as if I feed it sample files, or snippets of website text that will work but this type of image, zero, zip, nadda.

Is there any hope for this type of image?

[Ananth Subramanian]

Hi Gunsuka,

                        These images may need some morphological operations on them before you can OCR using tesseract. However, you may use Amazon Rekognition to get faster and better results. In case you are fine, I will provide you the code. Meanwhile I ran through these examples and was able to get 90- 95% accurate results.

Regards,

Ananth