Research On Visual-Inertial Navigation Of Paddy Transplanter

Intelligent agricultural machinery and equipment is an important part of smart agriculture.In addition,it is also the key research field of "Made in China 2025".In recent years,China’s agricultural machinery automatic navigation technology has developed vigorously,but most of the existing agricultural machinery automatic navigation research is based on satellite positioning technology.Aiming at the problems of high cost and poor intelligence of agricultural machinery for a single highprecision satellite-positioning device,this research takes the ride-on rice transplanter as the navigation platform to carry out the investigation of visual-inertial integrated navigation.The main research contents are as follows:The extraction technology of the visual reference baseline of the rice transplanter is studied.The state-of-the-art deep learning method is applied to semantic segmentation of the seedling-transplanted area in the image of paddy field.The navigation baseline is extracted from the segmented masks by the machine vision method.The navigation parameters are calculated from the navigation baseline by using the posture of camera and vehicle.The visual-inertial fusion technology of rice transplanter integrated navigation is studied.The calibration and error analysis of each sensor are carried out to obtain various structure and performance parameters of the sensor,which provide an important basis for integrated navigation.Taking the posture and navigation parameters of the rice transplanter as the system state,an extended Kalman filter is designed to fuse the output of vision and inertial modules.The control technology of visual-inertial integrated navigation is studied.According to the calibration results of the steering wheel,the steering wheel is modeled,and a PID controller is designed.The PID parameters are adjusted by the decay curve method to realize the stable control of the steering wheel.Using Simulink simulation software,a pure tracking algorithm is designed,and the look-ahead distance is determined.On one hand,based on the Yanmar VP6 D rice transplanter,the construction of the visual-inertial integrated navigation hardware system was completed.On the other hand,the software and algorithm design of the visual-inertial integrated navigation of the rice transplanter was completed based on the ROS,a kind of robot operating system.Using RTK-GNSS to obtain the real-time positioning of the rice transplanter relative to the crop row,the lateral deviation of navigation was calculated,and the integrated navigation experiment was completed.The experimental results show that the lateral deviation of straight-line navigation is 0.044m±0.040 m and that of turning navigation is-0.088 m ±0.143 m.This research provides a reference value for the research on visual-inertial integrated navigation of rice transplanter without relying on any high-precision satellite positioning.It also lays a foundation for further improvement of the intelligence level of rice transplanter.