The　urban　polycentric　structure　is　connected　to　the　economy　and　enormously　impacts　socioeconomic　development　and　policies.　Unlike　traffic　data　and　big　geographic　data,　remote　sensing　data　have　shown　an　accessible　way　to　measure　urban　spatial　interaction.　However,　most　existing　studies　only　focused　on　the　interaction　among　cities　rather　than　within　cities.　Meanwhile,　the　urban　spatial　interaction,　which　should　be　directional,　was　always　expressed　as　an　undirected　graph.　Therefore,　this　article　developed　a　network-based　radiation　model　using　nighttime　light　remote　sensing　data　and　mapped　a　directed　interaction　network　(inward　and　outward　direction)　among　urban　centers.　Taking　the　region　within　the　outer　ring　of　Shanghai　as　an　example,　the　taxi　trajectory　data　were　adopted　to　validate　the　result　with　the　R-2　of　0.61.　We　discovered　that:　the　urban　polycentric　interaction　network　is　dumbbell-shaped　with　an　east-west　development　corridor　crossing　the　main　center　and　connecting　two　main　urban　center　clusters.　The　in-strength　and　out-strength　interaction　of　each　urban　center　have　a　similar　distribution.　The　urban　centers　with　higher　in-strength　and　out-strength　are　mainly　concentrated　toward　the　main　center,　especially　in　the　east-west　direction.　At　the　urban　center　level,　the　total　inward　interaction　is　slightly　higher　than　the　total　outward　interaction　of　most　urban　centers.　Spatially,　an　unbalanced　distribution　was　found.　In　summary,　our　proposed　method　effectively　indicates　the　urban　polycentric　interaction　and　is　applicable　to　other　regions　since　it　requires　no　arbitrary　parameters　and　the　input　data　(e.g.,　nighttime　light　data)　is　readily　available.