Recent　studies　have　shown　that　the　synaptic　plasticity　induced　by　development　and　learning　can　promote　the　formation　of　multiple　synapse.　With　the　rapid　development　of　electron　microscopy　(EM)　technology,　we　can　closely　observe　the　multiple　synapse　structure　with　high　resolution.　Although　the　multiple　synapse　has　been　widely　researched　by　recent　researchers,　the　classification　accuracy　for　the　type　of　multiple　synapse　has　not　been　documented.　In　this　paper,　we　propose　an　effective　automatic　classification　method　for　the　type　of　multiple　synapse.　The　main　steps　are　summarized　as　three　parts:　synaptic　cleft　segmentation,　vesicle　band　segmentation,　multiple　synapse　classification.　The　experiments　on　four　datasets　demonstrate　that　the　proposed　method　can　reach　an　average　accuracy　about　97%.