Object anchoring is a very important and useful concept, especially for robotic systems which use two different levels to represent objects, symbolic and sub-symbolic. The right sub-symbolic information acquired from sensors have to be combined with the symbolic description that points to the same physical object which is often a time-consuming and error-prone procedure, e.g. due to unsuccessful detection using machine vision. In this paper an extension of this anchoring concept is given using relative anchors which presents an intelligent object representation and allows the reduction of necessary machine vision operations. Also the inaccuracies given by the algorithms used for detection are taken into account and used to enhance the object anchoring. The concept is evaluated on a robotic system in a library scenario and the benefit is presented.