A technique for a vision application is proposed by means of which it is possible to obtain the three dimensional reconstruction of surfaces and objects. This technique uses a number of images from a single camera that is moved around the object by a robot arm. In this way the calibration of the video system is no more necessary because the robot kinematics is known. The images are processed by an algorithm that has been developed starting from our previous investigations on computer vision applied to robotics. The reconstruction of the images can be easily used to reproduce the object itself by means of tools installed on a robot arm. Some examples are given.