0
$\begingroup$

I am working on a problem where I have to find how many people are passing through a point and how many times they pass through. A good example use case for this problem would be counting how many times an person enter and exit the building.

Detecting people is somewhat easy, I have relatively high SNR images so I use machine learning (pedestrian detection) to see how many people come and go. Achieving this has been relatively easy. However identifying people throughout the day has been a more complex challenge.

I do not want to use faces since they just walk in so it is not always feasible to get clean face shots. They don't change cloths during the day so visually, I can relatively easily identify who is who when I watch the video without seeing faces.

Since I am able to pick out the pedestrians in bounding boxes nicely, I now need a way to compare them and barring light changes, perspective changes etc. identify if they are the same person or not.

Where should I start for this type of algos?

$\endgroup$
  • $\begingroup$ Not just the color, size of the person and the velocity at which the person moves can also be used.. $\endgroup$ – Navin Prashath Nov 27 '16 at 5:26
  • $\begingroup$ @NavinPrashath I am trying to avoid color, these are outdoor images and depending on the time of the day, the colors shift. It is difficult to assume colors stay the same. $\endgroup$ – Ktuncer Nov 28 '16 at 3:19
  • $\begingroup$ As a naive suggestion, I suggest to do some search on gait recognition. $\endgroup$ – MimSaad Nov 28 '16 at 14:32
  • $\begingroup$ As a naive suggestion, I suggest to do some search on gait recognition. $\endgroup$ – MimSaad Nov 28 '16 at 14:32

Your Answer

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

Browse other questions tagged or ask your own question.