In　order　to　make　full　use　of　the　massive　vehicle　networking　data　and　dig　out　the　characteristics　of　the　vehicle　driving　behaviors,　the　analysis　of　the　massive　vehicle　networking　data　is　indispensable.　And　clustering　analysis　is　an　effective　way　to　extract　the　information　which　can　guide　and　supervise　the　vehicles.　In　this　paper,　we　propose　a　novel　initial　clustering　center　selection　algorithm　based　on　the　image　recognition.　In　addition,　in　order　to　shorter　the　time　of　the　clustering　algorithm,　we　improve　the　-Link　based　on　the　massive　vehicle　networking　data.　We　refer　to　image-based　clustering　algorithm　applied　on　matching　map　of　the　vehicle　networking　data　and　ultimately　obtain　the　traffic　hot　spot　map　of　the　urban　roads　by　the　clustering　analysis.　We　evaluate　our　algorithm　by　doing　thorough　simulation　on　the　massive　data　of　vehicles　in　Fuzhou　city.　The　result　shows　that　when　compared　to　the　traditional　clustering　algorithms　with　great　randomness　in　selection,　the　image-based　clustering　algorithm　consumes　less　time　in　the　case　of　massive　vehicle　networking　data.　And　through　the　analysis　of　the　trajectory　data,　the　traffic　hot　spot　map　can　be　formed　and　practically　reflects　the　practical　traffic　flow　of　Fuzhou　city.　©　2016　IEEE.