| Grafting is a technique that gives vegetables the ability to overcome the obstacles of continuous cropping caused by the continuous cropping of vegetable bases,and to increase vegetable yield,cold tolerance and resistance to soil-borne diseases.With the rapid development of computer technology,machine vision and artificial intelligence,the field of agricultural production is also constantly developing towards automation under the influence of these technologies.Most existing cucurbitaceous grafting machines can only automate the grafting process,and the seedling loading process still needs to be completed manually.Therefore,the realization of automatic loading of seedlings for the grafting machine has become a problem that needs to be solved urgently,and it is the key technology to realize the whole process automation of the grafting machine.Aiming at the problems of cucurbitaceous grafting machine requiring manual seedling loading,heavy workload and low grafting efficiency during the grafting process,this paper studies and designs a machine vision-driven automatic seedling loading device for plug rootstock seedlings,which provides the basis for the development of a fully automatic grafting machine.The research work of this article mainly includes the following aspects:(1)Rootstock seedling vision acquisition system and image preprocessing.According to the actual working environment and the actual growth parameters of the seedlings,the camera and lens parameters are selected,and a simple visual acquisition system is formed with RaspBerry Pi;the color characteristics of target pixels such as cotyledons,stalks,and stem whites are studied and analyzed,and the color space of HSV is determined,the expression range of the target pixel is constructed by multi-threaded line scanning method,and the image is denoised and the contour chain of the target image is extracted through the relationship analysis of the connected domain.The average preprocessing time is 33ms;the experiment analyzes the environmental light intensity and influence of image noise,preprocessing speed,and target image contour extraction results determine the optimal ambient light intensity to be 150Lux to 250Lux.(2)Research on the visual extraction method of rootstock seedling grafting parameters.According to the analysis of grafting agronomic requirements,the orientation angle and the longest section width of the cotyledons of the rootstock seedlings were extracted,and the top and side view positions of the growth points were positioned and estimated.First,the traditional minimum circumscribed circle algorithm is improved.The improved algorithm is 7 times faster than the original traditional algorithm,with an accuracy rate of 99.2%.The improved minimum circumscribed circle algorithm is used to extract the direction angle parameters of the rootstock seedling cotyledons with an accuracy rate of 99.55%;The flip method was used to extract the longest cross-section width of the cotyledons of the rootstock seedlings,with an average error of 3.2 pixels.Multi-method fusion is used to estimate the location of the growth point.When the cotyledons are bonded,the chain angle span of the contour chain is gradually enlarged,and the corner positions are explored to locate the growth point.The average positioning error is 2.6 pixels;The Zhang-Suen method is used to extract the target cotyledon skeleton,and the skeleton intersection point is used to locate the growth point,the extracted top-view growth point position average error is 3.9 pixels,and the side-view growth point position average error is 5.5 pixels.(3)Design and analysis of seedling loading device based on mechanical arm.First,a terminal mechanism for fetching seedlings is designed,the terminal can detect the edge of the plug tray,and can realize the action of clamping and lifting the seedlings,the mechanical arm is improved and designed,one is to increase the arch structure,the large and small arm motors arranged on both sides are changed to the same side arrangement,which reduces the moment of inertia,the second is to improve the design of the terminal self-holding mechanism,which increases the rotation range of the forearm from 320° to 360°,and experimentally analyzes the self-holding level of the terminal ability,the average deflection angle of positive rotation is 0.878°,and the average right deflection angle of negative rotation is 0.317°;The forward and inverse kinematics equations of the robotic arm are established by the DH method,and the rotation center of the robotic arm terminal to the base is analyzed through experiments,the installation deviation of the robot arm is 2.528mm,after the calibration is corrected,the positioning error of the robot arm is corrected from 1.984mm to 0.264mm.(4)The control and experiment of the device on the plug seedling.In the hole tray center positioning experiment,a robotic arm is used to drive the terminal probe to detect the edge of the tray,the effective positioning area accounts for 95.3%,the average positioning error of the tray center is 0.302mm;In the experiment of seedling and seedling lifting,under different clamping speeds of the seedling splint,the robot arm is coordinated to control the insertion speed of the seedling terminal,the experimental result shows that the average movement deviation of the splint during the insertion and seedling removal process is 0.75mm;In the terminal angle adjustment experiment,the rotation angle of the robot arm terminal platform motor is adjusted under the driving of the machine vision from the top view,the experimental results show that the maximum deviation of the horizontal rotation adjustment angle of the seedling terminal is 2.18°,the minimum deviation is 0°,and the average deviation is 1.07°. |
