The　in-motion　alignment　for　the　underwater　strapdown　inertial　navigation　system　is　still　a　challenging　problem　due　to　various　disturbances　in　the　underwater　environment.　In　this　paper,　a　novel　in-motion　alignment　method,　based　on　the　Lie　group　representation,　is　developed.　In　this　method,　the　process　model　is　rewritten　using　the　Lie　group　of　the　constant　attitude　matrix　between　two　inertial　frames　as　the　state.　An　exact　linear　measurement　model　is　constructed　by　analyzing　the　effect　of　the　sensor　errors　in　calculating　the　velocity　vector.　Next,　the　state-dependent　Lie　group　filter　is　designed　basing　on　accurate　derivation　expressions　for　the　covariance　matrices　of　state-dependent　noises.　The　simulation　and　experiment　results　demonstrate　that　the　proposed　method　can　achieve　better　alignment　accuracy　and　time　than　the　existing　method.　The　accuracy　improves　by　70%　with　the　quaternion　Kalman　filter.