Animprovedself-organizingfeaturemap(SOFM)neuralnetworkispresentedtogeneraterectangularandhexagonallatticwithnormalvectorattachedtoeachvertex.Aftertheneuralnetworkwastrained,thewholescattereddatawerepidedintosub-regionswhereclassifiedcorewererepresentedbytheweightvectorsofneuronsattheoutputlayerofneuralnetwork.Theweightvectorsoftheneuronswereusedtoapproximatethedense3-Dscatteredpoints,sothedensescatteredpointscouldbereducedtoareasonablescale,whilethetopologicalfeatureofthewholescatteredpointswereremained.