In environments where Global Navigation Satellite Systems are denied, a common solution to recover one’s position on the Earth is to use stars as inertial references, as it was done centuries ago by navigators using a sextant. Nowadays, sextants are replaced by star-sighting devices composed of inertial sensors, precise clocks and one or more star sensors, combining the short-term precision of inertial navigation techniques and the long-term precision of celestial ones. In this context, this paper aims at developing a star classifier for geopositionning purposes, i.e. a way to discriminate stars in the sky so that an observer can choose the stars that would provide the most precise estimate of its position regarding the sighting performances of the device used (sensor definition, precision of the inertial sensor, etc.). The star classifier proposed in this paper is based on differential calculations and spherical trigonometry, and leads to closed-form expressions which are easily embeddable to evaluate the potential of a star. These closed-form expressions are then validated on an experimental setup.