My experience with OCR's is mixed.
The first time, I was trying to use TextBridge to read part numbers from digital images of catalog magazine pages. It was able to capture characters, but most of the time, its guesses were wrong. Worse, validation of the characters wasn't a viable option since there could be hundreds of part numbers per page and very little intuitive language.
The second time was on my last contract and I actually wasn't using the OCR myself. The app I worked on was taking extremely hi-res pictures and capturing OCR's from images of semi-trucks, trailers, etc. In fact, I think they actually assemble the hi-res pic from a mosaic of vertical slices. Either way, they're remarkably successful at the OCR capture and the numbers they're catching have a checksum that lets them know that they have it right.
Overall, if you don't have a way to make sure that the captures are accurate, OCR can be a real pain in the backside. In my first effort, I abandoned trying to capture the part numbers from images...it just wasn't accurate enough, especially with the time it was going to cost me to even get close.
Good luck in what you're doing, whatever you decide. |
