Welcome to Atheist Discussion, a new community created by former members of The Thinking Atheist forum.

Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
License Plate Identifiers
#1

License Plate Identifiers
Just saw an interesting series "Unbelievable", a police procedural concerning a serial rapist based on a real case that ranges across numerous themes that is superbly done; I recommend it.  In it was a scene where a license plate was run through enhancement software to discern its ID, and the outcome was a failure due poor source resolution and extreme oblique angle.

It struck me watching it that the approach was backward:  attempting to sharpen a blurry image into readability, instead of generating all license plate combinations and blurring them to find which blurred images match the blurred original.  This latter approach seems feasible in that license plate graphology conforms to an exact specification and the number of possible combinations only in the millions - present computational speed should be able to zoom through the comparisons in hours, maybe even minutes.

Even if the outcome produced a thousand or so blurred candidates that field would be narrowed significantly by including the make/model and color of automobile.  The final output could be a dozen or so license plate/vehicle combos to put on an APB.

Anyone familiar with this technology know if this sort of approach is being developed or in operation?
Reply
#2

License Plate Identifiers
You'd have to deal with its admissibility as evidence in a court.  Which means convincing a judge and then a myriad of appeals courts.

There have been repeated stories coming out that all this supposedly hot shot forensic evidence is not quite as reliable as it is claimed to be.
Robert G. Ingersoll : “No man with a sense of humor ever founded a religion.”
Reply
#3

License Plate Identifiers
Just off the top of my head, there's no deterministic way to generate a specific blurred image from a non-blurred image, much less one that matches the characteristics of the device that originally took the blurred image, so, I rather doubt what you suggest is feasible.
Mountain-high though the difficulties appear, terrible and gloomy though all things seem, they are but Mâyâ.
Fear not — it is banished. Crush it, and it vanishes. Stamp upon it, and it dies.


Vivekananda
The following 1 user Likes Dānu's post:
  • Assassin
Reply
#4

License Plate Identifiers
(09-18-2019, 08:04 PM)Minimalist Wrote: You'd have to deal with its admissibility as evidence in a court.  Which means convincing a judge and then a myriad of appeals courts.

There have been repeated stories coming out that all this supposedly hot shot forensic evidence is not quite as reliable as it is claimed to be.

It wouldn't necessarily need to be admissible in court, only strong enough for "probable cause" allowing further investigation by police.
“Religion is excellent stuff for keeping common people quiet. 
Religion is what keeps the poor from murdering the rich.”
― Napoleon Bonaparte
The following 2 users Like Chas's post:
  • skyking, RobbyPants
Reply
#5

License Plate Identifiers
The police, at least here, already use enhanced photos of suspects at or near crime scenes to try and identify them. I've seen blurry photos shown on the news that explicitly say they were enhanced in an effort to help identify the person in the photo. So I wouldn't be surprised if they also try to enhance blurry license plates considering the police already surveil American drivers' license plates on a widespread basis.

-Teresa
There is in the universe only one true divide, one real binary, life and death. Either you are living or you are not. Everything else is molten, malleable.

-Susan Faludi, In the Darkroom
Reply
#6

License Plate Identifiers
You still have to deal with a Probable Cause hearing, Chas.  That's an easy one for the defense.

Quote:Generally, a probable cause hearing happens together with the defendant’s first court appearance after their arrest. The judge will determine whether probable cause supported the arrest. If it did not, law enforcement will not be able to continue holding the defendant in custody if they have not been released on bail or on their own recognizance.

Sooner or later the courts would have to rule on the the procedure.  Polygraphs, as an example, never did make the grade.  And there are turning out to be problems with ballistics, fingerprints, fiber evidence.
Robert G. Ingersoll : “No man with a sense of humor ever founded a religion.”
Reply
#7

License Plate Identifiers
(09-19-2019, 12:30 AM)Minimalist Wrote: You still have to deal with a Probable Cause hearing, Chas.  That's an easy one for the defense.

Quote:Generally, a probable cause hearing happens together with the defendant’s first court appearance after their arrest. The judge will determine whether probable cause supported the arrest. If it did not, law enforcement will not be able to continue holding the defendant in custody if they have not been released on bail or on their own recognizance.

Sooner or later the courts would have to rule on the the procedure...

That's assuming a particular case ever goes to trial. Criminal defendants take plea bargains in over 90% of cases in the US and never appear before a trier of fact. But I'm getting way off topic now, sorry, airportkid.

