In this paper, we propose an object recognition system which integrates features of an object using sequential images. Recently, it has been recognized that an active vision based on the control of camera position is a hopeful approaches to realize a robust object recognition system. Our system integrates features of the unknown object extracted from sequential images while the camera moves to the best position to recognize the object. Integration of features promises a reliable recognition at early step of camera control. The experimental results show the proposed approach promises a reliable and high speed object recognition.