Research On Combined Navigation Of Agricutural Machine Based On Visual Navigation And GPS Navigation

Agricultural machine navigation is an important way to make the machine driving automatically and improve the intelligence of agriculture.Agricultural machine navigation technology is a technology which uses light,electromagnetic waves and satellite signals which are get from sensors to obtain information about surroundings and locate itself and navigate according to maps.At present navigation technology contains visual navigation,GPS navigation and so on.The effect of visual navigation is influenced by light.Besides,images contain a large amount of data that needs to be processed.The signal strength of GPS is quiet weak indoor and high accuracy positioning requires high price equipment.Aiming at the problems which are mentioned above,this paper proposes a combined navigation method based on visual navigation and GPS navigation which is combined with advantages of two navigation methods.The combined navigation uses Unscented Karman Filter(UKF)to fuse data from the two methods and filter information with the purpose of improving accuracy of navigation.This paper also researches Ackerman turning model and proposes an auto turning algorithm based on forward-and-backward strategy so that the machine will not drive on crops.Base on this algorithm,this paper proposes a searching optimization algorithm to find out a turning way which has better performance.This paper solves the straight driving problem by combined navigation and solves the turning problem by searching optimization algorithm so that we will avoid the loss of driving on crops.In the solution of straight driving problem in driving automatically,this paper gets data from environment by visual and GPS navigation and process data.After the data is processed,UKF is used to filter data from two different ways so that the method predicts and update navigation points.The process of visual navigation contains image preprocessing and extracting the feature points.This paper chooses the method which has best performance to preprocess image.This paper proposes a method to extract feature points which costs less time and gets feature points directly.After getting the coordinate of points,it should be transformed to the world coordinate system.The process of transforming involves many coordinate systems,which contain pixel,image,machine and world coordinate system.The GPS coordinate also needs to be transformed to world coordinate system.After the coordinates is in the samesystem,UKF is used to fuse and filter two kinds of data.The experiment is runned in a simulated method and uses GPS navigation data as standard to compare the performance of using visual navigation alone and using combined navigation method.The experiment shows that combination navigation improves the effect of location and performs better than visual navigation.In the solution of turning problem in driving automatically,this paper designs and implements the turning algorithm based on Ackerman turning model and forward-backward strategy.To verify the feasibility of the algorithm,this paper constructs model based on real machine and run simulating experiment.The experiment searches angle between 30 degrees and 45 degrees on which the method has best performance.The result shows that searching optimization algorithm can search the best method quickly and fulfil the turning task.