Maxim V. Pavlov
Maxim V. Pavlov

Reputation: 10509

Tracking people is camera scope with computer vision

I need to be able to track people (and have a current count) in scope of an IP camera. As far as I understand (correct me if I am wrong), OpenCV is the most feature-rich computer vision library. Does it provide a ready-to-use routines for achieving this task?

Upvotes: 1

Views: 2806

Answers (2)

Eric
Eric

Reputation: 2341

Mean-shift methods based on distance between histograms of color (HSV space better) is easy to implement using opencv. You should have a look at this article for the detection of people which is the first task for tracking. It use histogram of oriented gradient as features and linear SVM as classifier. You can find codes in opencv for people detector. For the tracking, to handle complex events like occlusion, you probably need to investigate probabilistic tracking using particle filter.

Upvotes: 1

jlengrand
jlengrand

Reputation: 12827

Opencv does not embed directly such a complex application, as It is often conditions dependant. It does however contain a lot of useful tools to achieve this objective: http://opencv.willowgarage.com/documentation/motion_analysis_and_object_tracking.html

In addition, you can find some good open source code on the web related to theses issues : http://www.youtube.com/watch?v=bWl33urh2w8

And you can also find some good tutorials and explanations: http://www.neuroforge.co.uk/index.php/tracking-methods-in-opencv http://www.geckogeek.fr/tutorial-opencv-isoler-et-traquer-une-couleur.html (warning, french inside)

Upvotes: 2

Related Questions