Is there an image processing library / AR technology suitable for recognizing objects in a known parameter?

I need to detect objects in a scene (on iPhone and Android). The environment is limited so that the task is simpler and more accurate:

• the environment is small and well-known ... users study one room or a small open area, which I can do in advance to "train" or restrain the algorithm.
• the user's location in space is often limited ... even when the space is large, the user can be limited to certain paths in space
• detected objects are relatively static ... they are part of the environment and do not move

BUT, making the problem harder:

• I cannot change the environment by putting markers on objects, so I need to recognize the objects themselves
• The objects look pretty similar, so we may need to use the surrounding scenes as input, not just individual elements

For example, imagine a walk through a historic cemetery along a path (you are not allowed to walk on the grass). When the user points his phone to the headstone, I would like to be able to identify the headstone and evaluate where the user is relative to the headstone (so I can estimate the user's location on the way). Many of the tombstones are quite similar if you look only at the tombstone. Ahead I can go this way and take a few shots of objects from different angles.

, ?