Research On Vision-based Navigation For A Fixed-wing UAV In The Process Recycling

With the development of science and technology,UVAs are widely used in civil and military applications,such as reconnaissance,surveillance,environmental monitoring,network connectivity,and information relay.The safe,effective and quick recovery of the fixed-wing UAV can reduce the operating cost of the UAV and improve the efficiency of the UAV.Therefore,various countries have actively conducted research on fixed-wing UAV recycling technology.In this paper,the research on visual navigation is carried out in the context of fixed-wing UAV collision recovery.The recovery of the fixed-wing UAV collision line requires high precision and real-time navigation.To this end,this paper carried out the following research work:According to the imaging principle of the camera,different coordinate s ystems and their corresponding conversion relationships are established.The camera linear imaging model,nonlinear imaging model and weak perspective imaging model are established respectively.The parameter matrix,radial distortion and tangential distortion of the camera are calibrated by Zhang calibration method.Finally,the effectiveness of the algorithm is proved by simulation examples.Then,a cooperative visual logo was designed based on the camera imaging model.In order to make the obtained image accuracy meet the image processing requirements,two sizes of cooperation logos are designed to identify the line cooperation mark when the drone is far away from the cooperation mark,and identify the cooperation mark when the distance is close.At the same time,image preprocessing,image feature extraction and recognition algorithms are studied.Two feature extraction methods are adopted for the two cooperative markers,namely,dot mark recognition detection and line mark recognition detection.By studying the Hough line detection and LSD line detection algorithm,the straight line detection algorithm is improved for the problem that the same line on the LSD line detection easily recognizes two lines.The effectiveness of the algorithm used is p roved by simulation experiments.Afterwards,the pose solving algorithm was studied for the features extracted from the two cooperative signs.In the process of UAVs pose calculation,an accurate orthogonal iterative algorithm using the robust of PnP algorithm as the initial value is proposed for the problem of low precision,low anti-interference ability and weak real-time performance in the DLT method of n-point perspective problem.The simulation experiment is compared with several other pose calculation methods,and the calculation accuracy,real-time and robustness of different methods are compared.At the same time,the robust of PnL algorithm based on line feature n-line perspective problem is studied,and the simulation experiment is carried out to prove that the pose calculation algorithm meets the recovery requirements in accuracy and real-time.Finally,in order to solve the problem of low frequency of visual navigation measurement,the information of the visual navigation system and the inertial navigation system with high output frequency is integrated by the extended Kalman filter algorithm,and the attitude,position and speed of the unmanned aircraft relative to the recovery point are designed.And the drift error of the inertial component on the drone and the state quantity as the EKF filter to obtain the visual navigation information as a Kalman filter for the quantity measurement.Simulation experiments show that the algorithm has good real-time performance,high navigation accuracy and no cumulative error.