Research On Visual Detection Of Landing Area And Autonomous Landing Guidance Of UAV

With the increasing complexity of vertical takeoff and landing UAV applications,visual guidance technology has been widely concerned by researchers.In order to break through the present situation of independent research on vision and guidance,and avoid the dependence on altitude sensor,this paper studies the vision guidance technology based on terminal constraint in the background of UAV landing guidance,from the aspects of cooperative and non-cooperative environments.Firstly,the scheme of visual guidance system is discussed.The current research situation is discussed from three angles of input,output and regular design of guidance module.At the same time,the applicable situation and research malpractice of each scheme are comprehensively analyzed.According to the demand of the system under the concrete scene,the best research scheme for the guidance module is determined.Then,in term of the vision guidance system based on cooperative target,aiming at the limitation of the existing planar cooperative target to UAV landing,a new three-dimensional cooperative target is designed to improve the anti-occlusion robustness of the vision detection system.With the imaging of the cooperative target height vector as the guiding parameter,the relationship between the image information and the velocity command is established directly.The vision and guidance modules are combined without relying on the height sensor and the position solution module.Meanwhile,the efficiency of visual guidance system is improved.Next,in term of the visual guidance system for non-cooperative environment,aiming at the complexity of effective detection of non-cooperative targets,the location of the geometric center of the secure region is detected and extracted based on the convolution neural network classification scheme.With the angular velocity of the target line of sight and the three-dimensional optical flow of the target motion,the relationship between the image information and the acceleration command is directly established.The two modules of vision and guidance are fully integrated without relying on the height sensor.The efficiency of visual guidance system is further improved.Finally,the effects of the above system scheme,3D cooperative target design,landing area detection algorithm and related guidance methods are tested and verified.By theoretical analysis and simulation verification,it is proved that the 3D cooperative target has better anti-occlusion performance.With the guidance law of corresponding design,the simulation proves that the guidance system can get rid of the altitude sensor and the landing path is short for lower mobility requirements.The detection index of non-cooperative landing area reaches 0.98,which is superior to the existing algorithm,and the real-time performance meets the practical application needs.With the guidance law of corresponding design,the simulation results show that the guidance system does not depend on other height sensors.The stability of the landing system and the correctness of the steady-state error formula are verified,and it is proved that the designed system meets the needs of practical application.