Design And Realization Of Autonomous Landing System For Rotor UAV Based On Machine Vision

With the vigorous development of UAV technology,computer vision technology and artificial intelligence technology,rotor UAV is more and more widely used in various civil and military fields.At the same time,higher requirements are put forward for the autonomy and intellectualization of UAV.The autonomous landing technology of rotor UAV is one of the key technologies to realize its intellectualization.Therefore,vision-based autonomous landing technology of rotor UAV has become a research hotspot in the field of UAV in recent years.Compared with traditional navigation methods,vision-based autonomous landing has the advantages of high accuracy,low cost and strong anti-jamming ability,which can better realize the autonomous landing of UAV.In this paper,the research of autonomous landing technology of quadrotor UAV is carried out on the basis of ground cooperative target on the research platform of quadrotor UAV.In order to quickly and accurately identify ground cooperative targets,a landing mark composed of a triangle and a concentric ring is designed.Based on this,a quadrotor UAV autonomous landing algorithm is proposed.According to the flight altitude of UAV,the algorithm is divided into two stages.In the first stage,the position parameters of UAV relative to landing mark center are calculated by identifying the regular triangle.In the second stage,the position deviation and yaw angle between the UAV and the landing mark center are determined by the concentric circle and its internal course reference line,thus the autonomous landing of the UAV is completed.Because the triangular landing mark is relatively simple and the feature points are few,the position information and yaw angle of UAV can only be calculated,but the attitude data of 6 degrees of freedom can not be obtained.Therefore,according to the geometric characteristics of the standard helicopter apron,an attitude estimation method of UAV based on mark detection is proposed.Firstly,five-step mark extraction algorithm is used to extract visual marks from the images captured by airborne cameras.Secondly,the proposed corner detection algorithm based on distance three-point method is applied to obtain 12 corner points of H-shaped mark.By classifying and numbering the corners and matching them with the corresponding corners in the reference image,the homography matrix containing relative attitude information can be calculated.Finally,the attitude angle of UAV is obtained by using direct linear transform to decompose homography matrix,and the position of UAV relative to visual mark is calculated according to the principle of camera imaging.Although the attitude estimation algorithm based on the standard apron,which can calculate the 6-DOF attitude data of UAV,and the estimation error is small,can meet the requirements of autonomous landing of UAV.However,when the initial landing height of UAV is high,especially after more than 5 meters,landing mark can not be detected accurately,which affects the safe landing of UAV.For this reason,AdaBoost cascade classifier is further trained with LBP and Haar features,and the trained classifier is used to recognize landing marks in the scene when the UAV is at a higher altitude.Thus,the initial landing altitude of the UAV is increasing,and the practicability of the algorithm is improved.Aiming at the above visual attitude estimation algorithm proposed during the autonomous landing of quadrotor UAV,simulation experiment platform and quadrotor UAV experiment platform are built respectively.The validity and accuracy of the proposed algorithm are verified and analyzed through simulation experiments.The experimental results show that the proposed attitude estimation algorithm can meet the requirements of autonomous landing of UAV.Finally,the visual landing effect of quadrotor UAV is tested experimentally.