Development Of Automatic Navigation Method For Combine Harvester Based On Laser Scanner

Cereal is a kind of widely consumed food for a large part of the world’s population,harvesting cereal crop also plays a very important role in agricultural production.In traditional form of harvesting work,the operator of combine harvester should operate the steering and supervise the crop field at the meantime,which is pretty a tedious and monotonous job,and can be easily result in declined operation quality and efficiency,more seriously cause security concerns.While automatic control offers many potential advantages over human control,such as high quality and efficiency and relieve the demand for labor force,thus developing automated combine harvester for cereal has become an urgent issue.The objective of this research was to develop an automatic navigation guidance method for combine harvester based on Laser Rangefinder technology,and carry on the automatic harvesting experiment on wheat field.The main contents for this research were as follows:(1)A set of 3 Dimensional data collection system based on Laser Rangefinder technology was developed.The set of data collection system was composed of a Laser scanner,an inertial measurement unit(IMU),and a Pan Tilt unit(PTU).The PTU enlarged the detecting range in front of the combine harvester with the precisely rotations in both pan and tilt directions.A mathematic model was built for transferring the 2 dimensional information into 3 dimension space,which realized the 3D field information reconstruction in front of the set of sensors.An automatic position correction method was proposed,the initial installed angular deviation for the set of data collection system with the platform was calculated based on the relative position between the coordinate of Laser scanner and inherent feature of the installed platform,and was then eliminated by the rotation of PTU in the pan direction,which could overcome the difficulties in post data processing caused by the initial mounting error.(2)A Laser Rangefinder technology based crop edge detection method was developed.An error points removed method based on improved linear regression method was put forwarded,to pre-process the noise in the Laser point cloud so as to guarantee the accuracy and reliability of the collected data.An improved gradient based method and an Otsu’s method based were presented to detect the edge between the cut and uncut crop.Static field experiment was conducted to compare the two methods during harvesting season,to verify the error of the detection system,the real position of the edge was measured by RTK-GPS for accuracy evaluation,and the results showed that Otsu’s method provided a better performance with an average offset of 5.4cm,max offset of 8.3cm,and a RMS of 3cm.(3)An automatic cntroll system for navigation harvester was developed.An AG1100 combine harvester equipped with the edge detection system was adopted as a test platform for the guidance system.During the dynamic harvest test in the wheat field,the lateral error ? and heading error ? of the vehicle were worked out based of the navigation path,and steering angle of the combine harvester were calculated by a PI controller with the pre-set lateral gain λ1 and heading gain λ2.And then the steering angle was sent to the computer through the already developed CAN bus,the combine harvester steered following the received order.And an laser scanning based romote monitoring interface for navigation guidance system was developed for the purpose of safety.The real wheat filed harvesting test results indicated the auto guidance system could steer the combine harvester follow a crop row at an average accuracy of 8.0cm,with a max deviation of 23.4cm and RMS of 4.6cm,when the combine harvester speed was about 1.0 m/s.This accuracy was enough for wheat harvesting considering.(4)A vision based edge information acquiring algorithm was proposed to overcome the shortages of the Laser Rangefinder based edge detection method.Inverse perspective mapping(IPM)geometrical transform was used to remove the perspective effect on original mature wheat field.The coordinate systems of the image and the vehicle were matched,on the basis of which,the Region of Interest(ROI)was restricted to reduce the range of image processing.The image of ROI was then processed by illumination reduction and contrast enhancement to make the difference between cut and uncut wheat field more evidently,and then transferred into grayscale image.Threshold segmentation method based on histogram was used to convert grayscale image into a binary image,so as to distinguish the cut and uncut wheat.The target points were clustered by adopting cross correlation method on each horizontal scanning line in the binary image,and then least square method was used to detect the edge line between cut/uncut wheat.The results of field experiment showed that the vision based crop edge detecting system performed an accuracy rate of 97.7%,and can provide reliable navigation information under the conditions when the Laser based edge detection system was in quite low accuracy.