Research On Fixed-Wing Unmanned Aerial Vehicle Mono-Vision Landing Attitude Angle Measurement Simulation

The in-depth application of UAV in many fields has put forward higher requirements for autonomous control.The fixed-wing UAV lands in a sliding way.Because it cannot hover in the air during the whole landing process,the autonomous landing must be realized by controlling the attitude angle in real time.In order to improve the accuracy of attitude angle measurement,this paper proposes an measurement algorithm based on adaptive variable range cooperative target detection,which is a new method based on monocular vision and can guide the precise landing of UAV.The main innovations of this paper are as follows:At present,there is no open source data set of UAV landing scene for research,and it is difficult to collect data set in realistic scene,this paper uses Unreal Engine 4(UE4)to build a UAV landing simulation scene.On the one hand,the virtual scene we built can simulate the landing of UAV and obtain images during the landing process,providing data sets for algorithm training and verification.On the other hand,it can reduce the landing accident rate by verifying the performance of visual navigation algorithms in virtual scene.Due to the limited focal length of the camera,the UAV can only locate large landmarks at a long distance,but needs smaller cooperative targets to provide guidance information at a close distance.This paper proposes an attitude angle measurement method based on adaptive variable range cooperative target detection.And in this paper,YOLOv5 s network is used to locate the position of cooperative targets in the landing scene,and then the coordinate information of several cooperative targets is used as control points to calculate the attitude angle by using EPn P algorithm.Experimental results show that the“YOLOv5s+EPn P” method proposed in this paper has strong real-time performance,and the average measurement error at different distances is within 0.6°.Aiming at the situation of uneven illumination,reflection and blur of cooperative target images taken in the real environment,this paper proposes texture enhancement and detection efficiency optimization algorithm based on YOLOv5 s and MUnet to eliminate the influence of image degradation on detection results,thus realizing a detection scheme with robustness to illumination degradation.