Research On Obstacle Avoidance Technology Of Agricultural Rotorcraft UAV In Small Area Based On Visual Navigation

Visual detection and path planning technology are important guarantees to ensure the safety of Unmanned Aerial Vehicle(UAV)in plant protection operations,especially in the field with complex environment,safety operation has become the primary consideration.This paper analyzes the limitations of the traditional operation mode,further studies the shortcomings of the traditional algorithm in the application,and puts forward a new improvement strategy to improve the efficiency of the actual work and make it adapt to the complex operation environment.As one of the front-end technology modules of Simultaneous Localization and Mapping(SLAM),Oriented FAST and Rotated BRIEF(ORB)feature matching plays an important role.In this paper,an improved scheme combining K-Means ++ clustering preprocessing and sparse optical flow matching is designed.In the preprocessing stage,the image information is smooth-processed.After detecting the feature point set,the Harris response value is calculated to remove the pseudo-feature points and retain higher quality feature points.In order to solve the problem that the original algorithm relies on the exhaustive search and matching method to cause the calculation redundancy,the K-means++ center clustering algorithm,sparse optical flow method and traditional algorithms are merged to improve the matching accuracy,and then the Random Sample Consensus(RANSAC)algorithm is used to further improve the matching accuracy,so as to optimize the matching process.Based on the theoretical basis of Artificial Potential Field(APF),the paper summarizes its own defects and analyzes that there may be more constraints in complex environment.The main improvement strategies are as follows: In view of the unreachable target problem,the repulsion function is improved,and the relative distance between the UAV and the target point is introduced,so that the target point is always in the minimum potential field;Aiming at various situations of local minimum problem,a "touch guidance" subgoal node escape strategy is proposed.At the same time,considering that there may be multiple types of obstacles in the complex operating environment and the yaw problem caused by obstacle avoidance,this paper divides the boundary of obstacles,plans different threat ranges,improves the gravity function,introduces the concept of "line potential field",and adds the attraction of the original route to the UAV.In view of the possible influence of dynamic obstacles,based on the original algorithm,target points,obstacles,UAV speed and acceleration information are introduced to increase the UAV’s cognitive information of the environment.In order to solve the problem of excessive curvature and oscillation of local path,Bessel curve method is used to optimize the flight path,so as to meet the actual working requirements.In this paper,a combination of simulation verification and real aircraft test is adopted to verify the feasibility of the improved algorithm.The test results show that the optimization algorithm can effectively solve the corresponding problems,and has reference significance for the UAV plant protection operation technology.