We present a low cost acquisition setup for the accurate reconstruction of feet and shoe lasts geometry. It is based on the use of a single close range depth sensor (Primesense Carmine 1.09) that is moved around the object using a PC-controlled mobile robot. The mobile robot is built using a simple electric car constrained to move in a circular guide and controlled by an Arduino Uno board allowing the custom software to control the engine's speed. In the acquisition phase multiple point clouds are captured and registered in a common reference frame. The vehicle trajectory has been designed to maximize the reconstruction accuracy and robustness against changes in the scene illumination. Registered point clouds are filtered in order to remove the floor, reduce the noise and refine the alignment, and are finally merged in an unique simplified cloud that can be triangulated with standard techniques. The resulting foot model is then aligned in a standard way and segmented identifying relevant landmarks for measurement.