people counter

edit

I would like to develop a people counter:

"A people counter is a device used to measure the number and direction of people traversing a certain passage or entrance per unit time. The device is often used at the entrance of a building so that the total number of visitors can be recorded."

Something similar to the following figure:

I have the freedom to choose : 1) the camera type (depth camera, rgb camera, ...) 2) position of the camera and 3) orientation of the camera 4) resolution of the camera…..

For example if the camera is pointing perpendicular to the floor, a captured image could be something like this:

In this situation, I think that the best approach will be based on background subtraction + blob tracking, or even optical flow or something like this. I don’t think that the commonly approach HOG + SVM will achieve good results. By the way, this algorithm seems to be quite robust: https://www.youtube.com/watch?v=OWab2...

Furthermore, If the camera is pointing to the floor at 45 degrees (for example) I think that HOG + SVM approach could be applied. Something like this:

So, any ideas or thoughts about the position of the camera, resolution, algorithms … in order to develop a robust people counting system?

Thanks in advance

EDIT: similar question here: http://answers.opencv.org/question/22... but two years ago and no answers.