Abstract: Simultaneous localization and mapping (SLAM) is a key technology for unmanned aerial vehicle (UAV) autonomous flight and navigation. This paper presents a visual localization and environment mapping method for micro UAV. In the method, the point-plane ICP point cloud matching algorithm is used to achieve visual localization based on RGB-D sensor, the parallel computing acceleration is applied to meet the real-time requirements of UAV control, then the truncated signed distance function (TSDF) algorithm is used for multi-frame point cloud data fusion to achieve environment model reconstruction for an unknown environment, and at last the fusion of visual SLAM system and UAV IMU sensor data is applied to enhance the precision of autonomous localization and model reconstruction. Finally, we set up a micro UAV system with full visual SLAM algorithm and test the whole system in a typical indoor environment. The result shows a 0.092 m positioning deviation and 60 Hz update rate, capable for real-time UAV controlling accuracy requirement, thus verifying the effectiveness of the method.