Optical tracklet association with a defined probabilistic correlation measure

Optical tracklet association is a probabilistic problem in nature due to the uncertainties of angular observations and orbital propagation. This study focuses on the short tracklet association in space-based optical space surveillance scenarios. A probabilistic tracklet association approach incorporating angles-only initial orbit determination and nonlinear orbital uncertainty propagation is proposed. The probabilistic correlation measure is defined as the likelihood score between two probability density functions describing orbital uncertainties. This measure provides a quantitative assessment on whether two tracklets are correlated and does not hinge upon Gaussian uncertainty assumption. State transition tensors for modified equinoctial elements under spherical harmonic gravity field are derived to achieve accurate nonlinear orbital uncertainty propagation. The capability of the developed probabilistic tracklet association approach is demonstrated by both numerical simulations and real data tests. 99.45 % TP rate and 0.09 % FP rate are achieved in the simulated scenario covering 30 LEO space objects and 254 optical tracklets. Tracklets from satellites in the same orbit (i.e., constellation tracklets) are also processed with the developed method.