Here’s my dumb question for today…
We are smart enough to build self-driving cars, right? That’s a camera and computer vision software that needs to see everything happening on the road, including pedestrians.
It doesn’t take specialized training for a human to work as a crossing guard, right? The human sees a pedestrian approaching an intersection, walks out into the crosswalk holding a STOP sign, and stops the traffic.
Why would we wire up the light poles surrounding intersections with switches to activate WALK signs? Why not mount a camera up on the pole to watch for approaching pedestrians? Then, if no cars are coming, turn the light red for cars and activate the WALK sign before the pedestrian has to break stride.
How hard can this be? Maybe this could ease traffic congestion slightly by making it more pleasant to walk. At a minimum, we could save a lot of money installing and maintaining the under-pavement sensors for cars. The same camera can simply watch for a car approaching or stopping at an intersection. That should also save fuel (and the planet!) by changing the light before the car has to hit its brakes.
Camera plus microprocessor plus software should be cheaper than sensors, wires, and maintenance, no?
Good idea, but we might still want those buttons to give impatient pedestrians something to do while they are waiting. Elevators still have “close door” buttons, even tho they do not do anything anymore.
Already mostly done. See http://gridsmart.com for details. There are several around MA. There is one at the entrance to Stop and Shop on Great Road in Bedford. I think the pedestrian module is optional and may not actually control the signal. It does eliminate the wires in the road.
If only machine vision worked as well as valuations of machine vision startups. The current state would have an awful lot of cars stopped for imaginary pedestrians. Better to sell a company that makes traffic cams than sell traffic cams.
ditto good idea. Re walk buttons, I once read (long ago) that an inventory of the buttons in NYC found that a majority didn’t do anything.
Similar vein: Some air traffic control (getting queued for landing for example) is already little more than a human relaying computer generated information to the pilot–do we really need the human?
Z,
We need the human to override the permissions so the system achieves more throughput. If the basic rules were enforced, the system would be much less productive.
It’s been done. I don’t know how common it is, but I’ve seen it in use at an intersection between a pedestrian pathway and a bus transitway. Works well, as far as I can see.
How does the camera know which way I might go at the intersection? As a tourist I might not even know. You can give a wave to a crossing guard, should I wave at the camera too
Waving at camera could be made a gesture understood by video-processing software. And tracking walking direction for few yards towards road edge should work.
I work for a company which makes support equipment for computer vision systems, so on one level I support this, although I question the need for AI in this application (and many others, as well) Here’s my dumb question: how about we install buttons at crosswalks as an input to the traffic light so that pedestrians could press the button when they want to cross the street?
jay c # 10:
Not sure whether AI is needed here either, humans and their gestures are quite uniform and crossing recognition can be done with defined statistical algorithms and heuristics, to limit application software surface attacks and un-intended use.
Dumb button: 1. not sanitary and 2. is subject of denial of service attacks by a single individual. That’s why most buttons are not functioning properly.