Research On Vision-based Autonomous Landing System Of UAV

In recent years,UAV control technology has become increasingly mature.Because of its low cost,easy control,and flexible flying capabilities,the quadrotor has been widely used in aerial photography,plant protection,electric power inspection,and field reconnaissance.The demand for autonomous flight of UAVs has attracted wide attention,and UAV autonomous landing technology is an important prerequisite for autonomous flight.Visual sensors are characterized by their low cost and high flexibility.Applying computer vision technology and visual feedback to the control loop of the autonomous system like UAVs is of great significance for the development of its intelligent control.This subject comes from the air-ground coordination project about UAVs.This subject is in order to realize the replenishment and recovery of UAVs during the process of going out,and to achieve the autonomous return charge,so that we can realize the goal of continuous autonomous operation.The project requires the use of visual sensors as environmental sensors and uses an embedded system to achieve autonomous landing of quadrotors on moving targets.The main research contents of the paper are as follows:Firstly,a visually assisted autonomous landing system is designed.Firstly,the form without ground station assistance is determined.Based on the calculation requirements,the onboard computer and flight control processor selection are determined.In order to improve the accuracy of visual positioning and the working radius,the appropriate airborne cameras and their associated dual-axis cradle head are selected.Secondly,a marker identification algorithm based on binary coding is designed.Several mainstream algorithms for identifying and locating with the aid of QR code identifiers is compared,and visual positioning methods based on Ar Uco algorithm is determined.An algorithm for automatically generating a marker dictionary is deduced,which reduces environmental interference and the probability of confusion of markers in the dictionary.This algorithm is universal and can be used in other binary coded positioning systems.Then,the problem of navigation and positioning in the autonomous landing process of UAVs is studied.The linear model of the pinhole camera is introduced,the relative relationship between the marker coordinate system and the camera coordinate system is deduced,and the program of visual positioning is written.The workflow is introduced.The algorithm compensates the problem of the mechanical vibration of the UAV during flight is designed.And the accuracy of visual positioning is verified.Finally,the autonomous landing control algorithm is studied and tested.According to the position information,the velocity information of the target is deduced,and the Kalman filter method is used to predict the target 's motion track when the target is temporarily covered by the environment.It also reduces the calculation amount of visual recognition.Finally,the autonomous landing test is conducted under the condition of airborne calculation.Both the visual positioning and the UAV control achieves high accuracy,and finally the autonomous landing for the mobile platform is achieved.