|
Detection and measurement of parked cars
quentin from France [1 posts] |
10 year
|
Hello everybody,
I tried roborealm months ago, but I didn't really take the necessary time to finish the little project I was working on...
So now I bought a licence and have the necessary time to work on it and discover.
Anyway, I don't really know how to begin with, I tried a lot of stuff but didn't manage to get what I wanted..
I'd like, from my window, check if there are any available parking places and measure the length of the parking place to see if my car could fit.
I'm sending you the picture i'd like to process, there are empty spots for you to see the texture of the road. (I had some information on a previous post i made 2 years ago, but unfortunately I didn't manage to get it work...)
So if anyone could give me some steps to be able to get the information i'm looking for, that would be awesome !
thanks !
|
|
|
|
quentin from France [2 posts] |
10 year
|
hello!
nobody has a little idea to help a bit? i'm trying so much stuff with so little success, help would be deeply apreciated... thanks!
|
|
|
|
Steven Gentner from United States [1446 posts] |
9 year
|
Quentin,
Thanks for the image. That helps me understand what you are after. You'll need to take a bunch more in different lighting (cloudy day, morning, evening, etc) to verify any technique will work.
The obvious technique is to subtract the background from the current image. Ideally the background would have no cars in it which would then indicate when something has changed and therefore be a car occupying a spot. The issue is that the background image needs to change over the course of the day since the lighting will change.
Instead, you might be able to look for the edges that a car generates against the asphalt and look for the presence for something in that way. I've attached a robofile that focuses on just the nearest side (for now) and uses the Sobel module to get a sense of if anything is in the way. The largest gap is then detected as the largest area that has low edge noise in it.
As you probably have found out, there are many ways to accomplish a task. Once you get more images another technique might be better but you have to start somewhere. Perhaps you can post some of the trials that you tried.
STeven.
 program.robo
|
|
This Thread