Would a person be arrested and charged solely on a blurry license plate photo? I'm not sure. It seems like weak evidence on its own and there would be more to investigating and prosecuting a crime than tracing blurry photos.

-Teresa
There is in the universe only one true divide, one real binary, life and death. Either you are living or you are not. Everything else is molten, malleable.

-Susan Faludi, In the Darkroom
Reply
#8

License Plate Identifiers
This is actually a fairly complex legal process, T, and what the OP seemed to be suggesting was obtaining an arrest warrant on the basis of this kind of evidence.

Under the law:

Quote:When an Officer May Make an Arrest
There are only a very limited number of circumstances in which an officer may make an arrest:
  • The officer personally observed a crime;
  • The officer has probable cause to believe that person arrested committed a crime;
  • The officer has an arrest warrant issued by a judge.
An officer cannot arrest someone just because she feels like it or has a vague hunch that someone might be a criminal. Police officers have to be able to justify their arrest usually by showing some tangible evidence that led them to probable cause.

So as noted above:

The term "probable cause" is noted in the 4th Amendment to the Constitution and the citation goes on to say:

Quote:As seen in those words, in order for a court to issue a warrant -- for someone's arrest, or to search or seize property -- there must be probable cause. In situations where police are allowed to effect an arrest, search or seizure without a warrant, they also must have probable cause and it's required for prosecutors to charge a defendant with a crime as well.

The second clause would not apply because obviously someone has to sit down with a computer and play with the images in order to try to figure out if the suspect is involved.  It isn't just some cop seeing a guy with a broken tail light and deciding "I'm going to hassle this guy."
 
So before a cop could even get a warrant he would have to convince a judge that his affidavit is based on solid facts which would probably require the testimony of experts in the case of a new technology.  The OP was hypothesizing about a potential method which does not now exist and has never been certified as to its accuracy.

Does that help.
Robert G. Ingersoll : “No man with a sense of humor ever founded a religion.”
The following 1 user Likes Minimalist's post:
  • Tres Leches
Reply
#9

License Plate Identifiers
I do know that there is software used by the astrophysical community to clean up images of stars. But that relies on arrays of telescopes placed at specific locations to reduce data aliasing, and have a priori knowledge that the item under "surveillance" is actually pretty much round. License plates? That's going to be tough. One thing to know- offenders are usually known to the local police. That fact means that in a given neighborhood, even a blurred picture of a particular vehicle will likely give law enforcement an idea of who the perp actually is.
If you get to thinking you’re a person of some influence, try ordering somebody else’s dog around.
Reply
#10

License Plate Identifiers
(09-18-2019, 07:35 PM)airportkid Wrote: Just saw an interesting series "Unbelievable", a police procedural concerning a serial rapist based on a real case that ranges across numerous themes that is superbly done; I recommend it.  In it was a scene where a license plate was run through enhancement software to discern its ID, and the outcome was a failure due poor source resolution and extreme oblique angle.

It struck me watching it that the approach was backward:  attempting to sharpen a blurry image into readability, instead of generating all license plate combinations and blurring them to find which blurred images match the blurred original.  This latter approach seems feasible in that license plate graphology conforms to an exact specification and the number of possible combinations only in the millions - present computational speed should be able to zoom through the comparisons in hours, maybe even minutes.

Even if the outcome produced a thousand or so blurred candidates that field would be narrowed significantly by including the make/model and color of automobile.  The final output could be a dozen or so license plate/vehicle combos to put on an APB.

Anyone familiar with this technology know if this sort of approach is being developed or in operation?

I watched that too. It was set in 2008. Image sharpening was far cruder "way back then". You'll note that everyone was using flip phones, not smart phones. Hard to remember sometimes how far we've come in just the last decade.
Reply
#11

License Plate Identifiers
(09-18-2019, 07:35 PM)airportkid Wrote: Just saw an interesting series "Unbelievable", a police procedural concerning a serial rapist based on a real case that ranges across numerous themes that is superbly done; I recommend it.  In it was a scene where a license plate was run through enhancement software to discern its ID, and the outcome was a failure due poor source resolution and extreme oblique angle.

It struck me watching it that the approach was backward:  attempting to sharpen a blurry image into readability, instead of generating all license plate combinations and blurring them to find which blurred images match the blurred original.  This latter approach seems feasible in that license plate graphology conforms to an exact specification and the number of possible combinations only in the millions - present computational speed should be able to zoom through the comparisons in hours, maybe even minutes.

