Aiming　at　the　poor　performance　of　convergence　of　ant　colony　optimization　(ACO),　in　this　paper,　a　novel　pheromone　initialization　strategy　of　ACO　for　the　traveling　salesman　problem　(TSP)　is　put　forward.　More　precisely,　the　pheromone　matrix　of　a　specific　ACO　algorithm　is　initialized　by　the　minimal　spanning　tree　(MST)　information.　Simulation　results　demonstrate　that　the　proposed　strategy　could　improve　the　convergence　performance　of　ACO　both　in　term　of　quality　of　solution　and　speed　of　convergence.