| With the rapid development of plant tissue culture technology,the method of breeding seedlings has been widely used in the seedling production of trees and economic crops,and has become one of the main ways to obtain high-quality seedlings of various plants.With this method,the seedlings can be cultivated quickly and multiply.To grade the tissue culture seedlings before they harden and then to implement differential management of water and fertilizer according to the growth state of different groups can achieve homogenization of seedlings quality by avoiding abnormal growth of the weak seedlings beside the large ones.At present,the grading of tissue culture seedlings are generally handled manually.The manual operation is unstable and inefficient,and the labor cost increases year by year as well,so it is very urgent to develop automated grading equipment for tissue culture seedlings.Aiming at this,the key technologies of automatic classification of tissue culture seedlings of spathiphyllum floribundum“Vicki”,a kind of potted flower produced in quantity,have been studied in this thesis.Taking into account the characteristics of spathiphyllum floribundum tissue culture seedlings,such as tender stems,overlapping leaves and irregular seedlings,a fast and non-destructive on-line classification method based on machine vision has been proposed.The work involved in this thesis includes:（1）With reference to the national standard for grading of flower seedlings in the same family,the visual grading indices of spathiphyllum floribundum tissue culture seedlings and the weight coefficients of them have been specified.it is specified that the grading indices of spathiphyllum floribundum tissue culture seedlings are leaf area,diameter at ground,crown diameter and seedling height,and the weight coefficients of the above indices are subjectively set to 0.30,0.25,0.25 and 0.2 according to the actual experience of seed production enterprises.A hierarchical modelη=a×Z1+b×Z2+c×Z3+d×Z4is established,where a,b,c,d are the weight coefficients of leaf area,diameter at ground,crown diameter and seedling height of the seedlings,while Z1,Z2,Z3and Z4their grades respectively,andηis the comprehensive grade of the seedlings.（2）Through the analysis of the morphology of tissue culture seedlings,the measure and acquisition method for the grading indices of them has been determined.Based on the established grading model of spathiphyllum floribundum tissue culture seedlings,the four indices,i.e.leaf area,diameter at ground,crown diameter and seedling height,of the spathiphyllum floribundum tissue culture seedlings need to be measured by machine vision.Leaf area is a very important grading index,but because the leaves are delicate and overlapping,it is difficult to directly measure the leaf area by machine vision without any contact or damage.In this thesis,the geometrical morphology of tissue culture seedling samples has been studied.The results show that the leaf area y1of spathiphyllum floribundum tissue culture seedlings is linear with respect to the overall projection area x,with an expression as y1=0.8113x+9.2473,and the regression coefficient R2is 0.9344;and it is found that the diameter at ground has a polynomial function with the overall projection area,and the regression coefficient R2is 0.9067.We have also found that both the crown diameter and the seedling height have weak correlation with the overall projection area.The above studies show that the leaf area and diameter at ground of the spathiphyllum floribundum tissue culture seedling can be determined by measuring the overall projection area,while the crown diameter and seedling height need to be measured separately.（3）Through the research on the machine vision model of each grading indices of Grade 1 spathiphyllum floribundum tissue culture seedlings under standard nursery environment,the machine vision measurement algorithms of them have been constructed.Based on the COGNEX machine vision system,the automatic measurement of the overall projection area,crown diameter and seedling height of the spathiphyllum floribundum tissue culture seedlings has been realized.A method for extracting the overall projection area using the color template matching algorithm is proposed.Taking the stalk color and the root color as the color template,the spathiphyllum floribundum tissue culture seedlings can be better separated from the background when the color depth is 0.4,the brightness is1.5,and the color tone is 2.0.In this case,relative errors of the maximum,the minimum,the average and the standard deviation of the leaf area are all within 5%,and relative error of the variation coefficient is 0.35%;while the highest value among the maximum,the minimum,the average and the standard deviations of diameter at ground is 4.91%,with relative error of the variation coefficient being 1.77%.The relative errors of the maximum,the minimum,the average and the standard deviation of the crown diameter and seedling height measured by the minimum rectangle algorithm（MBR）are all within 5%,with a relative error of the variation coefficient of 0.4%.Using the constructed machine vision algorithm,based on the weight coefficient of each grading index given by the seedling production enterprises,the spathiphyllum floribundum tissue culture seedling classification test has been carried out,and the results show that the classification accuracy for Grade 1,Grade 2,and Grade 3 seedlings reaches 100%,94.74%,and 90.00%respectively,and the overall classification accuracy is 96.7%.（4）According to the results of visual monitoring during seedling stage of Grade 1spathiphyllum floribundum tissue culture seedlings under standard nursery environment,the growth curves of the geometric characteristic parameters of the spathiphyllum floribundum seedlings have been extracted and the growth functions have been established.By now,the weight coefficients of the spathiphyllum floribundum seedling grading indices are given subjectively,which may lead to one-sided grading results.In order to accurately reflect the importance of each index on the grading and determine its weight coefficient,it is necessary to further investigate the growth trend of tissue culture seedlings during the“post-grading”period.In this thesis,the growth state images of the sample spathiphyllum floribundum seedlings were taken every 10 days during the nursery period through machine vision,and the growth curves and growth functions were analyzed statistically.The overall projection area,diameter at ground,crown diameter,and seedling height of the seedlings,as well as their geometric characteristic parameters such as the angle between petiole and stem,the length of petiole,the diameter of petiole and the leaf area,are represented as quadratic polynomial functions of time,and the regression coefficients are all above 0.85.（5）Based on the growth functions of the geometric characteristics of spathiphyllum floribundum tissue culture seedlings,a simulation model of the growth state of the seedlings has been constructed.This simulation model provides external structure information of seedlings in different growth periods,thus can be used in the development of seedling automation equipment,and in providing seedling quality evaluation standards for nursery users.According to the growth functions of each geometric characteristic parameter of tissue culture seedling,the production way of three-dimensional model of spathiphyllum floribundum seedlings has been determined.By using the parameter L system and taking the growth time as the independent variable,the dynamic geometry parameter growth model of the seedlings has been established.The number and distribution of the branches and leaves generated from the three-dimensional model are consistent with the growth state of the actual seedlings,and the highest similarity of them is 86%when comparing with actual seedling by machine vision template matching algorithm.（6）The weight coefficient and classification model of each grading index of spathiphyllum floribundum tissue culture seedlings have been determined based on the rough set theory.In order to correct the subjective weight coefficients,according to the growth curve of the geometric characteristic parameters of spathiphyllum floribundum tissue culture seedlings,the weight coefficient of each grading index has been determined based on the expert weighting method and the rough set theory.The results show that the four grading indices can be listed in order of importance as leaf area,diameter at ground,crown diameter and seedling height,and their weight coefficients are 0.35,0.27,0.23 and0.15,respectively.Compared with the subjective values,the weight coefficients of leaf area and diameter at ground increase by 0.05 and 0.02,while those of crown diameter and seedling height decrease by 0.02 and 0.05,respectively.With these corrected coefficients,the classification accuracy for Grade 1,Grade 2,and Grade 3 seedlings is improved by 0%,2.63%,and 5%respectively,and the overall classification accuracy by 1.97%.（7）Based on the above key technologies,a prototype of on-line grading equipment for spathiphyllum floribundum tissue culture seedlings has been constructed.The prototype consists of tissue culture seedling automatic conveying device,machine vision grading system and gas jet grading collection device.The performance test of the prototype shows that,under the condition that the conveying distance between two tissue culture seedlings is0.25m and the conveying speed is 0.5m/s,the productivity of the classification operation could reach 7200 plants/h and the classification accuracy rate could be above 96%. |
PDF Full Text Request