Research On Navigation System Based On Field Seedling Vision Recognition

The working environment of agricultural robots is mostly open,complex and unstructured farmland environment,which causes the complexity and difficulty of machine vision navigation in agricultural applications.How to extract the available navigation information from it has become a bottleneck when the current application of machine vision in agriculture,especially in paddy fields.This paper analyzes the characteristics of field seedling images and develops a research for a visual navigation system.The main research contents include:(1)Aiming at the influence of light intensity acquired from image feature extraction of field seedlings,a method for extracting the reference line of seedlings that can remove light effects is proposed by comparing and analyzing the grayscale results of each component under the RGB,HIS and YUV color models.he feasibility of U component of YUV color model as the optimal component of image graying is verified through actual comparison.Due to the limitation of viewing angle due to the camera image acquisition,the seedling list shows irregularities that gradually intersect with each other,and the actual orientation information of seedlings cannot be obtained.This paper proposes to use the improved IPM transformation formula for inverse projection transformation to obtain seedling images.In the vertical top view of the plant,the actual field position was restored,and finally the simulation experiment showed that the improved IPM formula can effectively eliminate the large errors caused by the near-field bending of the camera under the premise of almost the same processing time.In order to improve the accuracy of the fitting straight line for the extracted seedling contour feature points,a clustering method was proposed to cluster different seedling feature points.Finally,the Hough transform based on known points is used to extract the seedling navigation lines.(2)The coordinate systems of the mobile robot platform are constructed and a fuzzy controller is designed.Based on the accurate extraction of seedling navigation lines,this paper solves the pose parameters of the mobile robot platform through the geometric relations between the coordinate systems,and selects the pose parameters as input to design a fuzzy control algorithm.The system transferring function and state space equation of the robot mobile platform is established,and the effectiveness of the fuzzy control algorithm control is verified by MATLAB simulation.(3)Simulation tests are conducted on the visual navigation system of the robot mobile platform.The experimental tests mainly include: on the one hand,the lateral position deviation and the heading angle deviation are measured when the mobile robot platform is stationary,and compared with the calculated deviation value,and the error of the calculated deviation value and the actual measurement deviation value are both controlled within a relative small range.On the other hand,a simulation experiment is conducted on the autonomous navigation system of the robot mobile platform through self-guided navigation of different seedlings,10 groups of position deviation and heading angle deviation data are obtained.Comparative analysis showed that the visual navigation system developed in this paper is feasible and effective.The experimental results show that the navigation control algorithm of the robot mobile platform designed in this paper can be applied to field seedling autonomous walking control,including paddy field environment.The visual navigation system developed in this paper can accurately identify seedlings and extract seedlings’ navigation lines,and can steadily track navigation lines at the same time.