Abstract: Focusing on the low precision of the altitude sensors for the small unmanned aerial rotorcraft (SUAR) due to the ground-effect and external disturbances in the process of taking off and landing, a method based on monocular vision is proposed to get altitude information. With the improved Ostu method and the affine invariant moments, the system can realize the high precision target recognition. Furthermore, to deal with the problems of the image noise, the measurement error for the barometric sensor and GPS (global positioning system), and the limited measurement range of ultrasonic sensor, an adaptive Kalman method based on residual error is proposed to fuse the data of altitude information from vision system, barometric altimeter, GPS and ultrasonic sensors. Thus, the SUAR can get altitude information of high precision in whole measuring range. Finally, the effectiveness of the proposed method is tested by the static test, hovering flight test and automatic landing test.