An　improved　CNN　neural　network　based　on　GRU　sequence　model　license　plate　recognition　technology　is　mentioned　in　this　article.　The　network　realizes　the　recognition　of　the　7-digit　string　of　Chinese　blue　license　plates　and　yellow　license　plates,　and　has　the　advantages　of　higher　recognition　accuracy　and　faster　recognition　speed.　The　specially　labeled　license　plate　data　set　is　input　into　the　built　neural　network　for　training　and　the　model　is　obtained.　Interact　with　the　front-end　user　through　the　GUI　interface　to　obtain　the　recognition　result.　This　paper　specifically　analyzes　the　license　plate　location　detection　method,　describes　in　detail　the　CNN　neural　network　construction　process,　and　the　trained　neural　network　model　analyzes　the　results　of　license　plate　recognition,　and　finally　summarizes　and　looks　forward　to　the　improvement　of　the　recognition　system.　©　2022　IEEE.