Design And Experiment Of Self-propelled System With Multi-target Variable Dosing

In recent years,the greenhouse facilities are increasingly popular,and the problem of pesticide pollution in the greenhouse has been paid more and more attention.Precision target application technology is of great significance to reduce pesticide pollution and protect workers.However,due to the complex environment of greenhouse facilities and special crop planting methods in greenhouse,precise target application technology usually has the problems of low target identification accuracy,weak robustness of recognition algorithm and low accuracy of target.Therefore,In this paper,cucumber seedlings in the greenhouse were used as research objects to study the precise application of target variables.The main research contents are as follows:(1)A target application system based on up and down machine was designed.The system uses machine vision technology to provide support for crop identification,automatic navigation car as a moving platform,X-Y two-axis lead screw sliding table group as the target robot arm,and mobile spraying sprinkler model to achieve variable application.(2)In order to improve the accuracy of target recognition and the robustness of the algorithm.Firstly,the images of cucumber seedlings under different lighting conditions in the greenhouse were collected to obtain the sample images of plants and backgrounds,and the evaluation index of gray histogram was designed.Secondly,genetic algorithm was used to determine the optimal coefficient of gray-scale operator.Finally,Otsu method and morphological processing method were used to fast threshold segmentation and noise reduction of gray image.The experimental results shows that the average F1 value of the binary images was more than 96.5%,and the average false positive rate was kept within 0.8%.The image segmentation shows strong robustness.(3)The connected fields in crop binary graph were extracted by region labeling method and the application points were determined according to the center of the minimum peripheral circle of connected fields.The image acquisition model of moving vehicle was established to remove repeated application points.According to the moving speed of the vehicle,the point of application was marked in real time.According to the moving speed of the vehicle and the ladder speed curve,the trajectory control of the nozzle was realized.(4)In order to verify the performance of the application system,the variable application model of sprinkler head moving was tested in the greenhouse environment.The target performance of the system was better at different vehicle speeds,and the accuracy of the recognition rate and the success rate of the target were both above 97%.The experimental results showed that the dosage saving rate could reach more than 57%.After verification,the target application platform designed in this study could meet the demand of precise target variable application,and its crop recognition algorithm has strong robustness,high target accuracy,and can achieve a higher rate of liquid medicine saving.