The commissioner of the Independent Police Complaints Commission (IPCC) has said there are severe difficulties in running automatic numberplate recognition systems. Nicholas Long said that ANPR systems are often overwhelmed with information and cannot be monitored properly. He was speaking following an investigation by the IPCC …
I remember helping with an ANPR trial North Wales Police were doing in the mid-1990s with a portable camera set up on a bridge over the A55. It had no live link to the PNC, only cached local data for the trials. Within the hour or so that the trial took, if they had responded to every hit the system registered they would have tied up the entire force's response capability and that of the two neighbouring forces.
You don't say
They forget to mention the innocent people stopped and harrassed by the Police because their database is out of date with insurance and other databases, or the 16 year old girl who was killed by a policeman (convicted) who had a false hit on his system and decided 70+ mph was a good speed to approach a pedestrian crossing at.
"Hello sir, our system says that you don't have insurance, can you tell me who you're insured with please?" <phone call> "OK, that all seems to be in order, sorry for wasting your time".
I'm happy to be inconvenienced occasionally in return for fewer scumbags driving without insurance, and I'm pretty sure that if you're co-operative with the friendly officer, he'll return the favour, and everything will go smoothly. If he doesn't, you have grounds for complaint.
I'm not really sure what relevance ANPR has on the other case you quote, unless you're suggesting that the policeman wouldn't have been driving recklessly had the information come from a different source?
Mountains of data by own admission largely useless.
So, what point is there in retaining the data? While disk is cheap, it isn't free. And properly caring for the data is too hard; it's already too hard to make proper use of it.
Since the police do their thing ostensibly for the public good and in the public trust, then this is one more reason to stop gathering large amounts of data willy-nilly, and moreover to stop with it right now.
What the police must do is choose very carefully what to pursue, and have their systems geared to picking out the tidbits they need so they can then act on the information right away. And they must then also act on it.
If not, throw the data away (safely, securely) and only keep records of how much of the data gathered was useful. If it turns out the usefulness doesn't outweigh the costs and trouble of having the systems around, well, tear'em down and think of some other tool to use, if any.
The system really should have been built with that in mind from the start. But it was not. To me, that is more proof that it was "because we can" technology, banking on "inevitable" future but as of yet unproven usefulness, that has now been shown less than unequivocally useful. When is the police going to pack away their toys and go back to policing?
What is the point?
Well call me a cynic but I suspect the data will be kept and when processing power and automation is improved and Mr Plod can afford to hire a few fresh post grads, we will have a rash of retrospective investigations and or prosecutions based on things we can no longer disprove, but they have "evidence" of.
This highlights the difference between a sensible policy of collecting data for a particular purpose and not using it for anything else, and what we have now. When you have used it for what you started out wanting it for, get rid of it. Don't just keep it so you can go on a fishing expedition in the future.
The point is that it's abuse of data, that therefore has to stop.
The ANPR system needs no storage. All it needs is to know what to look for, look for that, and raise alerts if and when it sees what it's looking for. The basic storage is to be in case files and only contain that which is useful and legal to be kept there. The fact that there is a big database behind the ANPR now means that we have to hit the government over the head with a large and pointy cluestick until this changes.
Gosh .. at those trivial data rates you could very easily have even a simple Perl script with a huge regexp that tail'd the feed and fired off an SMS or Tweet to the relevant person whenever a car of interest was spotted.
The easiest and cheapest way to use ANPR is to register only vehicles you have on a hit list, doing a forensic search requires you to store the hit data for a period of time, but how long, there are are 35 Million vehicles in the UK, and we pass around 10 ANPR cameras a day, so thats 350 Million ANPR records per day, give or take, probably 200 bytes per record, more if it includes a picture, say 350K, and suddenly you start to need a lot of storage.
Ignoring an image, why 200 bytes? Surely, Camera location, <=4 bytes (4 billion locations should be sufficient across the UK), number plate <=7 for UK plates (let's be generous and allow 10), and a timestamp, again let's be generous and allow 8 bytes (UNIX per-second timestamps can still be represented by 4 bytes until 2036).
Total, 22 bytes per record, maybe with some overhead.
Multiply by your 350 million records a day (I'm not sure if this is actually accurate, as not every vehicle is driven every day, and I want to be shown the 10 ANPR cameras a day that I pass), and you still only come up with around 10GB a day. Deem that data useless after 6 months, and you're talking about 2TB, or one current generation high end Sata disk in total.
The recorded image is only necessary if you intend to use record for enforcement (for say, driving without road tax, MOT, or insurance), but probably not for locating people by spotting their cars. If you had automatic discard of images of vehicles that were legal and not on a hit-list, then you would not need to keep all of the images anyway.
Bearing in mind that many more times that volume of data is transmitted and stored daily in any number of data applications, it does not seem too far fetched after all.
