Counting bounding boxes inside ROI

6 views (last 30 days)
Hi all, I am currently working on a project on vehicle detection. One of my goal of this project is to count the number of car inside ROI such a lane. the method I think of is to using the centroid of the bounding box to see if it is inside the ROI. but I haven't got a clue on how to start. Since the ROI is a polygon and the centroid is a x,y coordinate, how to know if it has leave the ROI?

Accepted Answer

Image Analyst
Image Analyst on 23 Mar 2014
Yin Lok, if you're tracking - something like on this page - then you have one ROI per tracked/detected object. If so you know how many objects there are . If you have multiple ROIs and some of those ROIs have multiple vehicles inside them, then post your image.
  2 Comments
Image Analyst
Image Analyst on 25 Mar 2014
If you have multiple lanes and you have, say, 3 boxes in the left lane, and 5 boxes/cars in the right lane, then if you have polygon coordinates that define each lane, and you have the middle of each box, you can simply use inpolygon to determine if a box middle is inside a lane ROI polygon.
Andre
Andre on 21 Aug 2016
The link goes to features of Computer Vision System Toolbox. Which example there are you referring to?

Sign in to comment.

More Answers (2)

Joseph Cheng
Joseph Cheng on 21 Mar 2014
There are many ways to do this but it depends on how you are defining or receiving the ROI. Is the ROI defined by its corners such that you get 3+ (x,y) points and anything within the triangle to polygon is the ROI? If so it is like if i give you 3+ "y=mx+b" lines and ask if the point is above or below each line. If the ROI is just a mask of zeros and ones then you can just check the (x,y) location in the mask.
  2 Comments
Yin Lok
Yin Lok on 23 Mar 2014
Thanks Joseph, my ROI is defined as a binary mask so i just need to just the x,y coordinates in ones?
Joseph Cheng
Joseph Cheng on 24 Mar 2014
Sorry for the late reply, I've been thinking about it more and using the information in your latest comment, you can use the binary mask to get rid of the image outside of your ROI. Then if you are doing anything like the bottom example in the link Image Analyst supplied, you can probably put a minimum size threshold for the vehicle size and remove that tracked vehicle as it leaves the ROI.

Sign in to comment.


vaishali masal
vaishali masal on 29 Oct 2016
I am doing project of vehicle-number-plate-recognition. the whole i got http://in.mathworks.com/matlabcentral/fileexchange/40426-vehicle-number-plate-recognition on this link. but that code is valid only for 6 characters. But in India vehicles have 10 characters numbers for eg: TN32 BZ 7701 Can you please tell me how i should modify above code. mail id: masal.vaishali18@gmail.com
  1 Comment
Image Analyst
Image Analyst on 29 Oct 2016
You posted this as an answer to Yin's question mistakenly. This is not an answer to Yin's question. You should post this as your own question.

Sign in to comment.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!