| Kiwifruit is a cross-pollinating plant,and artificial assisted pollination can effectively improve the quality and yield of kiwifruit.In view of the problems of low efficiency,high labor intensity,high cost,and large pollen consumption of mechanized pollination equipment,taking the trellis-type standardized kiwifruit orchard as the object,by analyzing the distribution characteristics of flowers in the field,determining the flower identification and positioning scheme,designing the air-fed metering pollinator,building a pollination test platform,determining the optimal control parameters of the pollinator,designing the structural parameters and control system of the Delta target mechanism,studying the motion control method of the target mechanism,and building a prototype of the air-fed dry powder pollination device for the target.Through field tests,the performance of pollination operations was verified.The main research contents are as follows:（1）Analysis of target pollination agronomic parameters.Through the investigation and review of relevant data of standardized"Hayward"kiwifruit orchards,the environmental parameters and flower distribution rules of standardized kiwifruit orchards were obtained,and the flowers were concentrated in the canopy about 160cm-180cm high from the ground.By analyzing the number of pollen grains required for full pollination of kiwifruit combined with pollen quality,the amount of pure pollen deposition in the stigma area of fully pollinated kiwifruit flowers should be greater than 4.3×10-2mg.（2）Research on flower identification and localization method based on binocular vision.A field image acquisition device was set up,and kiwi flower image data was collected for different weather,different time periods and different collection distances,and the flower images were expanded using image enhancement technology.In order to avoid the inconsistency between the center of flower mass and the center of stamen due to the defection or occlusion of petals,the stamen was selected as the label feature of kiwi flowers,and the kiwifruit flower recognition model was trained.Based on the application requirements of embedded devices,the CSPDarknet backbone feature extraction network of YOLOv5n is replaced by ShuffleNetV2 lightweight network,the ordinary convolution of the original network is replaced by Ghost convolution,the GhostConv module and C3Ghost module are constructed,and the Conv layer and C3 layer of the original network neck are replaced,and an improved YOLOv5n identification network is built.The improved network mAP is 98.3%,which is 0.2%lower than that of the original network m AP of YOLOv5n,and the model parameters,calculation amount and model size are 0.46M,0.9GFLOPs and 1.27M,respectively,which are reduced by 70%,78.6%and 66%compared with the original network model parameters,computation and model size,and the detection speed is 37.1ms,which is38.4%higher than that of the original network.The binocular positioning method was used to construct the flower positioning algorithm,which realized the spatial positioning of kiwi flowers.By building an indoor test scaffold,the optimization parameters of flower positioning were analyzed,and the recognition distance was 500mm,and the field of view was about35cm×and the positioning accuracy of 10mm could be reached at 35cm.（3）Design and optimization of control parameters of air-fed dry powder metering pollinator.The air-fed dry pollinator was designed,and the inner cavity structure of the pollinator was simulated by Fluent software,and the influence of straight pipe structure and throat structure on the pressure at the pollination position and the pollinator outlet wind speed was analyzed,which verified that the throat structure was more conducive to pollen input at the pollinator inlet.Through the construction of the pollinator test platform and the multi-factor test analysis,the pollinator optimization parameter combination was determined as follows:pollination time 0.7s,fan voltage 9.6V,pollination distance 50mm,corresponding pollen deposition quality of 1.314mg,pollen utilization rate of 68.55%,pollen coverage rate of 49.48%.The results of field experiments showed that the pollinator pollinated with an optimized combination of parameters,with a fruit set rate of 96.7%,an average fruit weight of 94.57g,and a fruit shape index of 1.21,which was consistent with the pollination effect of the electric pollinator,but lower than that of manual spot pollination,and the pollen dosage was about 2mg per flower,which was 40%and 60%less than manual point pollination and electric pollinator.（4）Delta parallel target mechanism design and control algorithm construction.According to the camera’s field of view space and scaffolding height,calculate and determine the structure and motion parameters of the robotic arm;The simplified model of the manipulator was established,the kinematics analysis of the manipulator was carried out,the kinematics equation of the manipulator was established,and the forward and reverse solution operation method of the manipulator was determined.The working space of the manipulator is simulated,and the rationality of the structural parameters of the manipulator is verified.Based on the ant colony algorithm,a mathematical model of pollination task planning was established,the specific process of the ant colony algorithm to achieve task planning was analyzed,and the pollination path planning algorithm of robotic arm based on ant colony algorithm was tested in MATLAB.The path planning method based on PH curve and the trajectory planning method based on S-type velocity curve are determined,and by analyzing the trajectory of the end of the robotic arm in the X and Y directions with time,it is verified that the proposed method can effectively avoid the shock and vibration phenomena caused by the sudden change of velocity and acceleration during the movement of the manipulator.（5）Prototype construction and test of air-fed kiwifruit dry pollination operation device.Combined with the camera layout parameters,the combination of the optimal control parameters of the pollinator and the relevant structural parameters of the target mechanism,the prototype of the pollination operation device was built.Designed and built the whole machine control system and circuit system;Through the field pollination test,the performance of the whole machine identification system and the target operation system were tested.The average identification accuracy of the prototype identification system of the air-fed kiwifruit dry pollination operation device was 93.3%and the average positioning accuracy was 95.7%.The average target accuracy of the target operation system was 85.7%;The pollination accuracy was 80%,the fruit set rate after accurate pollination was 91.7%,and the pollination time of single flower was about 4 s.Compared with artificial pollination,the prototype of this institute has no obvious advantages in pollination efficiency,but based on artificial intelligence recognition,automatic pollination operation of robotic arm and accurate measurement of pollen,it has comprehensive advantages in pollen saving,consistency of pollination effect and manpower substitution. |
PDF Full Text Request