2
$\begingroup$

The idea is to get three classifications for crowd: High, low and medium based on an image captured inside a train compartment. Something like this: enter image description here

This gets classified in the 'High' category

http://www.spiceflair.com/wp-content/uploads/2012/05/14a.jpg This perhaps in the 'Medium' & so on.

$\endgroup$
  • $\begingroup$ This may be a more suitable question to ask in tbe cross validated forum - they're for machine learning type questions. My best guess is if you have access to a training set, compare some test statistic of a wavelet transform which could allow you to classify. $\endgroup$ – Tom Kealy Sep 12 '13 at 15:51
0
$\begingroup$

There is research being done on crowd monitoring/crowd counting, which is similar to what you are looking for. The UCSD Vision Lab is doing work on this.

Here is a paper from this lab that estimates the size of crowds [PDF].

It is not exactly the same as your train scenario, but the idea is something that could be used for this problem.

$\endgroup$
0
$\begingroup$

This sounds feasible (and fun!), but it will be hard to get very accurate. My guess is that, like your tags suggest, this would be a combination of edge detection / space detection to get a measure of density and face/hand detection to get a measure of human content. You could then train up an SVM or RandomForest model on the data, tweak, and repeat.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.