Design And Experimental Research Of Variable Spray Control System For Kiwi-fruit Orchard Based On LiDAR

At present,the control of diseases and insect pests in orchards mainly relies on chemical pesticide spraying.The large-capacity and full-area spraying of the traditional sprayer ignores the difference between the target of fruit tree canopy and the density of branches and leaves,a lot of pesticides are wasted and lost,not only the quality of the fruit was affected but also environmental pollution was caused.The precise variable spray system can determine the amount of pesticide application according to the characteristics of the canopy detection information,thereby the purpose of safe fruit production and environmental protection was realized.Based on the Li DAR sensing system,the method of deep learning was used to semantically segment the information of kiwifruit canopy branches and leaves,and a variable spray control system was designed for the segmented canopy leaf information.The main research contents and conclusions are as follows:(1): Canopy leaf density measurement method for kiwifruit orchard based on point cloud semantic segmentation.In this paper,a Li DAR-based kiwifruit canopy measurement system was developed,and the Li DAR was used to scan the kiwifruit canopy to establish point cloud datasets.Rand LA-Net neural network model and six-fold cross-validation method was used to semantically segment leaves,branches and T-frame point clouds of kiwifruit canopy.The average OA of the model reaches over 92%,and the average m Io U is 81.4%,which can meet the segmentation requirements of the information of each part of the kiwifruit canopy.The leaf attribute measurement software was developed to calculate the number of leaves and leaf area.The least squares method of weighted minimum residuals was used to perform regression analysis on the real leaf attributes and the surface area of the leaf point cloud.The correlation between the measured leaf area and the surface area of the leaf point cloud was high,the correlation coefficient R is 0.78,and the corresponding relationship was got.Alpha-shape algorithm was used to reconstruct the three-dimensional canopy of kiwifruit and calculate the canopy volume.The canopy area of kiwifruit was divided into some grids,and the prescription information of leaf density was generated by calculating the leaf area and volume of the small grids,which could provide effective decision-making for the variable spray system.(2)Canopy droplet deposition discipline and pesticide application decision-making model construction based on CFD technology.Through the method of numerical simulation,the fluid analysis of the existing orchard air-assisted sprayer was carried out.The droplet deposition results and spraying range of each nozzle of the spraying system on the canopy of kiwifruit were obtained,and related verification experiments were designed.The experimental value of the droplet deposition amount and the simulated value have a high degree of fitting,the correlation coefficient R2 is 0.92,and the RMSE is 0.042.It is proved that the method of using fluid simulation can reflect the real canopy deposition.Based on the canopy leaf area and leaf density prescription data of kiwifruit,a canopy application rate model based on decision coefficient was developed.Aiming at the coincidence phenomenon of nozzle spray in the simulation results,the spraying strategy was adjusted,and the calculation method of the valve opening degree at each nozzle was obtained.(3)The hardware and program design of the variable spray control system.On the basis of the existing orchard air-assisted sprayer,the relevant hardware of the variable spray control system was identified.The installation position and power supply mode of each hardware in the sprayer were determined.Aiming at the problem of insufficient computing power of the upper computer,intranet penetration technology and TCP/IP communication method was used to realize the communication between the semantic segmentation server and the upper computer.The spray strategy program and spray execution program were also designed to ensure the smooth operation of the control system.(4)Performance test of variable spray system based on Li DAR.For the developed variable spray control system,a simulated canopy of kiwifruit was built,clear water was used as the spray solution and the position of water-sensitive paper was set according to the spraying area of each nozzle.Spray experiments were carried out on two kiwifruit canopies with different branch and leaf distributions using constant and variable models.The results of the canopy deposition test showed that the droplet deposition caused by the spraying of the variable group and the constant group all met the requirements of the general test method for plant protection machinery.In the variable spray,the droplet deposition density had a good correlation with the decision coefficient K.Compared with the constant model,the variable model had a pesticide saving rate of 54.3% and 34.8% for the two canopies,and the ground loss was reduced by 75.5% and 38.6%,respectively.