Characteristics　of　plaque　are　extracted　by　the　gray　level　co-occurrence　matrix.　The　four　effective　characteristic　values　including　energy,　entropy,　moment　of　inertia　and　correlation　are　selected　to　compose　the　eigenvector.　And　then　support　vector　machine　(SVM)　is　applied　to　construct　the　classifier　combining　particle　swarm　optimization　(PSO)　algorithm.　The　parameters　of　SVM　are　optimized　based　on　Gaussian　radius　basis　kernel　function.　The　result　shows　that　the　method　spends　less　time　and　the　average　recognition　accuracy　rate　of　the　four　common　plaques　reaches　92%.　It　verifies　the　effectiveness　of　the　proposed　method.