Development of an Algorithm for Absolute Visual Navigation of an Unmanned Aerial Vehicle

A hybrid method for visual-inertial navigation of unmanned aerial vehicles is presented. It combines absolute navigation using a pre-downloaded satellite georeferenced map, visual odometry based on optical flow analysis, and inertial measurement unit data. The method utilizes key image feature matching using the SIFT algorithm for both absolute positioning and sparse optical flow computation. Integration of estimates from both approaches using an extended Kalman filter ensures smooth trajectories, compensates for outliers typical of absolute navigation, and eliminates coordinate drift inherent in visual odometry.