After the fact there are a few that could be applied. Without understanding what is normal and response times it might be a waste to apply any. Could use loitering beyond 3 minutes. They certainly met that. Could use classification of Vehicle too long since it appears to be an alley and 2 or 3 minutes is long for this car. Now, what is the response time? Security guard nearby? Owner gets message or central station to review.....then what. There could be numerous outside influences that would make any analytic difficult.
As my retired LAPD father-in-law said for years.....jail is full of stupid criminals.
Could use loitering beyond 3 minutes...
Or maybe BRS Lab's generic "something funny" is going on detection.
BRS would detect it but I think it would be a bit much for a place selling small propane canisters. Pitch it and let me know how that goes.
Your best option would be loitering, and/or possibly multiple people (>2).
The important thing to understand is that there is no analytic that would be 100% effective here. This scenario would be best served with some kind of remote monitoring component. That could be an outsourced company, or an employee of the propane company if they already have some kind of 24/7 ops center.
The goal is, generally, to find a way to monitor a large number of sites so that you can have 1 operator handling a large number of locations. Realistically they would receive several events that were not crimes. You might set a loitering alarm for 3 minutes, and you would have some really slow customers that trip the alarm, but the remote guard is able to easily see that it's not a crime, it's just a slow customer.
This is also why pre-recorded messages and things like that do not end up being very effective for sites like this. You risk alienating good customers, and the criminals might run away the first time, but they will come back and soon realize the message is just a recording and has no real threat.
Analytics for sites like this are very good as a force multiplyer, they can make it cost effective to monitor things that could not be done before. Analytics are not a "perfect" solution, where you can expect that every alarm is a real crime, and also that you don't miss anything either.
Super sirens that penetrate the upper spectrum so as to disable there ability to reason.
Very high pitch. Software would give you email messages but would you really want to respond.
Industrial Great Sirens would give such a headach they would thing twice before coming back
Most video VMSs give the ability to trigger from the software. install perimeter link around the inside and when it goes into alarm then send emails, text mssges , and triggers .
Bottom Line is they got away with it, now they believe that can do it again and get away with it.
so you will see at least one, if not all of this group in the headlines again, and again.
Using a High Pitch Siren above 20khz would Scare off the perpatrators, say 195 db , and using the wire would send out emails when contact opens. Oviously you already have a camera with nite vision.
The pain level needs to be high enough to scare them off.
Another way is to use a high impact protected strobe which when activated would output lites .
Or use a wireless trigger to trigger the out put of the cameras to start a recording sequence or another sequence of horns, lites, strobes.
Very easy to set up
If they really want into the area , the only real thing you can do is harden, notifiy, alert whoever and then send out someone to respond, which creates another situatation , are the responders qualified to take on these criminals, and can the evidence be used against them.
Seems to me the obvious analytic here would be loitering combined with a schedule that detects loitering at 1AM and asserts an alarm.
Analytics that distinguish between 'normal' and 'abnormal' human activity are more difficult than analtyics that classify activity where none should exist (as in, somebody in your store when it should be shut up tight). This has worked for relatively crude motion detection systems for years--it can work with video analytics as well with lesser false positives than when those same analytics are applied to scenes that include normal human activity.
If you look at your time constants here you see they had all nite to complete thier mission.
If you would have exacqvision with email triggers from a camera with trigger inputs and a wireless transmitter to trigger the cameras you would get perfect pictures sent at time of action or event. If you set up sound or lites this would possibly scare them away.
If the are so determined to complete thier mission of theft , they will be back for the rest .
If you damage thier hearing , well then they cannot plan future events
either way you detect and notify of events.
deterance goes a long way.
And signage helps even more with lots of pictures so they can understand.
Lites , Sound , Cameras