Research On Vision Positioning System Of Autonomous Landing Of Unmanned Powered Parachute Vehicle

With the rapid development of China’s soft-wing unmanned aerial vehicle technology,unmanned powered parachute vehicle（UPPV）is widely used in military and civil fields because of its excellent aerodynamic and controllability.In order to realize the high-precision autonomous landing of UPPV,this paper studies the visual positioning system of autonomous landing of UPPV;the sliding mode controller is designed and the runway recognition model is trained,so that the UPPV can complete the autonomous landing.Firstly,the 6-DOF mathematical model of the UPPV is established and a sliding mode controller is designed.Calculate the additional mass,mass characteristics,aerodynamic force and moment of the UPPV system,and establish dynamic equations and kinematics equations;a sliding mode controller is designed and compared with PID controller by MATLAB software,the result shows that the designed sliding mode controller has obvious advantages in control response speed and smoothness,and can better control the flight attitude of UPPV.Secondly,the lightweight runway recognition model is trained and the offset correction algorithm is designed.Using Mobilenet-V3（Large）neural network and YOLOv3 deep learning algorithm to train a lightweight landing runway recognition model with a comprehensive accuracy of 97.47%.Compared with other models trained under neural networks,it has obvious advantages in detection speed and accuracy;an algorithm for the offset correction of the landing runway of the UPPV is designed,and identification tests are carried out in a variety of scenarios,which proves the feasibility of the algorithm.Finally,the hardware design and overall construction of the testing platform for the vision positioning system of autonomous landing of UPPV,including the design and selection of the hardware,the model is transplanted and the communication between flight controller and airborne image processor is designed.The testing platform to be built has carried out multiple autonomous landing flight tests in a safe area and analyzed the test data.The results show that the vision positioning system of autonomous landing of UPPV designed in this study has strong feasibility and high reliability.