For optical observations of Earth-orbiting objects, multiple observations must be combined to determine the orbit of the observed object. Whether two arbitrary tracks are of the same object, however, is generally unknown a priori, and solving this problem with traditional approaches requires iterative procedures that do not guarantee convergence. This paper proposes a technique of correlating multiple optical observations using highly constrained probability distributions in Poincaré orbit element space. These distributions are divided into subregions, each of which are mapped linearly, to reduce computational burden, but without significantly losing accuracy. A proof-ofconcept implementation, in which simulated observations were processed, performed well. The technique proposed serves as a solution technique for initial orbit determination given two precise optical observation tracks.