Research On Plug Grid Detection And Automatic Seedling Picking Method Based On Line Structured Light

Automated grafting can effectively improve the speed and efficiency of grafting and reduce manual workload.However,automatic grafting has not been able to achieve fully automatic operation at present,and automatic seedling extraction is a difficult point that cannot be overcome at present.At present,the acquisition of plug parameters and its pose relies too much on manual labeling,resulting in a low level of automation in the seedling picking process.Aiming at the low degree of automation in the rootstock seedling extraction process during the grafting process,based on line structured light and computer vision technology,this paper carried out the research on the automatic seedling extraction method of plug rootstock seedlings,which provided a basis for the development of automatic grafting machines.The work of this paper mainly has the following aspects:(1)Three-bit detection based on line structured light.A line laser point cloud data acquisition terminal with rotation function is designed,with a rotation step angle of 1.8°,16 subdivisions,and a rotation range of 90°;a connected domain and bounding box extraction algorithm based on line segments is proposed.The running time of each image of the algorithm is 1145μs,while the traditional eight-connected domain algorithm takes6448μs;a line laser extraction algorithm based on connected domains and bounding boxes is proposed,and the extracted line laser pictures account for 2% of the original image area;A laser line removal algorithm based on a bounding box is proposed,with a success rate of 98.2%;a sub-pixel laser centerline extraction algorithm based on the fusion of the laser edge and the gray center of gravity based on the bounding box is proposed,and the extracted laser centerline is continuous It is better than extreme value method,skeleton refinement method and Steger method in terms of sex,smoothness and stability.The average running time of the algorithm is 766μs,which is less than 163μs of the extreme value method,but shorter than 1786μs of the skeleton thinning method and6293μs of the Steger method.(2)Three-dimensional reconstruction and parameter extraction of acupoint grid.A bounding box-based point cloud data connected domain extraction algorithm is proposed to process the point cloud data of the tray wall of different specifications,and the average time is 2185.3μs,while using the traditional clustering method based on Euclidean distance The average time is 206.27ms;a 3D reconstruction and parameter extraction process of the tray grid is designed,which can obtain the height of the tray,the side length of the upper opening curve of the tray grid,the side length of the bottom curve of the tray grid,and the groove of the tray wall.The distance,the three-dimensional model of the surface of the tray wall,and the straight line part of the surface at different heights.Experiments show that the average error of all parameters extracted by the algorithm is0.68 mm.(3)Recognition and parameter extraction of acupoints and field ridges based on line laser.A 3D point cloud data acquisition algorithm based on the perception domain is proposed.Compared with the direct acquisition of point cloud data,the impurity content of point cloud data is reduced by 19.7% in the whole point cloud data;the point cloud based on the bounding box is used The data connected domain extraction algorithm extracts the connected domains of point cloud data(the average number of point clouds is 40555.7),and the average running time is 9473.7μs,while using the traditional clustering method takes 187.71ms;a process for extracting the parameters of the hole plate field is designed,The success rate of extracting the hole plate field ridge by this process is 92.2%,the average error of the extracted position is 0.152 mm,and the average error of the angle is 0.191°.(4)Design a terminal that can automatically take rootstock seedlings,which can change the distance and rotation angle of the two seedling shovels respectively.The distance range is [35mm,75mm],and the angle range is [0 °,360 ° ].The seedling picking terminal was used to pick seedlings from plug trays at different positions and angles,and the average seedling picking success rate was 87.4%,which met the needs of picking seedlings.