Ask Your Question

objecttracking methods, best solution?

asked 2016-06-23 07:26:53 -0500

atv gravatar image

updated 2016-06-23 07:31:20 -0500

Hi, I realize there are many, but If i wanted to check if an object had moved, what would be the best method for me to implement.

template matching, or use the tracker or multitracker solutions from opencv? (are these in any way similar to TLD/openTLD ?) i don't think the opencv methods use HSV color values or something?

And a unrelated question 2: why can't i use template matching for face recognition? Can i not use detectmultiscale and then use template matching to recognize the face?

Thank you.

edit retag flag offensive close merge delete


afaik openCV 3 has a TLD implementation, but I can't tell you the function names.

Micka gravatar imageMicka ( 2016-06-23 07:52:53 -0500 )edit

These are the ones i'm talking about

atv gravatar imageatv ( 2016-06-23 08:26:32 -0500 )edit

1 answer

Sort by ยป oldest newest most voted

answered 2016-06-23 20:35:30 -0500

Tetragramm gravatar image

To watch for motion, pyrLK Optical Flow is probably the simplest choice. Find a set of keypoints on the object, and track them to see if they move any significant amount.

Match Template is capable of using multiple channels, as are several of the other trackers, such as TrackerKCF. Whether that is HSV or RGB or YUV depends on the images you pass to it.

Template Matching is far more general than any decent facial recognition software. Template Matching would see only very small differences between a face with high cheek bones or not. Also, a turned face would be very different to template matching, with a different size and shape. Google some face recognition tutorials to see how they usually work.

edit flag offensive delete link more
Login/Signup to Answer

Question Tools

1 follower


Asked: 2016-06-23 07:26:53 -0500

Seen: 133 times

Last updated: Jun 23 '16