Ask Your Question
0

Ways of Doing People Detection

asked 2014-07-15 12:38:09 -0600

yohanrw gravatar image

updated 2014-07-16 00:02:50 -0600

Hi Guys,

I have a very challenging task; detect the people in images. This is my challenge.

  1. Detect people in images (but make it quick, should not take more than 5 seconds).
  2. People might be walking, standing, or sitting.
  3. The Race of the people should not matter.
  4. The background environment should not matter. (ex: Person might be sitting next to Home, Person might be standing infront of the car, people might be walking in beach)

How can I achieve this task? I can only think of Cascades. Any ideas? How good the Cascade training should be? any other ideas?

edit retag flag offensive close merge delete

Comments

Google part based modeling by Felzenszwalb or integral channel features by Dollar. No existing openCV techniques achieve your demands.

StevenPuttemans gravatar imageStevenPuttemans ( 2014-07-15 16:14:53 -0600 )edit

@StevenPuttermans: Hi, My requirement has been changed from real time to detect in images. Please have a look at the updated question.

yohanrw gravatar imageyohanrw ( 2014-07-16 00:04:25 -0600 )edit

1 answer

Sort by ยป oldest newest most voted
3

answered 2014-07-16 02:07:14 -0600

As @Witek already described, packages like doppia, created by rodrigo benenson are quite powerfull and will meet your real time requirements. However, if you want another approach, based on picture by picture analysis, these are the techniques that openCV supports.

  1. Cascade classification based on the AdaBoost principle using the Viola & Jones framework. Click here for more information. This can be done with HAAR, LBP and HOG features.
  2. You can use the Histogram Of Oriented Gradient approach created by Dallal & Triggs, which is explained right here. It combines HOG descriptors and SVM classification.
  3. The part based modeling has an early implementation in OpenCV called LatentSVM of which documentation is found here(http://docs.opencv.org/modules/objdetect/doc/latent_svm.html?highlight=hog).
  4. With a static background you could use efficient background subtraction using Gaussian Mixture Models to improve the detection results.

Start reading and experimenting!

edit flag offensive delete link more

Comments

And to add to this, even if you use these techniques, using knowledge like a fixed scale or a limited scale range combined with the efficient use of a tracker can get you real time performance!

StevenPuttemans gravatar imageStevenPuttemans ( 2014-07-16 02:11:45 -0600 )edit

Question Tools

1 follower

Stats

Asked: 2014-07-15 12:38:09 -0600

Seen: 9,219 times

Last updated: Jul 16 '14