Stuffed bunnies not the only things T drivers stop their trains for

Joe Pesaturo at the MBTA reports:

At 12:17 p.m. today, an intoxicated man stumbled and fell beneath a Green Line trolley that had just made a stop at Haymarket Station. Preventing a tragedy, customers on the platform alerted the trolley operator of the situation before the train started to move. The MBTA would like to express its gratitude to the customers who "Saw Something, and Said Something" before it was too late. The man was not injured.



Free tagging: 


I think...

They are looking not for people, but for objects that appear, and then do not move or leave.

The image processing doesn't seem to be well tuned for people, so I expect if one of those people were to leave a bag on the platform, the bag would get a red box, which if it stayed in one place for too long, would cause some sort of automated alert for the T cops or an attendant to go inspect.

See something, say something for computers.

Motion detection

The squares, though laggy and inaccurate, indicate that a moving object was detected. At times, the motion detection algorithm gets stuck on the darker tiles in the left foreground, which is interesting.
I think you're talking about a different surveillance algorithm, called abandoned or static object detection The algorithm behind the red squares in this video isn't doing that. The AOD algorithm filters out motion and flags introduced objects that remain inactive for a specified period. In this video, an AOD algorithm would probably flag the train and the immobile person on the ground after a minute or so.

Here's a YouTube video on "Static Object Detection - Video Surveillance System":

I want to know more about

I want to know more about those red boxes, too. Maybe they scan for new stationary objects in the scene, because they don't focus on the existing ones.

And, how do they work? Could a person standing very still on a platform, or sitting on a bench trigger an alarm?