Research On Indoor Localization Of Unmanned Aerial Vehicle Based On Vision And Inertial Navigation

The prerequisite for UAV to achieve autonomous flight function is good positioning ability.Real-time calculation of UAV motion state and accurate pose estimation are the basis for UAV to achieve stable navigation and other complex tasks.On the other hand,Global Position System(GPS)is the main tool for an UAV to locate its position in the outdoor environment.But when UAV flies indoors,since the GPS signal strength is usually weak or even missing,the GPS positioning method can not be appliedas in outdoor flight positioning.To deal with the above problems,this paper proposes a vision and inertial navigation fusion method to achieve the autonomous indoor positioning function of UAV.The main contents of this paper are summarized as follows:First,the UAV indoor positioning system is modeled in this paper.The motion of a rigid body in three-dimensional space consists of one rotation and one translation.Unmanned aerial vehicle(UAV)can be regarded as a rigid body in three-dimensional space,so the task is not only to determine the position of UAV,but also to determine the rotation of UAV relative to the inertial coordinate system when positioning UAV.After the rotation description and transformation matrix analysis of UAV,this paper introduces Lie group Lie algebra to describe the pose of UAV,and derive the Jacobian matrix of transformation coordinates relative to transformation matrix through Lie algebra perturbation model,which lays a necessary theoretical foundation for UAV positioning.In the next work on visually determining the positionand the attitude of the UAV,we use depth vision sensors,which can measure one-to-one corresponding color map and depth map according to UAV flying state.The proposed method can process these color map and depth map on image level and local map level.In the procedure of calculating UAV pose by vision,this paper develops the feature point method in detail,and analyzes the traditional direct method in detail.After analyzing the advantages and disadvantages of the two methods,the improved direct method is put forward,and the improved direct method is verified by the data set.In the final result of visual inertial fusion to determine the pose of UAV,this paper uses the extended Kalman filter algorithm based on Lie algebra.This chapter first introduces the measurement model,motion model and pre-integration model of the inertial measurement unit,then introduces the extended Kalman filter algorithm.Finally,we design an extended Kalman filter based on Lie algebra based on the pre-integration of the inertial measurement unit and the observations calculated by the visual method to fuse the visual data with the inertial navigation data to realize the autonomous indoor positioning of the UAVthen the algorithm was verified the data setexperimentally.