As　the　only　sign　to　record　the　basic　parameter　information　of　transmission　line,　utility　pole　signboard　plays　an　important　role　in　the　process　of　circuit　statistics,　management　and　inspection,　and　has　a　very　high　research　value.　At　present,　the　data　acquisition　of　utility　pole　signboard　mainly　depends　on　manual　work.　Due　to　the　influence　of　environment,　equipment　and　other　factors,　the　angle　of　utility　pole　signboard　in　the　image　will　be　inclined,　which　will　reduce　the　detection　accuracy　of　utility　pole　signboard.　Through　literature　research,　it　is　found　that　there　are　few　academic　researches　on　the　detection　and　identification　of　such　utility　pole　signboards,　while　there　are　many　researches　and　practical　solutions　in　license　plate　detection　and　recognition.　This　paper　proposes　a　new　detection　algorithm　based　on　Faster　R-CNN　model　and　ResNet-50　convolutional　neural　network,　which　improves　the　accuracy　by　increasing　its　depth　to　50　layers,　and　improves　the　generalization　ability　of　the　algorithm　combined　with　multi-scale　training　to　achieve　the　global　optimal　detection　effects.　The　experimental　results　show　that　the　proposed　algorithm　has　good　detection　effects,　and　the　detection　accuracy　is　87.6%,　which　verifies　the　feasibility　of　pole　signboard　detection.　©　2020　IEEE.