Planet–sun sensor revisited

Since the seminal work of Daniele Mortari (“Moon-Sun Attitude Sensor,” Journal of Spacecraft and Rockets, Vol. 34, No. 3, 1997, pp. 360–364), the concept of an attitude sensor using images of illuminated celestial bodies has been pushed forward through the years. The basic idea consists of extracting two independent directions from the image of a celestial body, namely, the camera-to-planet and the planet-to-sun directions. The former is estimated from the center of an ellipse fitted to the imaged limb points and the latter from the symmetry axis of the illuminated region. These assumptions, however, only hold for far-distant spherical targets. In this work, the problem is reformulated in the framework of projective camera transformations of quadrics and conics, and an algorithm estimating the line of sight to the planet and the illumination direction from the limb and terminator ellipses, respectively, is presented. The method is applicable to any ellipsoidlike celestial body having known orientation. The algorithm is first validated on synthetically generated images and then tested using real pictures of Dione and Enceladus satellites gathered from Cassini spacecraft. Results show that the sensor concept returns rms errors in the order of the angular width of a pixel in computing the nadir direction, and subdegree accuracy in computing the sun direction.