Research On Visual Navigation Technology Of Agricultural UAV Based On ROS

Agricultural unmanned aerial vehicle(UAV)is a new type of agricultural machinery with great superiority compared with general agricultural machines,and is a research hotspot in the field of precision agriculture.Among them,the visual navigation technology is an important prerequisite and key foundation for agricultural UAV to realize precise operation in the field.In order to realize the automation and precision of agricultural UAV,this project designs an agricultural UAV visual navigation system using ROS operating system as a simulation experiment platform.The system detects the navigation line by the farmland images acquired by the on-board camera,and then controls the autonomous flight of the UAV along the centerline of the crop rows in the monopoly crop fields.The main task in this thesis is as follows:(1)According to the task requirement of visual navigation line detection,this thesis proposes a detection algorithm combining real-time semantic segmentation network and random sampling consistent algorithm.An improved ENet model is designed to achieve inter-row segmentation of farmland images,introducing residual branching to efficiently extract low-dimensional information of images and improve segmentation accuracy.An improved random sampling consistent algorithm is proposed for the segmented image to achieve the fitting of crop row centerline.According to the crop row characteristics,a new model scoring index is defined to determine the optimal set of interior points.The experimental results show that the detection algorithm has good adaptability to all types of farmland images with high accuracy and small number of model parameters.Therefore,the visual navigation detection algorithm proposed in this study can be deployed in embedded devices to provide accurate navigation information for subsequent agricultural UAV flight operations in agricultural fields.(2)According to the task requirements of visual navigation control,a fuzzy adaptive PID controller is designed to realize the precise flight of UAV.Based on the problem of poor flexibility and applicability of the string-level PID algorithm,this study introduces the fuzzy control algorithm that can automatically adjust the parameters.And Simulink is used to simulate the UAV modeling and compare the performance of each serial-level PID control algorithm and fuzzy adaptive PID control algorithm.The experimental results show that the fuzzy adaptive PID controller has stronger stability and more outstanding anti-interference ability,which can realize the accurate control of UAV visual navigation.(3)Based on the Gazebo simulation software in ROS,a virtual farming environment with monopoly crops was designed to simulate the whole system to test and verify the feasibility and reliability of the UAV autonomous navigation system based on vision detection.The experimental results show that the vision navigation system can operate well on the agricultural UAV equipment with good navigation effect.In summary,the agricultural UAV visual navigation system proposed in this article can achieve the expected effect,which has certain practical significance for the development of agricultural UAV in smart agriculture.