The jobsite trailer was exposed to public street because temporary fence abutted either end of trailer. We recommended a sonic detector if anyone broke into trailer from underneath through floor. We also recommended backing up all data and bolting computer, a sign-in book with policies for deliverymen entering site, and better maintenance of water and mud under gate used by pedestrians to avoid slips.
Beyond the Rules
And now for a complete shift of view, really a revolution to the next level of artificial intelligence video analytics. BRS Labs (Behavioral Recognition Systems) has created a non-rule-based analytics. Unlike the traditional type where you must set rules for each camera, this requires no setting of rules. If you think about it you will quickly understand that the rule-based type works very well for unoccupied, restricted areas. It tells you as soon as someone has trespassed there. But what about occupied areas like shopping plazas, university campuses, hospitals, loading docks, work yards? The rule based cannot work in these areas because people belong there all over the place all the time.
The non-rule-based BRS system watches the cameras and learns, on its own, what is normal behavior for humans, vehicles, machinery, and the environment. It then reports when it observes something that looks abnormal in the patterns of behavior. So this is much more subtle and complex. It can detect an employee breaking a safety practice. For example, if everyone uses a ladder to get down somewhere, someone jumping down would be detected. Someone “tailgating” behind an authorized person who used their proximity card or tag to get through a controlled door. A vehicle driving the wrong way. A fire breaking out in a dumpster outside. Someone who fell off his bicycle. Things you could never anticipate or even think to look for. Phenomenal! This can reduce workers comp and public liability issues as well as crime, both internal crime and outsider crime. But there are some limitations to understand.
One is that due to the complexity of the analysis, this samples down to one cif resolution which is very low. This means its effective range to analyze humans might be around 100’ for most camera lens settings. (Higher mm lens or zoom would increase distance but decrease width of field).
The other thing to recognize is that the artificial intelligence (A.I.) is not yet as smart as Star Trek’s Lt. Commander Data. It’s still 2015. So the first time the A.I. sees something unusual such as someone suddenly opening an umbrella, or its first snowfall, it will send an alert. A different mindset must apply to the security professional utilizing this. No human can monitor 300 cameras. The A.I. will pre-monitor all 300 at once and give you a “tap on the shoulder.” It’s saying, “Here, human, take a look at this. It might be something.” If it is nothing —and usually it will be that — the security officer just glances at it and moves on. The security officer is already employed sitting in that command station. This greatly increase situational awareness, given the teamwork of A.I. and human. If in the course of a year it finds one outdoor fire, or one person who accidentally fell down a flight a steps to a basement and would otherwise have lain there unobserved, then it was all worth the year’s looking at alerts.
So Which Type is Right?
I love that question because it is so easy to answer. Both types are right. You can mix and match them. If you have a long fence line and a large restricted area, for example a manufacturer’s grounds and facility after hours, then the rule-based analytics will give great proactive protection. Coupled with an outdoor public address system, the remote security officer can immediately talk down to the intruder. This dissuades nearly all intruders. Nearly all are opportunists. A live person talking to them is pretty convincing that their best opportunity is elsewhere. If not the police give such verified breach reports their highest priority for response. (Unlike burglar alarms, which they give very low priority due to such high instance of false alarms —nearly 98%). On the public address system, high quality amplified speakers with large horns, fed from a sufficiently large amplifier, assures crystal clear voice over very long distances. Cheap little squawk speakers won’t work. But a good system is not expensive.
Where you have areas that are occupied and active and where bad things can happen, there you would want the BRS non-rule-based video analytic artificial intelligence.
If you are at a large corporation, utility, or institution, talk to your risk manager and insurance person. The benefits of your implementing these improvements should be rewarded by your insurer in the form of scheduled premium credit or discount. This is not “security” in the former sense of getting a video recording the next day once it is already too late. The value of such recordings for outdoor areas for forensic identification are dubious or at best hit or miss. This is real security. Something preventative and proactive. It not only can reduce crime but other major sources of loss such as workers comp and public liability.