BTW. I wish to state that I do not condone this information being kept and used for tracking purposes at all. I don't mind to get illegal cars/drivers off the road, but that's it as far as I am concerned.
Too much information, my arse
The only thing this serves to demonstrate is Forces' (and NPIA's) complete bloody inability to identify how their business works and translate it into accurate representative use cases. This isn't about too much information, it's about the fact that none of the ANPR systems were capable of notifying external systems/users of hits against a categorised "hit list" of plates. If your ANPR system is just sitting doing passive monitoring, every time you go through a dodgy area, of course it'll generate enough crap to keep you busy for months.
I've been working in development of Police IT systems for nearly twenty years, and I've yet to see a requirements spec for a system which comes anywhere close to actually defining what they need. I'm not surprised to see senior staff blaming the system, as usual.
"I've been working in development of Police IT systems for nearly twenty years, and I've yet to see a requirements spec for a system which comes anywhere close to actually defining what they need. "
Have you ever seen a cost/benefit analysis explaining *why* it was needed in the first place?
Or was this one of those "Oh look at the super-duper new tech. We *must* have it" sort of buying decisions?
It's *very* curious to outsiders that while the 43 forces make much of their independence they seem very coordinated on some things but go in near random directions on others.
Not a flame but actual puzzlement given that *all* these systems ultimately send their information to the same place and presumably update PNC2 as part of the process.
@John Smith 19
Good point about the cost/benefit analysis - no, but that's probably because those aren't generally visible from the supplier's side. But the tender process for most of these type of systems are typically bragged about for ages in the press by either the home office, NPIA or the forces, so you can generally put two and two together. To be fair, most of them aren't "must have new tech" decisions, because the govt wouldn't sign them off. Having said that, there does seem to be a degree of oneupmanship between forces, so an element of that creeps in later.
The real problem with forces going in different directions is that the IT body "representing them" (was PITO, now NPIA, next God-knows-who) can recommend, but ultimately has no teeth. Every time someone gets close to giving them the ability to mandate purchasing, a bunch of Chief Constables kick back saying that their force needs something different because of the way they work, yada yada yada.
Since Soham->Bichard, the appearance of the PND gives NPIA a carrot to use on the forces instead of a stick. Since the forces *have* to work with PND, they start moving in the direction of common data formats, at least for centralised intelligence. However, for all other systems (Custody, Case Prep, Command & Control etc) the bets are still off.
Re: your last paragraph - no, only a small subset of the forces' info is sent to PNC/PND, and that's only at the will of the local coppers. Everyone still has their own datasets and they're all fiercely protective.
The only things that will change the way the police buy their IT are a real effort by government to align the way forces do their day-to-day work, and some serious investment (both money and time) in the tendering process (instead of just handing it to some bored fifty-something coppers on the way out to retirement).
Its madness, to me, that each force uses this data feed so differently and there's no centralised reporting. Designing a system to respond in real-time shouldn't be that hard. 1) forces or investigating officers register vehicles of interest with PNC or other central system, 2) stream all ANPR hits (just the plate, time, location) to central system, 3) send prioritised alerts to the relevant and interested forces or officers. Basic "publish-subscribe" pattern.
If banks can do their wizz-bang algorithmic trading sub-second off live feeds, then surely this system could do it in sub-minute timeframe?
No rocket science, at all. How such a useful crime-fighting system can be so ballsed up is amazing, but then, its government IT.
Why isn't there a centralised IT strategy for this. Madness. Again, government IT; why do something right once, when we can line the pockets of consultants time and time again.
A victim of its own success ?
It sounds more like a victim of its own complete failure to provide a meaningful, working system.
Which will always happen when the number one requirement on the system spec is "Collect as much data as possible," rather than "Collect only meaningful data that may be of help to an investigation".
Technology For Technology's Sake
I can see the benefit of ANPR in police vehicles, catching people with no insurance etc. Having static cameras which report back to a (possibly unmanned) control room seem like a waste of money to me. You just need one of the controllers to retire to the bog with yesterday's Sun for a rest break, and that axe-murderer's left the area...
Signal to noise ratio
I've frequently said before, rather than *fighting* the database state, which will only encourage politicians to dig their heels in, we should *embrace* it. The more the merrier. Bring it on. Because as the state gathers more and more information, and finds new ways to make connections between it (bearing in mind we are looking at an exponential increase) they will eventually sink under the weight of data.
Take a plan to fingerprint everyone, for example. Currently fingerprints work, because the sample held is small, in comparison to the general population. Therefore any match is more likely to be what you are looking for. Now expand the sample to be a much larger subset of the general population. Running a search could throw up 2,3 or more matches. Add to that the fact that people move, die, leave the country, and just give false details, and you *will* end up with a court case where the defence get the prosecution to admit they had *another* match, which they couldn't eliminate.
