The　problem　of　estimation　of　a　semiparametric　errors-in-variables　functional　relationship　model　under　longitudinal　data　is　studied.　By　using　a　nonparametric　kernel　weight　function　and　total　least　square　method,　the　estimators　of　unknown　parameters　and　functions　are　constructed.　Strong　convergency　of　the　estimators　is　investigated.　The　performance　of　the　simulation　is　encouraging.　The　results　obtained　in　the　paper　can　be　used　to　deal　with　the　data　processing　in　the　fields　of　biological　engineering,　financial　analysis　and　network　engineering.