Ask Your Question

Detection of moving objects in a real scene

asked 2012-11-05 05:21:50 -0600

rBellsound gravatar image

updated 2012-11-05 05:25:27 -0600

I have series of photos of one scene that was photographed with unstabilized camera with shaking hands, for e.g. 10 seconds. The time interval between photos is not constatnt: 0.2s - 2s.

Is it possible with openCv to detect mooving objects - just get the rectangle they are filled in - like people or cars? The objects can not be colorod with any special color and the scene contains partially moving objects such as trees that "moves" by the wind, but these should not be detected.
Also because of quite long time interval between each photo, the object, depending on its velocity, can be quite distant from its previous position (on previous photo).

I am new to OpenCV, could you please suggest topics I should learn to get this done, ideally on Android ? Thanks!

edit retag flag offensive close merge delete

2 answers

Sort by ยป oldest newest most voted

answered 2012-11-05 11:00:41 -0600

imran gravatar image

Hi a very good tracking algorithm which I would recommed was developed by Zdenek Kalal. Just search for OpenTLD. There are MATLAB and C++ versions. The C++ version makes extensive use of OpenCV. There is also being research done on porting OpenTLD to Android, see

edit flag offensive delete link more


Thank You very much Imran for pointing to OpenTLD. I am not sure, if I can handle this technology for current project on Android, but it is definitely worth of study in longer term.

rBellsound gravatar imagerBellsound ( 2012-11-06 15:53:23 -0600 )edit

Not a problem, all the best! You should add images to your question to show exactly what you need, then you might get more responses and a better push in the right direction.

imran gravatar imageimran ( 2012-11-13 15:42:20 -0600 )edit

answered 2013-06-24 01:26:15 -0600

Shaban gravatar image

updated 2013-06-24 01:28:01 -0600

Try this tutorial to detect any moving objects on the screen:

And to handle the tree shaking (false object detection) you can use the average of color as a seed to grow pixel from the center of the blob/contour. if the number of pixels covered by the extended region is more than the number of original blob, then this object is false. SOURCE: Real Time Robust Human Detection and Tracking System by Jianpeng Zhou and Jack Hoang. :)

edit flag offensive delete link more

Question Tools


Asked: 2012-11-05 05:21:50 -0600

Seen: 5,373 times

Last updated: Jun 24 '13