This paper addresses the problem of partially captured objects within the image scene that is acquired from an eye-in-hand robot vision environment. Two methods were proposed to locate full object view. The first is based on a logically oriented programming whereas the second implements neural networks. The aim is to assure that the object is completely visible within the camera field of view. Experimental rig and computer programs are developed employing both methods and to train the neural networks. Both methods showed validity to locate the missing parts of the object in the scene and to feedback the required movements to the robot arm to achieve the goal, though to overcome a dead zone case of moving the camera in forward and backward directions in a closed repeated loop manner, a developed strategy is suggested and showed success.
Robot vision, Object tracking, Neural networks
