Research On Machine Vision Based Graft-Seeding Positioning Method And End-Effector Design

In this paper,the problem of manual operation and inability to dynamically adjust the grafting process of semi-automatic vegetable grafting machine is designed.A machine vision-driven grafting device for toaster crops is designed to provide a basis for improving the automation level of the grafting machine.The innovations of this research are as follows:(1)According to the agronomic requirements analysis,several key parameters of grafted crops are proposed,and the automatic identification and localization methods of key parameters in grafted seedling images are proposed.(2)Three dynamic adjustment grafting execution terminals were designed to realize the grafting of Solanaceae.The device designed the rootstock seedling and the scion seedling cutting execution terminal,and the scion seedlings execution terminal.The work of this study is mainly as follows:(1)Research on image processing methods of grafted seedlings.In this study,the image of grafted seedling leaves,true leaves,stems and root pixels in RGB and HSV controls was studied for the grafted seedlings,and the target pixel extraction method of grafted seedlings was studied to realize the extraction of connected domains and connectivity of grafted seedlings.The fast noise reduction of the grafted seedling image is realized by controlling the size of the connected domain.The multi-threaded grafted seedling image processing method was studied.The average time spent collecting images and extracting seedling contour chains through the embedded platform was 6.5ms and 11.5ms,respectively.(2)Study on extraction methods of key grafting parameters of grafted seedlings.According to the agronomic requirements of the grafting of Solanaceae crops,four key grafting parameters of root,cotyledon,true leaf position and cutting angle were determined.Through the analysis of the basic data such as the contour chain,horizontal intercept and seedling diameter of the grafted stem,cotyledon,true leaf and root position,the extraction method of key grafting parameters was studied,and the automatic positioning of the cutting point position and angle was realized.Several influencing factors causing the positioning error are analyzed through experiments.(3)Dynamically adjusted grafting execution terminal design.According to the characteristics of the Solanaceae crop and the grafting process,three kinds of execution terminal mechanisms of the cutting base of the rootstock seedling,the terminal for picking seedlings and the cutting terminal of the seedlings were designed.In the cutting terminal mechanism,the side view and the overhead view camera are designed to collect the true leaves and stems of the grafted seedlings,and drive the cutting blades to respond to the rotation,so as to realize adaptive grafting cutting of grafted seedlings with different growth forms.(4)Control and experiment of grafting devices.Based on the Raspberry Pi platform,the machine vision-based grafted seedling image analysis and key parameter detection are implemented.The Arduino DUE is used as the platform to realize the control system of the grafting execution terminal.The system includes the three-axis drive of the robot arm and the horizontal orientation of the grafting terminal.Cutting angle and drive control of the cutter.Through the side-view camera of the scion seedling cutting mechanism,the contour chain of the grafted seedlings was detected from far to near,and the change of the grafting chain of the grafted seedlings during the moving process was observed to realize the automatic positioning of the cutting point and the cutting angle of the grafted seedlings.