| In recent years,the continuous promotion of agricultural equipment upgrading in China has driven the increase of mechanization rate in grain harvesting.Among them,grain harvesting is one of the most important links.The application of navigation control technology in grain harvesting operations is considered the key to achieving efficient and low-cost harvesting.Currently used navigation systems for combine harvesters are mainly based on satellite positioning technology,which is suitable for unmanned farm global planning path driving work.However,there are shortcomings in real-time boundary identification,difficult secondary development of navigation system,and high cost.Some sensor fusion navigation systems based on filtering methods also have certain limitations in relative position recognition and attitude updating.In order to improve the positioning accuracy and efficiency of "full coverage operation" of combine harvester navigation systems and reduce the cost of navigation systems,this paper proposes a navigation control system based on visual SLAM-inertial fusion.The aim is to improve the stability and efficiency of the combine harvester navigation system.The main research contents of this paper are as follows:(1)Design of the Framework for Navigation Control System of Combine Harvester.The operational process of the combine harvester is described as a state estimation problem of visual SLAM,and the entire system is divided into four parts:sensor data acquisition,field navigation information processing,relative position calculation of the combine harvester,and steering control.A binocular camerainertial-carrier coordinate system is established,and coordinate transformation matrices are calculated.Finally,error analysis and joint calibration of the binocular camera and inertial IMU mathematical models are performed to provide theoretical support for the design of the navigation control system.(2)Acquisition of Field Navigation Information for Combine Harvester.Firstly,the field image information is obtained through a binocular camera.After filtering and enhancement processing,the crop boundary line is extracted as the navigation baseline of the harvester.Considering the special characteristics of crops in the field,an improved FAST-ORB feature point extraction algorithm is designed to improve the accuracy of binocular visual pose estimation.In addition,the IMU data is used to assist binocular stereo matching and outlier removal to improve the success rate of stereo matching.Finally,the pose of the combine harvester is estimated using the binocular camera,and a least-squares optimization model based on reprojection error is constructed.(3)Design of Sliding Window-based Visual-Inertial Fusion Pose Estimation and Steering Controller.Firstly,establish a framework flowchart for visual-inertial information fusion and perform joint initialization of binocular camera and IMU.Based on tightly-coupled sliding window optimization method,error optimization is performed on relevant information such as images and IMUs,and the objective function is nonlinear least squares optimized to obtain the pose of the combine harvester.According to the field environment and the driving state of the wheeled combine harvester,establish a path deviation prediction and tracking model for the harvester,analyze the relationship between the lateral deviation and heading deviation angle in the steering control variable,design the steering control method,and conduct simulation and dynamic response tests.The average response time is 2.56 s,the average delay is 50 ms,and the error is 0.7°.The designed steering controller can meet the requirements of steering control for the combine harvester.(4)The software design of the navigation control system for a combine harvester is based on the design requirements of the navigation system.The design is implemented using an NVIDIA TX2 development board as the processor,and QT development framework is used to design the navigation and steering control software for the combined harvester.This includes the design of an image acquisition program,a binocular image processing program,a sliding window pose optimization estimation program based on visual-inertial fusion,a steering control program and a system display interface.Additionally,in accordance with the operation requirements of the control system,corresponding program designs have been developed for overall system data processing,human-machine interaction and data storage.(5)Validation of the field trial effectiveness of the navigation and control system for a combine harvester.Firstly,an experimental platform for the navigation and control system of the combine harvester was constructed,and the corresponding sensors were selected and tested based on the established coordinate system.Secondly,corresponding field tests were conducted,indicating that the visual sensor can meet the requirements for spatial distance recognition and the steering mechanism can meet the navigation system requirements in the simulated test for lateral deviation during field operation.Finally,the field trial results showed that the experiment machine integrated with this system can achieve the expected harvesting effect when operated at speeds of 0.9-1.5m/s,with an average lateral deviation range of 2.21-8.62 cm and a standard deviation range of 0.13-4.21 cm,and an average cut rate range of 92.2%-96.0%.Under similar operating conditions,the average lateral deviation range of the combine harvester integrated with the vision-inertial navigation system based on the filtering algorithm is 0.35-4.48 cm at an operating speed of 0.6-1.2m/s,and the standard deviation range is 1.23-5.71 cm.The lateral deviation standard deviation of the navigation and control system developed in this paper is better than that of the above system,and has better stability.The developed navigation and control system is cost-effective and easy to promote and apply,providing technical support for achieving unmanned driving for combine harvesters. |
PDF Full Text Request