In survey series of unknown Earth orbiting objects, no a priori orbital elements are available. In surveys of wide field telescopes possibly many nonresolved object images are present on the single frames of the series. Reliable methods have to be found to associate the object images stemming from the same object with each other, so-called linking. The presence of cosmic ray events, so-called Cosmics, complicates reliable linking of non-resolved images. The tracklets of object images allow to extract exact positions for a first orbit determination. A two step method is used and tested on observation frames of space debris surveys of the ESA Space Debris Telescope, located on Tenerife, Spain: In a first step a cosmic filter is applied in the single observation frames. Four different filter approaches are compared and tested in performance. In a second step, the detected object images are linked on observation series based on the assumption of a linear accelerated movement of the objects over the frame during the series, which is updated with every object image, that could be successfully linked.

An unexpected space debris population has been detected in 2004 Schildknecht et al. (2003, 2004) with the unique properties of a very high area-to-mass ratio (HAMR) Schildknecht et al. (2005a). Ever since it has been tried to investigate the dynamical properties of those objects further. The orbits of those objects are heavily perturbed by the effect of direct radiation pressure. Unknown attitude motion complicates orbit prediction. The area-to-mass ratio of the objects seems to be not stable over time. Only sparse optical data is available for those objects in drift orbits. The current work uses optical observations of five HAMR objects, observed over several years and investigates the variation of their area-to-mass ratio and orbital parameters. A normalized orbit determination setup has been established and validated with two low and two of the high ratio objects, to ensure, that comparable orbits over longer time spans are determined even with sparse optical data.