Apparently a teen ran off with her soccer coach. And surveillance videos captured them.
There are probably a bazillion surveillance videos across the nation. How did these get selected?
Thanks as always.
I’ve seen the detectives access surveillance on The First 48. Cameras are one of the first things they look for at crime scenes.
It’s nothing fancy. A standard DVD player with fast forward and time code search.
The detectives even check video from businesses around that area. Hoping to see the car approaching or leaving the crime scene. I remember one case where the suspects gassed up a few blocks from the crime. I think it was right before it occurred. The camera got very good captures of their faces.
It’s tedious work.
Surveillance cameras are getting better. I remember cases from 10 to 12 years ago. The detail in the video was barely useable. It might show three men chasing the victim and shooting him in the street. The detail wouldn’t be sharp enough to make prints of the faces. Frustrating for the detectives. They saw the crime occur and couldn’t identify any suspects from the video.
The newer hi-def cameras are much more useful for police work. I’ve noticed a big difference in the current season of The First 48.
If the person is using a credit card I’d imagine it would be trivial for the police/FBI to follow them and have store managers pull up those pictures. In a case like that it may or may not help them get in front of the suspects, but if the police come to me and say ‘this credit card was swiped for this amount at this time, can you show me who did it?’, I can have a picture similar to that one in a few minutes.
From there, they can put that on the local news, knowing what the they’re wearing, about where they are and what they’re driving and that may be enough for other cashiers, motel clerks etc to recognize them.
Having said all that, finding some random person waltzing through your cameras is horrible. If I know that someone was in my store on a given day and no other information is known (time, type of car etc), actually finding them on the camera is a crap shoot. A small part of that, however, is due to the crappy controls our DVR has. But even so, scanning a bunch of cameras over a 12 hour day looking for someone you don’t know, is a lot of work.
Another thing, and Acey mentioned this, if something of note happens, they’ll check nearby cameras as well. In all the years we’ve been in business (almost 40) we’ve been held up once. As part of the investigation, the police spoke with a nearby gas station as well as the airport across the street to see if their cameras caught anything at all.
Our local PD also put out a voluntary request to business (and possibly residents) asking them/us to let them know if we have any exterior cameras. That way if any type of investigation takes them near one of those properties they know that there may be an extra set of eyes there to help them out.
Modern cameras (Govt owned ones, not so much private ones) have the ability to undertake facial recognition and then ping the authorities if they see a flagged individual.
Okay, I got it. They used a credit card. Would make it easier to find videos.
Thanks all.
Or it could be that someone thought they recognized the couple or their car and flagged the police to check out that gas station.
But no, there’s no centralized magic machine (currently) that allows them to simply ask every security camera in the land to tell them if someone appears. Most cameras (I believe) just go straight into a machine on the other side of the wall that records maybe a 1-7 days worth of video.
As for how one could filter, if you did want to build just such a magic machine would probably be that you have a thing that recognizes faces (i.e., “Is a face?” / “Is not a face?”). When it spots something that looks like a real, genuine face, it crops the picture down to just the face and bumps it up to the central server. (This is the sort of thing that new AI technologies are good at, but you can do a simple check by looking for something that’s moving, is within the range of flesh palettes, and has white bits (like eyes and teeth in the middle, and that would get you a reasonable guess.)
In the central server, they’d have a more complex system that figures out the angle of each shot. The system would wrap the 2D photo of the face onto a 3D shape of a head, covering all of it that the camera was able to capture. Then they rotate the 3D head to match some standard angles (from the front, from the side, etc.) and take some measurements - general skin tone, width of the eyes, width of the head, pointyness of the chin, etc. and that gives them a set of numbers that they’re looking for. They might have twenty measurements that the system can ferret out of a picture, but each picture might only provide 5-6 (for example). But so you take whatever measurements that you were able to get, check it across all the measurements of all of the faces that you are looking for, and if anything matches all of them with a sufficient confidence ratio, then you go back to the original machine, ask for the five minutes of video before and after the snapshot that was sent, bump that and all of the matched suspect’s photos to a human, and they do a visual inspection to decide whether it’s worth following up on.
Or, again, these days you just pipe it into an AI and the AI does all the same sorts of things as our brain to decide, “Have I seen this person before?” In essence, it’s doing the same as the above, but in a less clearly describable manner. The last check by a human would probably still be there, though.
This.
Our local PD puts out a weekly roster of events taken directly from officer’s reports. It’s real common in the narrative to read “Area canvassed for surveillance cameras. Made contact with owners of same.”
I imagine what happens next depends a lot on the severity of the incident and the tech of the nearby cameras.
Going back to the OP, between mobile phones and credit cards most of us leave a timestamped record of our whereabouts in two different gigantic surveillance systems. It’s not that difficult for detectives to pull the “hits” out of those over the time in question and visit places along the route that they know from experience have good surveillance.
