The invention provides a method and system for obtaining a grading partition tree of dual-reverse furthest neighbors on a road network. The method includes the steps of, as for the road network G and a certain inquiry set Q, building a furthest Voronoi graph of the road network G with respect to the inquiry set Q, defining a furthest Voronoi region of a certain inquiry node q, belonging to the Q, on the furthest Voronoi graph to be the nodes fvc (q,Q), enabling the equation fn(p,Q)=q to be met, namely enabling all the nodes p contained by the fvc(q,Q) to regard the q as the furthest neighbors of the nodes p relative to the Q, enabling BRFN(q, Q, VG) to equal the fvc(q,Q), in order to obtain the fvc(q,Q), firstly building a set S containing potential solutions of all nodes VG on the road network G, every time taking out a node q' from the other nodes of the Q, dividing the S into two parts according to the distance between each potential solution in the set S and the q and the distance between each potential solution in the set S and the q', deleting the part close to the inquiry node q from the S until all the other nodes q' of the Q are taken out, and enabling the final part which is not deleted in the furthest Voronoi graph to be the fvc(q,Q), wherein the potential solutions are certain nodes on the road network G, and single reverse neighbors of an inquired node can be rapidly searched on the road network.