Vision Based Autonomous Landing And Pose Estimation Of Four-rotor Aircraft

Aircraft landing is an important stage during the flight of aircraft, landing control has becomea research focus in the field of unmanned aircraft based on vision in recent years. In the process oflanding control, the position and attitude estimation algorithm using image features will directlyaffect the control precision of UAV. This paper takes the four-rotor aircraft as a flight platformand carries through a research on visual landing and position estimation of UAV base on vision.Firstly, this paper introduces the principle of aircraft visual landing, designed “H” typelanding image. Considering about the fact that the camera is fixedly connected to the vehicle,using the pinhole perspective projection model and the world coordinate system to the bodycoordinate system, get the vision based aircraft pose estimation model.Secondly, according to the characteristics of land marks, designed image processingalgorithm to detect corner feature. After the flow of threshold segmentation, median filter, cornerdetection, we can get the feature points in the image which is mapped from landing mark featurepoint. Also designed the algorithm of pose estimation based on the information of feature pointcoordinates, and takes it as the target tracking control input. In order to verify the validity of imageprocessing algorithms, built the image processing algorithm simulation verification platform basedon Visual Studio.Finally, vertical camera on the AR.Drone spacecraft is calibrated to get the camera intrinsicmatrix. Prepare the ground control program using C#language on the Visual Studio platform, theprogram mainly realizes the image processing algorithm, pose estimation algorithm and targettracking algorithm. Carried through the autonomous landing experiment for four-rotor aircraft, theresults show that the algorithm can be used for visual aircraft landing.