Research On Binocular Visual Obstacle Detection And Obstacle Avoidance Control Of Quadrotor UAV

With the development and application of artificial intelligence,more intelligent requirements have been put forward for the quadrotor UAV which is widely used in various fields.The ability to autonomously avoid obstacles has become an important indicator for the intelligentization of quadrotor UAVs.In a low-altitude complex environment,the quadrotor UAV obstacle avoidance system needs to take into account the requirements of rapidity,safety and accuracy.However,the navigation system of traditional aircraft is difficult to meet the requirements of obstacle avoidance applications.In view of the above problems,this paper studies the quadrotor UAV unmanned low-altitude environment binocular visual obstacle avoidance control system.The main research work of the paper is as follows:Firstly,the binocular vision calibration experiment is carried out and the dynamic model of quadrotor UAV is established.The basic model of the camera,the principle of binocular vision ranging and the method of binocular vision calibration are described.On this basis,the binocular vision calibration experiment is carried out.The flying principle of the quadrotor UAV is analyzed,and the dynamic model of the quadrotor UAV is established on this basis.The existing obstacle avoidance system is analyzed.Secondly,in view of the real-time requirements of binocular vision detection obstacles,an obstacle detection framework based on improved stereo matching and improved U＿V map method is proposed.Based on the semi-global matching stereo matching framework,a fast Census transform is used to obtain algebraic value,combined with SGM’s semi-global cost aggregation algorithm for stereo matching,and parallax optimization through left-right consistency checking and median filtering algorithm.The obstacle detection method based on U＿V map is studied.Based on the shortage of too many horizontal lines for U＿V map fitting,an improved method based on global K-means clustering is proposed.Experiments show that the proposed three-dimensional matching framework combined with improved obstacle detection methods can detect obstacles in the environment well,and provide an important basis for the path planning of obstacle avoidance systems.Finally,for the safety and accuracy requirements of three-dimensional obstacle avoidance of quadrotor UAV,a path planner based on improved artificial potential field and an obstacle avoidance control system based on adaptive sliding mode controller are designed.According to the proposed obstacle expansion model,an improved artificial potential field algorithm is used for path planning;considering the real-time nature of obstacle avoidance,a tangent point optimization method is used to improve the artificial potential field algorithm.Considering the uncertainty and external disturbance of the actual control system,an adaptive non-singular terminal sliding mode controller is designed for the quadrotor UAV dynamic model.The controller uses an adaptive term to estimate the upper bound of interference,and designs a terminal sliding mode controller for the quadrotor position loop and attitude loop respectively according to the interference estimate value;and uses Lyapunov theory to prove the stability of the system.The simulation results show that the path planner has good real-time performance,and the designed controller has good tracking accuracy and tracking speed.