Research On Acquisition Of Target Spray Nozzle Positioning Information And Spray Volume Decision Based On Potato Canopy Phenotype Information

During the application process,the relative position of the nozzle and the potato canopy and the spray amount have a significant impact on the deposition effect of the pesticide on the potato canopy.Unreasonable nozzle position and spray amount will cause the pesticide to be in the non-target area outside the potato canopy.Excessive deposition will harm the ecological environment and restrict the green and sustainable development of my country’s agriculture.In order to reduce the application of pesticides in potatoes and improve the utilization rate of pesticides,based on deep learning and machine vision technologies,this paper studies the rapid acquisition of nozzle positioning information and the decision-making problem of spraying amount during potato application.The main content and conclusions of the study:(1)Potato canopy identification and enclosing circle feature construction.By comparing the recognition performance of different deep learning models on the potato dataset,the YOLOv4model is determined as the potato canopy recognition model,and a canopy enclosing circle construction method is proposed based on the rectangular prediction frame of the YOLOv4 model.The middle potato canopy was identified and the surrounding circle features were extracted,and the obtained results were compared with the calculation results of traditional image processing methods to verify the extraction accuracy of the recognition model for the canopy surrounding circles.The experimental results show that the method of constructing the enclosing circle feature based on the rectangular prediction frame of the YOLOv4 model can extract the enclosing circle feature of the potato canopy,and the average error of the extraction result compared with the traditional image processing method is 5.39%.(2)Research on extraction method of nozzle location information.Aiming at the influencing factors of nozzle positioning,such as canopy three-dimensional coordinates,droplet deposition range and potato plant height,Using the depth camera to obtain the depth image of the potato canopy to extract the three-dimensional coordinates of the center of the circle surrounded by the potato canopy.The functional relationship between the droplet deposition range and the height of the nozzle is constructed by the method of combining the experimental measurement of the droplet deposition range and the data fitting,and the determination coefficient R~2 is 0.996,A plant height prediction model based on potato canopy surrounding circle is proposed.The model determination coefficient R~2 is 0.997,and the average error of plant height prediction model is 5.76%.(3)Canopy spray volume decision study.Use the same spraying time to track the change of the canopy on the potato,collect the image of the deposition and distribution of droplets in the potato canopy,obtain the droplet density and droplet coverage,and explore the optimal spraying time under different canopy sizes.It shows that with the continuous growth of the potato canopy,the spraying time needs to increase accordingly.When the diameter of the circle surrounding the potato canopy is less than 10 cm,the spraying time should not exceed 0.2 s.When the diameter of the circle surrounding the canopy is 10 cm-20 cm,the spraying time should not exceed 0.3 s.When the diameter of the circle surrounding the canopy exceeds 20 cm,the spray time can exceed0.5 s.