| In North China Plain where annual maize-wheat rotation is predominantly managed,the blocking issue occurs during wheat no-till seeding,and the resulted poor seeding quality have become the key constraints for the implementation of whole-process mechanization of conservation tillage.It has been demonstrated that the above problems can be alleviated by wheat row-follow seeding,which means to seed wheat between two adjacent maize stubble rows in order to avoid the aboveground standing maize stubble and consequently the belowground root system.To improve the efficiency and precision of wheat row-follow seeding,the maize stubble avoidance anti-blocking technology for row-follow no-till seeder was studied based on machine vision.The main results are as follows:(1)The image segmentation method for maize stubble row was studied based on regular color image.According to the analysis results based on sample images,the distinctive pale yellow stubble incision was used to separate stubble row from the image background in RGB color space.The coefficients for gray image acquisition was optimized using genetic algorithm;to reduce the difficulties in image segmentation,a region of interest(ROI)was selected by calculating the maximum of column gray value accumulation;and the ROI was transformed into monochrome using Otsu method.A noise elimination method was developed based on erosion operation,which can efficiently remove the noise aside the maize stubble row in ROI monochrome.(2)The image segmentation method for maize stubble row was studied based on hyperspectral image.Hyperspectral images which cover a range of spectrum from 400nm to 1000nm,as well as images which cover a range of spectrum from 1000nm to 2500nm were collected outdoor for maize stubbles.By differential spectrum analysis and principle component analysis,three optimal wavebands(929nm,968nm,1000nm)were selected from VIR-NIR(400-1000nm),while three optimal wavebands(1260nm,1658nm,2131nm)were selected from SWIR(1000-2500nm),both two waveband subsets could achieve accurate maize stubble row segmentation under intense illumination conditions.(3)A guidance baseline detection for row-follow seeding was studied.Based on the obtained ROI monochrome,the standard Hough transform and the least square method were respectively applied for guidance baseline detection.By comparing the two detection results,the least square method was selected to guarantee the real-time performance of vision-based guidance system.Furthermore,an improved algorithm was put forward to improve the detection precision of traditional least square method.In the proposed method,all the feature points for baseline fitting in ROI monochrome were evaluated to judge their effectiveness,feature point that has been judged as noise point was removed,and the remainder were used for the final baseline fitting,in this way precise baseline detection was finally achieved.(4)A row-follow control system consists of actuator and software system was developed for furrow opening operation.The actuator was designed to allow the orientation adjustment and lateral movement of furrow openers.For this actuator,an electric motor combined with an worm-gear reducer were used to adjust the orientation of opener units,while the electric cylinder was used to achieve overall lateral movement of all openers.A decision control system was developed based on image processing results,in this system the principle computer was used for image processing and motion planning,the subord:inate computer was responsible for motion parameters calculation and the control of both electric motor and electric cylinder.The structure and function of the principle computer and the subordinate computer was designed,meanwhile the communication protocol between them were established.(5)A method for row-follow control was studied.Based on the structure of the actuator and software system,the preview control process was determined while a fuzzy controller responsible for the decision of the deflexion of furrow opener was designed.A synergic movement model for electric motor and electric cylinder was established to reduce the resistance during lateral movement of furrow openers.A simulation was conducted to analyze the row-follow control process,and the results indicated that when the preview time is 1.2 s,and the tractor speed was less than 1.2 m/s,the mean variation between the actual trajectory and expected trajectory of furrow opener was less than 3cm,which met the accuracy requirement for maize stubble avoidance.(6)A field experiment was carried out after the integration of vision system and row-follow control system.The results indicated that the furrow openers can successfully avoid the maize stubble when the tractor speed was less than 1.2 m/s,which met the requirement for row-follow operation. |
PDF Full Text Request