Mean-shift clustering approach to the tracklets association with angular measurements of resident space objects

Hundreds of too-short tracklets of angular measurements, which cannot be correlated to known catalogue objects, are usually obtained by an operation optical network every night. Although the proportion (≈1%) of these uncorrelated tracklets (UCTs) in the total collected measurements is not very high, the information of the changes in space situation hidden in them is vital. To group these UCTs for the determination of orbits is very valuable. Currently, most investigations on this topic focus on determining the pairwise relation of the tracklets to check the association. However, obtaining reliable tracklets groups and the related orbits may not be solved solely by the pairwise association methods because the number of the groups and the number of points in each group are unknown. This paper reveals the clustering nature of the tracklets association problem and establishes a mean-shift clustering approach to resolving it by treating tracklets as points, orbits as centres, and residuals as their distances. Using the real measurements from the International Scientific Optical Network (ISON), we verify the correctness and validity of the proposed method. In addition, in response to the critical situation of clustering tracklets for nearby objects with similar orbits, the proposed method works well by adjusting the bandwidths.