Even if the outcome produced a thousand or so blurred candidates that field would be narrowed significantly by including the make/model and color of automobile.  The final output could be a dozen or so license plate/vehicle combos to put on an APB.

Anyone familiar with this technology know if this sort of approach is being developed or in operation?

I would think this is stuff that an MNIST database and a suitably trained ML algorithm are made for.  I suspect the enhancement software had a different approach.
If it doesn't work, it doesn't matter how fast it doesn't work. ~ ???
Reply
#12

License Plate Identifiers
Number plate recognition fails totally of course if the plates are stolen.  Which is a more than common practice
amongst criminals utilising vehicles.  I was hit with two separate moving violations (later withdrawn) due solely
to erroneous plate recognition.  And a third fine for an alleged parking offence, once again due to an erroneous
digital image from a hand-held scanner (also withdrawn).

From a distance, plus crappy cameras, some plates can look very similar...

10O, 1OO, IOI, IO0, IO1, 10I, IO1, etc.
I'm a creationist;   I believe that man created God.
Reply
#13

License Plate Identifiers
Double post. (AD server not completing handshake, which is an ongoing issue not yet addressed)
I'm a creationist;   I believe that man created God.
Reply
#14

License Plate Identifiers
(09-20-2019, 09:32 AM)SYZ Wrote: Double post.  (AD server not completing handshake, which is an ongoing issue not yet addressed)

I have not had this happen even once (Mac OS 10.14.6 and current release of Chrome). Which tends to make me think it's more likely some kind of client-side or browser or network-related issue. Not that it couldn't possibly be addressed or mitigated on the server side of things.

Are you getting "503 Service Unavailable" or "Fatal alert-- handshake failure"? Or something else? If it's a TLS handshake failure you're talking about then I suppose it could be that the server doesn't support all possible protocols or cipher suites that a client could send. You should let the admin know what browser name/version and OS you're using, that might be relevant.
Reply
#15

License Plate Identifiers
(09-21-2019, 01:22 AM)mordant Wrote:
(09-20-2019, 09:32 AM)SYZ Wrote: Double post... no connection...

I have not had this happen even once...

I'm running Win 8.1 with Firefox v.69.0.1, 64 bit. My broadband speeds are 32 down, 7 up, and 30ms ping
average. (Pinging AD sometimes gives me 300ms or more.)  I've only been on-line here a few hours today,
and I've already got half a dozen "failed connection" errors. For me, a refresh doesn't necessarily work, for
anything up to 10 or 12 times. Extremely irritating.

(I note that in all cases, the handshake doesn't even commence.  
Maybe the initial SYN code's not getting to the AD server??  Why??)

I first reported this particular issue late last year.  AD is the only one of dozens of sites I visit every day, and
it's also the only one that I ever have this issue with.  It's obviously a glitch at the site server's end.
I'm a creationist;   I believe that man created God.
Reply
#16

License Plate Identifiers
(09-21-2019, 10:59 AM)SYZ Wrote:
(09-21-2019, 01:22 AM)mordant Wrote:
(09-20-2019, 09:32 AM)SYZ Wrote: Double post... no connection...

I have not had this happen even once...

I'm running Win 8.1 with Firefox v.69.0.1, 64 bit. My broadband speeds are 32 down, 7 up, and 30ms ping
average. (Pinging AD sometimes gives me 300ms or more.)  I've only been on-line here a few hours today,
and I've already got half a dozen "failed connection" errors. For me, a refresh doesn't necessarily work, for
anything up to 10 or 12 times. Extremely irritating.

(I note that in all cases, the handshake doesn't even commence.  
Maybe the initial SYN code's not getting to the AD server??  Why??)

I first reported this particular issue late last year.  AD is the only one of dozens of sites I visit every day, and
it's also the only one that I ever have this issue with.  It's obviously a glitch at the site server's end.

Huh. Well I'm unusual in that my client pays for a gigabit ethernet connection so maybe that's why I'm not seeing the issue. I am not a sysadmin though, I'm a dev, so I only known enough about stuff like this to be dangerous.

It's at moments like this I miss my son; it'd be an excuse to pick his very talented brain -- HE was the sysadmin in this family. Not only knowledgeable, but intuitive and fast.
The following 1 user Likes mordant's post:
  • SYZ
Reply




Users browsing this thread: 1 Guest(s)