Now, as a competent, dedicated law enforcement professional who is committed to making this project work for the good of society, you're going to need to test it. Thoroughly.
You need a lot of pictures of people with tattoos, preferably not covered by too much clothing. And you're going to need to double check that the image recognition is correct for every match.
Once you've done that, you'll need to start applying the same algorithm to video sources. Again, with careful checking to verify that the system is working correctly.
Where on the Internet are you going to find a huge volume of images and video featuring people with exposed tattoos? And how, exactly, did you just convince the government that it was your job to spend all day watching them while calling it science?
Probably the best way to roll this out would be to target and identify repeat offenders. If you have someone currently incarcerated, you can take good clear pictures of them revealing their tattoos which could then be matched to images of tattoos from crimes committed after they are released/before they were incarcerated.
MY one concern is this:how good would it be at recognizing altered tattoos or ones that have had additional tattoos added to or applied over the original tattoo? Want to help get away with murder? Every time you kill someone just add a tattoo to your sleeve, incorporating a previous tattoo so that it is altered enough to render this method ineffective. Plus you get the added benefit of having a souvenir/reminder of each criminal act. Think the Johnny-23 character from the movie Con Air.