Now You See It and Now You Don't!

Humans strive for 20/20 vision especially when playing “Where’s Waldo”! That level of vision is what is sometimes considered perfection (yes, if you get Lasik you can get to 15/15 but that is beside the point for this post). So, if we consider that to be good as a robot that can see we should strive for the same 20/20….that should be good enough, right? Well, not here at Knightscope, we push our robots harder than most humans.

Knightscope robots are intended in the long-term to be built to “see, feel, hear and smell”. Let’s focus on “see” today.

Every robot comes equipped with at least 4 cameras that cover 360 degrees around the machine for both day and night - and most importantly they are at eye-level. We do like to joke that sometimes CCTV cameras are very good bald spot detectors especially when positioned up high on buildings. :-)

Every camera is on all of the time (even when the machine is autonomously charging) so it gives the machine the ability of continuously seeing what happens around it 24/7/365. That is as far as most of today’s security solutions go. We go a step further, we ask our machines to not only see but also to make sense of what they see through a variety of calculations and report what is out of the ordinary. That way the machines can alert our clients of any suspicious behavior that should be further investigated by a human.

If a robot is going to be a good partner for humans, at least it should see just as well as they do. But what happens when even a 20/20 vision is not able to differentiate shapes at long distances or in blurry images? Take a look at the example below, there are 3 people in that image, 2 are pretty obvious but where is the third person? Can you “see”? Or is it actually 5 people?


Not as easy or obvious is it? But since robots are highly skilled and trained with hundreds of thousands of images, they can detect a person in a variety of positions, from far, far away and in many cases, robots can sometimes detect people or objects that a human cannot actually see or distinguish. Here is the image above but with the third person enlarged for visibility. But are there 2 more?


We have a lot of examples like the one above that we have collected now that we have operated over 700,000 hours with real clients in real world conditions 24/7/365.

This a good example of when machines can do more than what a human can do so there is a place for both on the security teams of the future - working together to give the 2+ million law enforcement and security professionals really smart, new eyes and ears for them to do their jobs much more effectively. And they need the help as they are trying to secure 328+ million Americans across a massive country made up of 50 states!

William Santana Li