| The working environment of this study is corn field.In view of the shortage of rural labor and the problem of chemical herbicide covering spraying,a navigation baseline recognition method for maize 3-5 leaf stage and 5-8 leaf stage was proposed,and a weed recognition strategy between maize rows based on the position of maize row line was proposed..It lays a foundation for the realization of autonomous navigation of agricultural robot and variable spraying of chemical herbicides.The specific research contents are as follows:(1)Aiming at the problem that agricultural robots based on visual navigation are easily disturbed by weeds and lack of seedlings during the navigation baseline extraction process.A method is proposed to quickly and accurately extract the center line of the corn row as the navigation baseline of the small agricultural robot under the field environment of the corn 3-5leaf stage.Using ultra-green gray method and Otsu method to achieve real-time acquisition of crop information.Use parallelograms to simulate the characteristics of crop row length,pixels,and direction to achieve preliminary positioning of crop row positions.The trapezoid simulates the shape features of the crop row in the image,which is narrow and wide at the bottom to achieve the final positioning of the crop row position.Use the statistical characteristics of Hough transform to realize crop line detection,compare the angle deviation between the current frame navigation baseline and the previous frame navigation baseline(the first frame image is not compared),and the deviation beyond a certain range will inherit the previous frame navigation baseline line.The test data shows that it takes an average of 76 ms to process a frame of 640*480 pixels,and the accuracy of the navigation baseline extraction is about 94%.This research algorithm provides a navigation method with good real-time performance and strong anti-interference ability for the autonomous navigation of small agricultural robot in the3-5 leaf stage of corn.(2)A method for obtaining the navigation guideline of corn rhizome during the silking period was quickly and accurately obtained in the corn field environment.The normalized super-green feature 2G-R-B algorithm is used to extract the green features.Before the maximum inter-class variance method(Qstu algorithm)is used to realize the binarization of the image,the linear stretching method is applied to enhance the image contrast.The vertical projection method was used to locate the corn rhizome candidate,and the dynamic variable size window was introduced to complete the first elimination of the corn rhizome candidate location.The corn root cane candidate location was secondly based on the relative navigation trend line offset.Elimination,improved the extraction accuracy of corn rhizome positioning points.The least squares method was used to fit the selected rhizome positioning points to obtain the corn line.Finally,vector addition is used to obtain the angle bisector of the two corn lines as the navigation reference line.The experimental data show that the processing time of a640*480 pixel image is only 100 ms.The extraction accuracy of the corn rhizome navigation baseline for different working environments is about 91.3%.The algorithm has good real-time and robustness for agriculture.Automated guided vehicles provide reliable navigation algorithms for driving in corn fields.(3)How to distinguish weeds from crops quickly and accurately is the first problem to be solved in variable rate application of chemical herbicides.Due to the variety of weeds,the crop information is relatively single for corn field.This study proposes that the frame selection and location of maize row area is completed based on corn line first,so as to distinguish corn crops from weeds.In order to avoid the interference of corn cotyledons to the recognition algorithm,the distance between the center of gravity of weed connected area and corn line was used to determine the final weed image information.(4)The traditional PID algorithm needs to establish accurate mathematical model of controlled object.The navigation control strategy of agricultural robot based on fuzzy control is proposed.Based on the previous research content,a fuzzy controller with offset and deflection angle as input and voltage as output is designed.And use Matlab/Simulink to build a simulation model for simulation analysis.During the simulation,keep one parameter of the offset distance and the deflection angle at 0,and make a jump to the other parameter at a certain moment,and observe the output curve oscillation amplitude and path tracking effect.The test results show that a jump of the deflection angle will drive a large oscillation of the deflection angle,while the jump of the deflection angle has a small effect on the amplitude of the deflection.At the same time,the test data also shows that the fuzzy control system has fast correction capability and robustness.The fuzzy control model is suitable for the navigation control of agricultural robot in the corn field,and it has reached the expected design goal. |
