Would be advisable to use Haar-classifiers for something else other than human body, face, etc? How to create a custom one with only few images?

Question

I need to detect object that aren't that much complex: pokestops.

Some images of a pokestop:

Screenshot examples:

I have tried different approaches (template matching, homography and feature detection) but sadly I wasn't able to detect correctly a pokestop. Then I remembered that I was able to detect faces using a Haar cascade classifier but of course I already had the .xml file. So, having only few images of an object, would I be able to create a proper xml for Haar classifiers? Maybe there is a better approach to detect this object? Thank you.