How to automatically count people in a bus or a room?

ChemAli2012-11-27 14:30:00

Algorithms

ChemAli, 2012-11-27 14:30:00

Transport is more interesting. The goal is transportation statistics, control without human intervention.

There are no thoughts. Something infrared? But heat spots can overlap. Recognize faces? The same. Photocells for input and output? Not very reliable, there will be false positives.

Answer the question

In order to leave comments, you need to log in

15 answer(s)

Talyutin, 2012-11-27
@Talyutin

100% accuracy will not work.
The best option is photocells for input-output. In many malls, this is how I count.

akral, 2012-11-27
@akral

Seat a student with a counter.

DbLogs, 2012-11-27
@DbLogs

There are traffic sensors.
I highly recommend this one:
npp-alion.ru/solution/avtotransportnyie-predpriyatiya/
npp-alion.ru/products/passenger-sensor/
Knows how to take into account the direction: that is, how much goes in and how much goes out. And also the error is ten times less than analogues.

dlinyj, 2012-11-27
@dlinyj

Issue an RFID tag to each passenger and count the number of its activations.
More exotic, to recognize the number of faces that pass through the doors. It depends on what statistical error and what requirements. If you can have a large error, then you can weigh the bus on special scales, and evaluate the amount of live meat in the cabin, and calculate the number of passengers from it.

rainwall, 2012-11-27
@rainwall

Turnstile at the entrance as in MGT buses.

ipswitch, 2012-11-27
@ipswitch

In Moscow, as an experiment, experimental meters were installed in MosGorTrans buses on the front doors - a combination of an infrared detector and photocells. The system did not count the actual number of passengers at each moment of time, but only recorded the number of people who entered (because the front door only works for the entrance). We add a similar one to the exit doors (they work only for the exit) and a little logic - and that's it.
It is also possible to use the turnstile rotation sensor. In Moscow (ASKP system) you can count only at the entrance - the number of turns of the turnstile turntable = the number of people who entered the bus. In the area (the CASCADE system), where double-sided reversing turnstiles are used, there is also an exit.

Ivan Trofimov, 2012-11-27
@cbone

Photocells for input and output? Not very reliable, there will be false positives.

Install two photocells at the input one after the other in order to reduce the error.

niko83, 2012-11-27
@niko83

As a rule, they step on the very first step in any way when they leave or enter, moreover, once, and they do not linger on this step. Can experiment with turning the first step into a button. + Photocell, the cutoff between the outgoing and incoming flow will be guaranteed.
the speed with a short time people enter and exit, as a rule, also lies in a very limited range - you can try to make some kind of filter, for those who just mark time in one place.

Akson87, 2012-11-28
@Akson87

Lots and lots of IR and/or ultrasonic sensors on the ceiling, counting distances, building a height map and isolating heads. But this is true, thinking aloud, it will be very expensive and very good processing algorithms will be needed.

Andrew, 2012-11-28
@OLS

It is interesting, but if you install a short, short (25 cm) metal handrail (so that it would be inconvenient for someone who comes in twice to take it) and measure its capacity. Any sharp jumps - someone is taken / released.

rPman, 2012-11-28
@rPman

After suggestions about photocells, handrail touch sensors, etc. I can offer an unusual idea ... install sensors on the bus (gasoline consumption, roadway inclination, speed, ...) and, based on them, concoct a mega-complex formula (or train a neural network) that calculates the weight of the bus ... then divide by the average weight of passengers (during training people can be counted 'manually').

AJ, 2012-11-28
@2ball

Photocells for input and output. 1 at the level of the legs, the second at the level of the head. + pressure sensor on the bottom step. Next, we write a synchronizer to count not pressed bodies, but real triggers.

Kirill Mamaev, 2012-11-29
@r00tGER

Weigh! Of course, rather a hypothetical proposal, but it's worth speculating.
Put the sensors on the suspension, analyze the empty and filled cabin based on the average weight of an adult.
Also, the accuracy of such sensors will allow up to tens of kg to track changes in weight during boarding and disembarking. And this is additional information that may be useful for a more accurate calculation. For example, track passengers with "not average" weight, such as children or passengers with luggage.
In general, it is possible to implement a rather complex algorithm for processing changes in the total weight during landings and landings, which will give a high percentage of accuracy.

alexk24, 2012-11-29
@alexk24

habrahabr.ru/company/croc/blog/151680/
Here they made a turnstile with face recognition. Perhaps they will be able to make a system that will count the passengers "over their heads." In general, the task set looks somewhat simpler than what has already been implemented, and at the same time, existing developments can be used. But the question arises in the budget of such a system.

Maxim Kotenko, 2017-09-07
@mistik_max

It is very easy to calculate by the number of tickets sold, as an option)