When the signal of Global Positioning System (GPS) is weak or rejected,the UAV cant implement autonomous navigation.To solve the problem,an improved ORB-LK optical flow algorithm is proposed to estimate the velocity of UAVs.Firstly,the feature points on the object are extracted by using the ORB algorithm,and the optical flow velocity of the image is calculated by using the LK optical flow method.Then,the forward/backward matching algorithm is used to determine matched pairs of the points,and the PROSAC algorithm is used to eliminate the mismatched pairs,so as to realize fine matching.Finally,the UAV rotation velocity measured by Inertial Measurement Unit (IMU) and the height measured by the sonar are used to calculate the translation velocity,so as to provide velocity information for the autonomous navigation of UAVs.Simulation experiments with Matlab show that the average operation time of the improved algorithm in the public image data set is 0.622 s,which meets the requirements of real-time application of the UAV.Compared with the original ORB-LK algorithm,the improved algorithm greatly improves the matching accuracy,and can estimate the velocity accurately.