The　initial　alignment　is　a　key　technology　for　strapdown　inertial　navigation　system　(SINS),　because　the　performance　of　the　SINS　is　largely　decided　by　the　accuracy　and　rapidity　of　the　alignment　process.　In　this　paper,　a　novel　alignment　method　is　developed　by　combining　the　quaternion　Kalman　filter　with　optimization-based　alignment　idea.　The　process　model　and　measurement　model　of　the　proposed　alignment　method　are　constructed　based　on　the　properties　of　quaternion.　Since　the　correlation　measurement　noise　in　the　measurement　formula,　the　adaptive　filter　of　the　quaternion　Kalman　filter　combined　with　innovation-based　adaptive　estimation　is　developed,　which　calculate　the　filter　gain　matrix　by　the　innovation　covariance　online.　The　simulation　and　experimental　results　demonstrate　that　the　proposed　alignment　method　has　more　accurate　alignment　results　compared　with　the　existing　alignment　methods.