Back when I was in IT in the homeland security space in the mid 2000s we saw state and local agencies that could track the location of a mobile phone in almost real time. No warrant required. This was the era of flip phones becoming the very early smart phones. E911 required the mobile phone providers be able to tell where every phone is all the time. Modern fancier phones make that easier and more precise. But the tech to do that at nationwide scale is now 10+ years old.
We have something called the Safe City Project(s)to be implemented in various cities and the process has been going on sonce 2014. The trick is to link all that information provided by sensors and other devices andcross-reference them with data from various Federal and provincial entities. And then connect them all vide a network to computers.
Back I college I used to have a saying: “Selling computers to the government is like selling guns to the Indians.” Referring to Native Americans, not South Asians.
Given the less-good features of human nature, sometimes it’s hard to see how the middle future can be anything but a techno dystopia for all but a few fatcats and their upper level thuggish henchmen.
Despite all the good intentions of the people who build all this stuff. Safe and convenient brick by safe and convenient brick.
Yup. When our Bar Association was first given a briefing on it, my reaction was "oh fuck the Government has so much technical ability, we are screwed. Then I remembered, it was the *Government *we are talking about. Sure an Orwellian nightmare is upon us. They might have the ability to undertake mass surveillance on everybody. But actually doing so requires efficiency which bureaucracies aren’t famous for. They are after all filled with thousands of CYA workers, who will only move once they have proper authorization in triplicate from the competent authority, but then will fail to act cause the goons union was on strike that week and somebody set the batons to East Africa by mistake.
There an interesting short story I once read set in Orwell’s Oceania. ‘Due To Circumstances Beyond Our Control, 1984 Has Been Unavoidably Cancelled….’ The author took Orwell’s descriptions and then wrote how he thought such a totalitarian regime would actually function. Spoiler; not well. Room 101 has no rats, since they all died and they forgot to send a new consignment in. The telescreens are broken and when Winston tries to get it fixed, he can’t get hold of the aurthorised dealer.
I’m not sure I get why the cameras are the limiting factor - can’t they just upload the video from any camera into a computer that can run the facial recognition software?
You can only recognize a face if you’ve got a high-def enough image to begin with.
The crappy cameras in a typical gas station that haven’t been cleaned since they were installed in 1995 don’t take good pictures.
Brand new high def cameras attached to high quality optics can take beautiful pix. And at long range. For $1000+ per camera. Governments can spend that kind of money en masse. Most businesses won’t bother except in specific narrow areas.
Rest assured the newer casinos in Vegas have very nice surveillance systems. The local bank branch in Las Vegas? Not so much.
Ah, I see what you mean. I thought there was something being said about the software being included in the camera, but it seems that’s not the case.
As LSLGuy says. Modern connected cameras and ones linked to a database can identify a person of interest and ping the authorities. On the other hand cheap commercial units, the type you will see at most businesses, the video feed will tell you that the individual had two arms, two legs and a head.
There is software that can examine poor pictures and sometimes get sufficient data for facial recognition. However, unlike the movies, there is no “enhance” option, if said detail is not in the original picture, the computer cannot see it either.
I recall one case where from the video feed we thought it was a woman, wearing a skirt. It turned out to be a thin guy. And in khakis.
That reminds me of something I forgot to mention in an earlier post. I brought up as a side note that my DVR lacks decent controls which makes slogging through hours of footage a nightmare and like you (and others) have said the quality is less than stellar.
Any time we’ve had something minor happen in our store and the cops want the footage, I just hit the play back button and take record it with my phone, then email that to them. There’s already so little quality that taking a picture/video of my monitor doesn’t lose a whole lot.
However, from what I’ve heard from both merchants that have done this and police, when there’s something more serious, say, a shooting, or they cops want you to look through days worth of footage, they’ll just take the hard drive from your DVR down to the station. It’s my understanding that they have tools there to help make the video look a bit better and if nothing else, a jog dial would make everything easier.
Get the camera oowners to do the filtgering. I have seen a couple of cases locally where after a crime, the police give an exactish time and ask anyone in the area who had a surveillance video cam running in that area to review their footage. A lot of store owners have cameras in their parking lots or facing a front window and dash cams are becoming very common, as is video recording/archiving. Unfortunately, most crimes without witnesses happen at night and video from store cams shooting through plate glass are garbage, so if they’re lucky, the cops get images as good as this, a Could-Be-Anythingmobile: Let's See If We Can ID This Car From A Fatal Hit And Run
Yes, having store owners look through their footage is pretty much what everyone said.
Also, the car in the jalopnik picture looks like a mid 90’s-early 2000’s two door Grand Am to me. Those cars had the wrap around brake lights and that slanted nose with the honeycomb/sectioned front part in the grill that I’m seeing.
And here we see the problem with distributed volunteer labor. The jalopnik car the police want is the one in the background whose grill we never see. 
If the cops ask 100 store owners to look at their 100 videos, the job will be done to 100 different standards. Most of which will be low. The upside is it’s free. If the police don’t much care about the outcome, free crappy work is better than none at all.