Research On Image Recognition And Automatic Control Of Field Boundary Turning For Tractor

In 1977,China set the target of achieving agricultural mechanization.Today,the comprehensive mechanization rate for crop cultivation and harvesting exceeds 72%.Heilongjiang reclamation area has realized large-scale agricultural production mechanization.Intelligent and unmanned is the future direction of agricultural mechanization development,and the intelligent and unmanned tractor is the key.At present,the driver still needs to manually judge and control the steering of the autonomous tractor in the field.If field image recognition is applied to the tractor,the labor intensity can be further reduced,and the operation efficiency and safety can be improved.The electronic hydraulic steering system of the tractor is the key research object to realize the tractor’s auxiliary or autonomous steering.The tractor head-steering system needs high-precision automatic control.In order to solve the above problems,this paper takes the realization of location image recognition and precise control of tractor hydraulic steering system based on deep learning as the research objective,and uses deep learning and convolutional neural network to realize location image recognition.The kinematic model of the tractor was established and the Carsim software was used to simulate the steering of the tractor.The fuzzy PID algorithm is used to realize the precise control of the electro-hydraulic proportional hydraulic steering system of the tractor.The specific research content is as follows:Analysis deep learning and the principle of image classification and recognition of convolutional neural network are elaborated.Res Net-50 convolutional neural network was deeply analyzed,and the field image data set is used for deep learning training to obtain the field image recognition model.The model is tested with test sets and the confusion matrix is made.The test results show that the recognition model is effective for all kinds of field images(soybean stubble,seedling belt,plot to be prepared,corn stubble,corn stalk;The average recognition accuracy of soybean stubble,corn straw/stubble,cement road,land to be prepared and seedling belt could reach 98.17%.Interactive software is written based on Runtime environment of MATLAB to realize real-time reading and recognition of RGB camera images.The image processing speed of the tested software is 4-6 images per second.For the Tractor Field Boundary Steering after image recognition,the tractor steering kinematic model is used for simulation study in Carsim.The parameters of the tractor in Carsim were set according to the John Deere 1204 tractor,and the kinematic simulation model of the local path tracking of the tractor head steering was established.The tractor steering wheel angle parameters are output through the kinematic model as the input signal for the control test of the electronically controlled hydraulic steering system of the tractor.Analyze the tractor electro-hydraulic steering system and its main components,establish the model of AMESim,simulate the dynamic characteristics of control current and cylinder speed(Angle speed),the results show that the proportional reversing valve control current has linear relationship with cylinder piston speed.Design the fuzzy PID control algorithm model and build the joint simulation model using Simulink to compare the PID algorithm and the fuzzy PID algorithm.The results show that the fuzzy PID algorithm has higher accuracy and smaller steady state error,which can accurately control the hydraulic steering system.The correctness of the simulation model and the feasibility of the control algorithm are verified by the bench test of the tractor electro-hydraulic steering test platform.Theoretical analysis and experimental research show that the tractor electrohydraulic proportional steering system has the characteristics of high control accuracy,fast response speed and high reliability,which meets the requirements of the tractor steering control.The electro-hydraulic proportional steering system can be installed on the traditional all-hydraulic steering tractor,but also applied on the autonomous tractor or unmanned tractor,providing a new idea for realizing unmanned automatic steering combined with machine vision.