Research And Implementation Of UAV Intelligent Landing System Based On Vision-assisted Positioning

At present,the autonomous landing of UAV generally uses GPS to achieve positioning,but GPS signals are susceptible to interference when landing to low altitude.In recent years,with the rapid development of computer vision,artificial intelligence and microprocessors,the use of visual information acquired by airborne cameras to assist in autonomous landing of UAV has become a trend of intelligent UAV development.In this paper,multi-rotor UAV is used as the experimental platform,and the images collected by the airborne camera are used as detection objects.A combination of image processing,object detection,and flight control technologies is used to propose an intelligent landing system based on vision-assisted positioning.The dissertation analyzes many problems such as the landmark image being judged by bad lighting,the landmark detection effect is limited,and the accumulated pose error is large during the landing process of UAV based on visual aid.Aiming at the impact of poor lighting on the landmark image quality,an image enhancement algorithm is proposed.According to the different conditions of overexposure and underexposure,the image adaptive algorithm and the image low-illumination algorithm are used for image preprocessing respectively,which effectively reduces the UAV The visual positioning error meets the needs of positioning,and the algorithm processing speed meets the real-time requirements of the UAV.In terms of landmark detection,an improved PC-YOLOv3 image recognition algorithm is proposed,which improves the running speed of the algorithm by trimming less important feature channel branches.Space issues.The experimental verification shows that the improved PC-YOLOv3 algorithm only needs less trainable parameters and lower computational overhead,which makes the recognition accuracy and processing speed of the UAV reach a balance.On the basis of visually assisted positioning,Kalman filtering was used to correct the UAV posture parameters,and finally the center coordinates of the landmark were sent to the flight controller to achieve the intelligent landing of the UAV.