I have questions about exactly where this data comes from. Does it include every single camera in the city, eg. security in shops or apartment buildings? Surely they're not linked into the government.
Or are they all supposed to be some kind of "smart" cameras that detect and isolate faces, vehicles etc and then send only some kind of abstracted data back to the mothership? That sounds like the only manageable methodology. It also sounds unbelievably expensive.
This article raises more questions than answers.
>Or are they all supposed to be some kind of "smart" cameras that detect and isolate faces, vehicles etc
The cameras are part of the "Skynet" (天网) system, which is intended to perform all of those things and a great deal more. It's powered by software from "the world's most valuable AI startup", SenseTime.
chongqing planed to have 500k camers in 2011. The auther believes that number grows 20% YOY. So it's 2.6 million now.
I have no words.
> A primary argument in favor of CCTV surveillance is improved law enforcement and crime prevention. We compared the number of public CCTV cameras with the crime and safety indices reported by Numbeo, which are based on surveys of that site’s visitors.
> For both indices, the correlation was weak (r = 0.168, r = -0.168, n = 120). A higher number of cameras just barely correlates with a higher safety index and lower crime index.
That's also before you consider the trade off, both in privacy and in money. How much does it cost to put up all those cameras, and how far would that money go if applied to traditional law enforcement methods? Or just not spent because the cost still exceeds the benefit even if the benefit isn't statistically indistinguishable from zero?
That makes an impressive case for cameras. I would have expected a negative correlation. People put up more cameras where there are more crimes.
There is no reason to expect that correlation to exist. People put up cameras where they can afford to buy cameras and have a local legislature that can be influenced by contractors who want to sell cameras, but having money is inversely correlated with high crime rates.
There are more cameras in NYC than Detroit.
If a criminal chooses to not commit a crime (or commits it elsewhere) because the risk/reward balance is changed, then it prevented the crime from happening (there), and made that location safer.
So there is no doubt that the cameras make places safer. Even fake cameras can make a location safer. The question is whether the cost to personal integrity makes it worth it.
Most importantly: who answers that question, and who ensures the cameras aren't used for bad purposes? That's probably a smaller problem in London than in Beijing.
Why not make cameras that output only encrypted video which may be stored by agencies and decrypted only when people holding the right keys come together during a trial or active police investigation? That would be strictly better for privacy than what we have now - which is total surveillance.
The companies are now selling to private communities. 
What about conviction rates? For example, a lot of shootings in Chicago go unsolved. Maybe for lack of a witness?
A couple of police chiefs ago the chief said people unwilling to talk was the number one reason crimes go unsolved in Chicago.
He didn't get a chance. He was fired for suggesting that people don't trust CPD because of over a century of corruption.
EDIT: They do credit the source to numbeo.com, which doesn’t rely on government stats (https://www.numbeo.com/crime/).
Given the answers to these question I am not sure helpful this analysis is.
Relative to Singapore about the same, assuming they used government-provided stats.
Edit: Estimates always skew reality!
Semirelevant digression: A few years ago someone found that in Munich, the shop façade cameras covered every meter of the main pedestrian street (Hbf to Marienplatz). Not planned by anyone, it was just that each shop added a camera to watch over its own expensive display window, and there are shops everywhere there. Now the government is proposing a new law giving the police real-time access.
There are 13,000 cameras  for 2.7 million people .
I have a security camera on my house. I don't know how that would be captured in the statistics at all.
> We focused primarily on public CCTV—cameras used by government entities such as law enforcement
> we have also tried to find the number of private CCTV cameras in use. However, as this isn’t possible for every city, we did not include private CCTV in the overall totals.
Anecdote: A manager of a Fortune 50 location I was working at received a call from HQ to keep an eye on a known perp who entered the store no more than 5 minutes prior to the call. Very impressive & very scary.
Human memory is both unreliable anyway and motivated in operation. CCTV acts as an impartial witness. It doesn't automatically have all the facts, but it is reliable and unmotivated. Knowing this impartial witness has their backs enables your employees to obey policy without worrying about how that will be remembered.
The operators are supposed to do some paperwork establishing they've thought about how they're going to treat all this data, and if they want to put CCTV somewhere you'd normally expect to be private (e.g. a changing room or toilet) they need to tell the cops what they're doing and why (e.g. to investigate a specific employee suspected of theft) before they do it otherwise that's a crime.