Research On High-precision Attitude Estimation Based On INS/Vision Fusion For Indoor Flight

Compared with conventional unmanned aerial vehicle (UAV), micro aerial vehicle (MAV) hasmany advantages like small size, low cost, flexibility and so on. It has broad application prospects inboth the military and civilian areas. However, due to the constraints of size, weight and cost, thetraditional sensors on MAV can’t give high-precision attitude information. This paper analyzes thestrengths and weaknesses of the current methods of attitude measurement and then does research onhigh-precision attitude estimation based on INS/Vision fusion for indoor flight.First, an attitude estimation method based on vision techniques is implemented, which uses theimage coordinates and the coordinates of world coordinate system of at least three non-collinearfeature points to calculate the attitude that camera coordinate system relative to the world coordinatesystem. After analyzing the error of INS and visual attitude estimation, the attitude estimation basedon INS/Vision fusion is adopted.Secondly, according to the different sampling rate of INS and vision, an algorithm of multi-rateKalman fusion is researched, which divides the updated process into two separate processes those aretime update and measurement update. When there is no visual attitude information, only the timeupdate is executed. In the conditions of visual attitude information, the time update and measurementupdate are executed. Experimental results show that this algorithm has better convergence effect andhigher data update frequency.Finally, a physical simulation system is builded to verify the visual attitude estimation algorithmand multi-rate Kalman fusion algorithm. Many issues of the visual attitude measurement system likecamera calibration, feature point extraction and so on are solved. Experimental results show that theattitude estimation based on INS/Vision fusion is feasible to apply.