Abstract: Computer Vision plays a great role in enhancing robot′s intelligence and ability to overcome uncertainties in unknown environment.Since spacial telerobot works in a hazardous and uncertain environment, a method is proposed to scale and position unknown objects based on eye in hand vision which can help the telerobot estimate whether it can grasp the unknown object, and determine where and how to grasp.In repositioning tasks, it can locate the object fast by using their prior information and avoid moving the manipulator frequently. The theory on which the method is based is addressed and then the method is validated by an experiment and the error is analyzed. The result confirms that this method can achieve high accuracy.