Reasonable doubt ?
Same for their databases of text messages, websites visited, etc etc ... bring it all on. You'll go mad first !!!
It's not about the all, it's about the individual.
If they store everything, they'll be able to go through old data to find things to use against selected individuals.
That's the real danger of these databases.
You already see it in other places - how many people have had embarrassing details of their life from many years before dragged up to their detriment?
Do you want to allow the state that power over everyone?
Ok, maybe you trust the current lot not to abuse it. But do you trust the next lot? or the ones after that? What if *insert diabolical party here* get into power?
In law, Parliament cannot bind future Parliaments to anything - the only way to ensure future Governments do not abuse such data is for the data not to exist.
One (joined up) system to rule them all?
O.K., not really one system, for everything, but for god's sake, the clue is in the name, PNC, Police NATIONAL Computer.
In reality, it's not, it is a series of disjointed, fragmented and dated computer systems. You then have a series of disjointed, fragmented, and budgetarially challenged forces having to run their own systems to feed into their own area.
Several of these failures were down to people process as much as system processes.
I would advocate ONE Police National Computer Database, that would be fed by ONE Automatic Numberplate Recognition system that fed into it's own database.
The ONE ANPR system would be manned 24 7 and could be used by both Police and Security forces with the correct authorisation to place a specific vehicle (or individual) on a Watch list.
Tie this in with the One driver licence and vehicle licensing body perhaps?
If we want to make some savings as a country, stop spending hundreds of large lumps of cash around the country supporting individual government departments, and spend two or three VERY large amounts of cash in one place.
Design it right, outsource the infrastructure, (or perhaps several massive government owned data centres across the UK, I'm sure there are loads of cold war nuclear bunkers not being used that would be great for data centre activities)
We, as a country need to ensure that the money the government is spending (our money) is spend wisely and efficiently. and that is large and centralised with mandatory use as opposed to siloed and pocketed repeatedly used across the UK (Think all the local governments, all the different departments that do the same thing, the same way on their own systems)
@Piri Piri Chicken
"In reality, it's not, it is a series of disjointed, fragmented and dated computer systems. You then have a series of disjointed, fragmented, and budgetarially challenged forces having to run their own systems to feed into their own area"
The PNC *is* one system - albeit a horrific mainframe-y ADABAS natural DB. That's the bit that was *supposed* to centralise police intelligence, but forces don't have to put everything on it. Each force has its own guidelines as to whether something ends up on PNC or not. However, you're right about all the other systems being disjointed and fragmented.
The much-touted PND won't help with this, as long as nobody is mandating what forces have to upload onto it.
@Piri Piri Chicken
You fail to ask the *key* question which will save the *most* amount of money and time.
Does the UK *need* a centralised ever expanding warehouse of license plate data with *no* oversight?
BTW I'm fairly sure your post will be either one of those 1/2 up, 1/2 down voted posts or unvoted, as to agree with one side of it would imply you agreed with the other.
Hint: You've described a sensible and efficient way to implement a centralised system (very reasonable) for an effective and highly authoritarian vehicle (and by extension person) tracking system. Not very reasonable.
Would you care to clarify your view?
One plate per car would help.
I wonder if the DVLA have eliminated the multiple plate registrations loopholes in thier systems/database?
Busy roads - Really?
Wow! just imagine... The police are amazed at the amount of traffic there is out there on our roads.
Who'd have imagined that there was so much that it would clog up their data systems? Just like it does to the roads.
I wonder - Is the high traffic level an equal surprise to to government? Obviously not if they are spending the cash that all that traffic generates and looking at all the revenue they're missing, which is the real point of all this interest.
Re: Signal to noise ratio
AC wrote: "Currently fingerprints work, because the sample held is small, in comparison to the general population. Therefore any match is more likely to be what you are looking for. Now expand the sample to be a much larger subset of the general population. Running a search could throw up 2,3 or more matches. "
2 or 3 hundred matches more like, and you can't eliminate most of the suspects like you can with the criminal fraternity - many either already behind bars, dead, wrong MO, etc, etc. If the police had the resources to follow up every match, you'd be seeing hell of a lot more of your local nick, and you'd better start keeping a diary so you can remember what you were doing on the night of the 17th six months ago.
Simple reasons for taking/keeping them
1: - to see if the image of the vehicle matches the registered details
2: - to see if it's only one vehicle using 'em. (london has a lot of vehicles using the same fake plate number to vex congestion charges)
It'd be nice if they could read the tax disc barcode but that's wishful thinking with today's optics.
It'd be even nicer if a few patrols simply checked all parked cars and lifted any with fake plates - with uninsured rates running at over 11% in London this is a prime reason why premiums are spiking rapidly
Hell, I'm surprised the unsewerants industry doesn't pay for that kind of action.