Compressive　Sensing　(CS)　is　a　promising　approach　to　compress　the　data　with　spatial-temporal　correlation,　so　as　to　reduce　the　communication　cost　in　Vehicular　Sensor　Network(VSN).　However,　the　current　research　on　the　application　of　CS　in　the　VSN　does　not　consider　dynamic　changes　in　data　sparsity　and　vehicle　distribution,　which　may　lead　to　unacceptable　reconstruction　accuracy.　In　order　to　ensure　the　accuracy　of　data　reconstruction,　this　paper　first　analyses　the　factors　that　affect　the　choice　of　measurement　quantity　in　VSN.　Then,　due　to　the　real-time　changes　in　the　data　sparsity　and　vehicle　distribution,　a　dynamic　measurement　policy　for　VSN　based　on　CS　is　proposed,　that　can　adjust　the　number　of　measurements　according　to　the　real-time　data　sparsity　and　vehicle　distribution.　Through　the　adjustment　of　measurement　quantity,　the　accuracy　of　reconstruction　is　improved　to　achieve　higher　quality　data　communication.　The　experiment　shows　that　the　proposed　dynamic　measurement　policy　improves　the　reconstruction　accuracy　by　15.3%　compared　with　the　existing　CS　approach　in　the　VSN.