Hello,
I have integrated the Tesseract OCR capability in an Augmented Reality software application (Unity3D|C#) where an Android tablet recognizes a product's serial-number upon taking a picture of it.
Now, I would like to "automate" this process, so that the user would not have to zoom the device's cam on the serial-number and press a button to take a photo, etc... but instead just hover the tablet on the product until Tesseract has recognized the serial-number's text.
In other words, I would like Tesseract to analyze the incoming video stream and perform recognition constantly on all frames until the text is recognized.
Of course, initially I had no idea if this would be possible, so as a first solution, I wrote a program to emulate the "photo-button-press" + "text-string-recognition" processes and repeat then every 1 second.
The performance is OK and the whole thing gets the job done, but I still wonder if Tesseract might be able to work on a live stream of video coming in through a tablet's cam and perform OCR on the incoming frames...
Joshua