In the early 00s, Cernium created Perceptrak, a proactive video alert management platform that used rich video analytics to assist operators of very large surveillance systems by “watching” the video from hundreds of cameras for them. When we first configured a system, it was not unusual for a customer to ask that we turn on all of the eighteen-odd possible behaviors on every camera, so that they “didn’t miss anything”! It would usually take a bit of persuasion, but eventually the customer would see the obvious – that lots of useless information swamps any good that the system might do – and agree to a smaller menu of operationally relevant behavior types for each camera.
With the advent of video-enabled mobile devices, manufacturers of video monitoring products have rushed to make mobile alert viewing a key feature. We’ve seen this movie before, and we know how it ends. Armed with motion detection, many of these systems send text or email messages every time a branch moves. Or, as a reviewer of one product put it, “A system that recorded three hours of swaying trees yet failed to notice the UPS man delivering a package is not something we want to rely on…” And with mobile devices now being asked to do everything imaginable for us, the flaws in any screen or communications hog are magnified. The last thing you want to experience on the go is a continuous barrage of information you don’t need.
Once you offer mobile alerts, how do you balance the desire to not miss anything with the possibility of seeing way too much? That is, can we satisfy the “Three Bears’ test” of not too much, not too little – just right? We were very sensitive to that issue with the creation of Archerfish, which was designed from the ground up as a mobile service. Rule #1 in any feature that we offer or modify is, “Thou shalt not annoy the customer.” We did three things that we think have made a major difference: 1) we deployed advanced analytics instead of motion detection, so if you’re looking for people, 99% of the time you’ll get an alert about a person instead of something else; 2) we balanced the sensitivity with some proprietary tricks, so while it won’t get every object right all the time, it also won’t trigger on everything that moves; and 3) we told Archerfish not to repeat itself, so that unless a situation persists, you will usually be told about it once, not over and over every few seconds.
This is our approach to the Goldilocks solution that manufacturers and mobile users are seeking. As new video applications proliferate, we expect to see a host of innovative answers that will quickly help the industry converge on the “just right” approach to mobile alerts.