An indoor quadrotor locating and object-following algorithm using monocular vision

Using quadrotor as an indoor robot asks for accurate locating and control methods. To solve this problem, a composite algorithm combining MIMU(Micro Inertial Measurement Unit) and monocular vision is used. The algorithm uses the measurement of MIMU as the source of position updating, while monocular vision algorithm provides the detection result of the feature of reference lines on the ground. The location of the reference lines detected is then compared with the result of MIMU, and the combination of these two results brings more accurate locating result for quadrotor. Compared to using MIMU, the error can decrease from ±50cm to ±10cm when the height of quadrotor is 1m. An object following platform is used for validating the composite algorithm, and system identification method is used for modeling the AR.Drone. Trajectory following algorithm and object following method are also developed. The scene of an AR.Drone searching at the navigation area and following the UGV is presented at last.