Aiming　at　the　problem　of　autonomous　navigation　for　Micro　Aerial　Vehicles　(MAVs)　in　unknown　environments,　a　monocular　vision　SVO/INS　integrated　navigation　method　is　proposed.　Firstly,　The　Semi-direct　Visual　Odometry　(SVO)　is　used　as　the　vision　odometry　front-end.　The　initialization　module,　keyframe　selection　criteria　and　tracking　failure　processing　strategy　of　SVO　are　redesigned　to　realize　the　application　of　SVO　with　forward-looking　camera.　Then,　an　error-state　Kalman　filter　is　developed.　The　filter　uses　the　IMU　measurements　to　predict　the　state,　which　would　be　updated　with　the　output　of　SVO.　The　proposed　method　is　evaluated　with　the　public　dataset,　EuRoc.　Experimental　results　show　that　the　proposed　algorithm　can　estimate　the　position,　attitude　and　velocity　of　the　MAVs　as　well　as　the　unknown　scale,　IMU　bias,　gravity　direction.　The　position　estimation　error　within　3min　navigation　is　0.15m,　the　pitch　and　roll　angle　estimation　errors　are　both　less　than　0.5°,　and　the　yaw　angle　estimation　error　is　less　than　1.2°.　The　advantages　of　the　proposed　method　are　the　low　computation　cost　and　real-time　performance,　which　make　it　very　suitable　for　MAVs　navigation　with　limited　onboard　computing　resources.　©　2019,　Editorial　Department　of　Journal　of　Chinese　Inertial　Technology.　All　right　reserved.