How to capture an object, not by coordinates but by the object itself when recording a Keyword test?

Can anyone help me, how to capture an object, not by co-ordinates but by the object itself? thank you