In　order　to　improve　the　real　time　and　precision　of　license　plate　detection　in　the　license　plate　recognition　system,　this　paper　presents　an　improved　license　plate　detection　algorithm.　In　view　of　the　existing　problems　in　the　license　plate　detection,　such　as:　the　feature　extraction　method　is　cumbersome　and　inefficient,　the　accuracy　is　easily　affected　by　external　factors　such　as　weather,　light　and　shooting　angle.　In　this　paper,　the　target　detection　algorithm　is　used　to　deeply　study　the　license　plate　detection　technology.　Based　on　the　SSD　algorithm,　the　residual　network　is　taken　as　the　basic　network,　and　several　parameters　are　trained　on　the　license　plate　data　set　to　obtain　the　appropriate　hyper　parameters,　which　improves　the　accuracy　of　small　target　detection.　Experiments　show　that　the　proposed　method　can　more　efficiently　extract　license　plate　features　and　effectively　improve　the　license　plate　detection　rate.