Implementation And Verification Of Position And Attitude Algorithm For UAV Based On Vision

The classic GPS/INS combined navigation system cannot be able to provide accurate navigationinformation for the UAV autonomous flight in all cases, so we attach great importance to computervision which has been considered as the secondary navigation technology for unmanned helicopterautonomous flight control and become the research priority. In this paper, the author regardsunmanned helicopter visual landing as the research object, designs the image processing algorithmand pose estimation algorithm and completes the estimation of position and attitude of unmannedhelicopter.Firstly, the thesis introduces the principle of unmanned helicopter visual landing, designs alanding target, analyzes the intrinsic parameters difference of commonly used camera models andprovides the unmanned helicopter position and attitude representation using camera perspectiveprojection relationship.Secondly, for the characteristics of landing target image, the paper gives the image processingalgorithm based on point features. A CCD camera is used to get the video image, then the landingtarget and background is separated during image preprocessing step. Four square’s edges of landingtarget can be detected with Canny algorithm, after a corner detection algorithm based on global andlocal curvature properties, the characterized corners can be marked in counterclockwise order.In the end, the camera intrinsic parameters are obtained through calibration, the linear positionand attitude estimation algorithm is designed according to camera perspective projection relationshipand dislinearized.Unmanned helicopter landing status monitoring software is designed by Matlab/GUIdevelopment tools which can realize estimation algorithms. In order to visually observe the motionstate of unmanned helicopter, the three-dimensional visualization simulation can be implementedusing FlightGear flight simulators which is regarded as the visual system. The Unmanned helicoptermovement is also simulated using Ground trolley, the simulation result shows that the imageprocessing algorithm, position and attitude estimation algorithm could be used for vision-basedunmanned helicopter landing.