In　this　paper,　we　propose　an　adaptive　on-line　registration　algorithm　based　on　generalized　likelihood　ratio　(GLR)　for　airborne　radars　in　bias　jumps　environment.　The　algorithm　tests　the　measurement　residual　of　the　registration　formula　in　order　to　decide　and　estimate　the　jumps　of　bias.　The　Monte　Carlo　results　show　that,　comparing　with　traditional　registration　algorithm,　the　algorithm　can　estimate　sensor　bias　effectively　in　bias　jumps　environment.　And　the　estimation　errors　are　reduced　remarkably　and　are　close　to　Cramer-Rao　lower　bound.