I'm guessing you mean something like 'how are surveillance systems adapting to the wide use of face masks in public'. *

There are a few techniques including

1) Gait analysis [1]

2) Ear shape analysis [2]

3) Running image recognition on something the person is wearing rather than on their masked face. I can't find it right now but I recall reading a Google research paper a few years back that managed a high (+87% certainty) just based on a person’s shoes. [Edit, after a bit of ‘Ducking’ came across [3] which is a 2016 patent awarded to Disney which is close enough]

Think about how often you see someone wearing the exact same shoe as you are. Same goes for things like backpacks, caps with logo on etc.

At some point 'Bob' or 'Alice' is likely to be carrying a cell-phone (sadly an almost certainty these days) or have to use public transport or leave their house or walk in to a shop and buy a drink on the way. There is a push in a lot of parts of the world to pay by card only and some shops are starting to refuse to accept cash and the same goes for public transport.

With sufficient coverage once you have managed to ‘lock on’ to a unique characteristic (shoe, gait, distinctive item of clothing) even if you don’t manage to get lucky that time you just let the image recognition trawl through all the archived footage and at some point chances are good that you can identify them well enough to know which door to knock on. Given that people are creatures of habit (daily commute, same route, same time of day) even if unable to track back to a physical location you just wait where they tend to be at certain times. The hunter only has to get lucky once, the rabbit has to be lucky each time they pop their heads out.

[1] https://www.scientificamerican.com/article/motion-capture-su... & https://apnews.com/bf75dd1c26c947b7826d270a16e2658a

[2] https://www.sciencedirect.com/science/article/abs/pii/S09574...

[3] https://thenextweb.com/insider/2016/07/29/foot-fetish/

* Where possible I have used links that are a few years old. Feel free to Google some of the terms used in the research papers for more up to date information.