Estimating parked car counts from occupied area
- This topic has 4 replies, 3 voices, and was last updated 13 hours, 32 minutes ago by
.
-
Posts
-
Hello-
Wondering if anyone has any idea about estimating number of parked cars based off a square foot measurement.
Workflow would look something like this: Measure estimated area of parked cars using google earth or something similar based off human observation -> take area and divide it by estimated cars per sq ft/sq m -> get estimated vehicle count. From there you could estimate low and high numbers of people attending event.
Not as straightforward as it seems as there still needs to be areas for the cars to pass through.
Thanks!
There are a few open-source examples with training datasets (like COWC) that you might be able to utilize by feeding them your own overhead images…
Example Data:
https://github.com/LLNL/cowcOS Tools:
https://github.com/JacksonPeoples/CarCounting
https://github.com/AlperenCicek/vehicle-detection-from-satellite
https://github.com/langheran/APKLOT
https://github.com/motokimura/cowc_car_countingOf course counting cars to get a headcount is a bit more nuanced.. e.g. counting cars in a office lot on a workday you’re more-likely to have closer to 1 person per car? A large gathering, like a sporting event, you’re maybe closer to 2-3 people on average?
I believe when designers try and work out a question like “how many parking spots should a stadium that holds 20k people have?” that the number is at least 2.5 per spot or higher.
A lot of places are still doing it old-school with hand-held “tally counters” at the ticket gates or on how many rotations of a turnstile. Modern places where they electronically scan your ticket and can provide real-time numbers..
IIRC the last time a similar question like this came up on here, I believe it was related to estimating protestor attendance, I had suggested capturing Bluetooth and Wifi device beaconing — ‘everyone’ has a cell phone these days, right? Bluetooth seems to be the most prevalent method. It’s already being used for advertising and by retail stores / malls to get traffic counts, or identify hay many people are in a particular section of a store..
E.g. https://hackaday.com/2023/12/13/bluetooth-as-proxy-for-occupancy/
Yeah this isn’t from satellite imagery or a photo.
Just an event where you can visually recall how full it appeared where the best way to figure it out would be measuring the area with a computer tool.
Or maybe there’s another way? Open to suggestions.
The formula I would use: SPOTS * %USED * 2.5 = ATTENDANCE
For ideal conditions marked parking spots (in the US) are ~7 1/2 to 9ft wide and anywhere from 18-20ft long. So about 152sq ft per car.
If it’s parking in a unmarked field, you’ll need to be much more generous with the space calculations. ~20% gets you about 10×23 or 230sq ft per car.
So.. if an event Location has ~100 spots and it appeared 80% full -> 80 spots * 2.5 people = 200 people in attendance.
- You must be logged in to reply to this topic.