A Test Method Of Binocular Vision-based Guidance For Tractor Based On Virtual Reality

The vision-based guidance system of agricultural machinery operates flexibly in complex field.How to design the image process algorithm to enhance the accuracy and speed of pathway determination is an intrinsic and popular issue.The classical test methods for agricultural guidance system are dominated by real field test.There are many problems for the classical test methods,such as high test cost,strong dependence on crop growing period,long test period,and easy to damage to crops and so on.To solve the above problems,a novel test method based on virtual reality for binocular vision-based guidance system was presented in this paper.The tractor was used as the operate machinery,the cotton at seedling stage was used as the target crop,the crop row field was taken as the experimental environment.The main contents of this paper are as follows:(1)A geometry modeling method of virtual scene for the guidance test.The geometry model of virtual scene consists of tractor appearance model,cotton model,crop row model,weed model and road model,which is built based on geometric characteristics of the actual field for guidance test.The presented method can simulate many scenes of cotton crop row in real time according to test requirements,which provides rich and effective conditions for the research of crop row detection and path tracking control.(2)A modeling method of physics engine for the tractor.The mathematical model of the physics engine was built,including the whole vehicle model,tire model,transmission system model,braking system model,steering system model and resolver model.Software of the physics engine was developed based on C++ language in Visual Studio 2008.The presented method can calculate the dynamic parameters of tractor quickly and correctly according to the real vehicle parameters and the information of experimental scene.The physics engine can further render the position and attitude of the tractor in real time in the virtual test scene.(3)The pathway determination of field guidance based on binocular vision.The centerlines of crop rows were determined by detecting and computing the spatial distribution of crop feature points.That method reduces the number of points of image matching,and improves the accuracy and speed of recognition.Results show that,in the situations without turnrows,the accuracy of crop row detection is not less than 92.11%;for the deviation angle,the absolute mean value and the standard deviation value are not more than 1.07° and 2.52°,respectively;for the image processing time,the mean value and standard deviation value are not more than 202.90 ms and 17.75 ms,respectively.A method of pathway identification based on the tractor driving position was presented.That method identifies the target crop row correctly and quickly,and ensure a stable tracking of the same target crop row in the process of vision-based guidance.Results show that,the accuracy of pathway identification is 97.33%;the mean value and standard deviation value of the processing time is 0.017 ms and 0.017 ms,respectively.(4)A control method of path tracking for the tractor.A computation model of front wheel angle was built based on pure pursuit method.The steering control method was designed based on the incremental proportion-integral-derivative(PID)algorithm.The parameters of the PID controller were optimized according to genetic algorithm.A control strategy of path tracking was designed to realize high tracking accuracy and fast responsibility.Results of the virtual test show that the path tracking control system designed by the proposed method can track the target path quickly and stably with a small overshoot.(5)Verification tests of virtual guidance system for the tractor.In order to test the effectiveness of the physics engine,the virtual tests of step obstacle,random road and steering performance were carried on.The proposed crop row detection method was tested with many images of crop row in virtual and real cotton field,respectively.The path tracking control system was tested by the virtual tests of tracking the parallel step straight path and the inclined straight path,respectively.Finally,the binocular vision-based virtual guidance system of the tractor was tested by the virtual tests of tracking the straight and curve crop row,respectively.