Runway Recognition And Tracking Based On Image Sequence Of Fixed-wing UAV Autonomous Landing

UAV autonomy technology develops with the expanding application requirements of UAV in themilitary and civil field. The autonomous landing of UAV relates to the safety of its landing recycling,become the key of modern UAV. Vision technology is regarded as an effective means of autonomousUAV landing, due to its passivity and full autonomy. Then the vision-based autonomous landingtechnique becomes a hotspot. Owing to the high speed and long distance of fixed-wing UAV landing,it is a technical challenge of vision-based autonomous landing for the fixed-wing UAV. Therecognition and tracking of landing runway becomes the key. This paper aim at the runwayrecognition and tracking method, which based on the image sequence acquired by the airborne cameraof fixed-wing UAV.Firstly, the landing process is divided into traditional method and visual navigation, positioningmethod, through the analysis of a large fixed-wing UAV landing process and research status of visiontechnology research. The working conditions of visual estimate system is cleared, we put forward thescheme of the system structure, and discuss the software and hardware composition.Secondly, the multi-resolution image technology is studied, then, the image resolution is changedto reduce the amount of calculation in image processing. Research on the recognition and tracking ofimage based on compressed sensing theory, and analyze the real time and accuracy of algorithm. Therunway edge line extraction algorithm based on LSD is also discussed, include the applicableconditions of the algorithm. And the estimation of relative yaw angle of UAV.Thirdly, an offline flight test research. The runway recognition and tracking algorithm is verified,depending on the landing flight video get by the front view camera, with a small fixed wing modelaircraft. Develop and design the visual simulation software based on PC, in order to display thesimulation results of the algorithm, and the verification based on DSP is involved. The results showthat the algorithm is feasible.Finally, summarize and outlook the research work.