It's quite a fun thought experiment to think about how you might avoid this bias.
Send Blackhawk helicopters to collect samples by force? Was there bias in the selection technique? What about the people genetically predisposed to avoid the Blackhawk helicopters? We'd need multiple collection methods, and still need to retain an unbiased distribution. You can continue until you get to "we should set up global search patterns so that no human can escape our dystopian DNA collection". And even then you're still biased by population effects owing to the probability that there is a genetic population who have been predisposed to suppress their own genetic population at this point in time.
Even then, you'd bias your sample toward people who use those trash cans (people who live near the trash cans you selected, people who don't litter, people who spend a lot of time in public if they are public trash cans, etc.)
Bias in surveys is an interesting and hard problem to think about, but not really a new one. Seems to me like the typical demographic control approaches would work here too.
[1]: https://en.wikipedia.org/wiki/UK_Biobank