| Agricultural informatization is the commanding point of agricultural modernization.Promoting the comprehensive and deep integration of information technology and agriculture is an important strategy to improve grain yield and quality.Agricultural informatization is closely related to crop high flux phenotype information acquisition.Accurate and rapid acquisition of crop phenotypic parameters can reflect the status of crop growth and development,varieties of quality,nutritional status and other information.It is a technical means to stabilize grain production and ensure food security.The research on the growth process of crops can understand the dynamic changes of the morphological structure of crops,understand the common influence of the environment and genes on crops,and provide assistance for the development of the seed industry and the development of major varieties.Therefore,the dynamic simulation of growth process based on accurate crop phenotypic information is of great significance for the formulation of reasonable planting scheme,excellent crop breeding and precise crop cultivation.Traditional crop morphological parameter measurement and growth process research mostly is studied by adopting the method of field measurements,which have problems such as process complicated,time-consuming and labor-intensive,and strong subjectivity.Through 3D laser scanning technology can accurately obtain the 3D laser point cloud of crop population,accurately measure the morphology parameters of crops,study the whole growth period of crop growth process,and realize the dynamic simulation of crop growth process.Maize is an important grain and industrial raw material crop in China,which is closely related to the improvement of grain yield.The cultivation of high-quality maize varieties is of great significance to food security.Obtaining accurate morphological parameters of maize and realizing dynamic monitoring of morphological parameters during maize growth are of great significance to the precise breeding,cultivation of improved varieties and phenotype research of maize.Based on 3D laser scanning technology,this paper took the maize population in the farmland environment as the research object,and conducted research on the registration,simplification,target extraction and segmentation of the 3D point cloud of the maize population,measurement of the morphological parameters of the maize population,and dynamic simulation of growth.Through the research,the extraction and refined measurement of the three-dimensional target of maize in the field were realized,the growth changes of maize morphological parameters were analyzed,and the growth dynamic simulation model of maize morphological parameters was established.The main work and achievements of the thesis include the following contents:(1)Maize population point cloud data collection and pretreatment: based on the maize field test scene,a collection plan for maize population point cloud data was formulated to obtain the complete maize population point cloud,and the maize population point cloud was preprocessed.The target ball information is the basis for registration.The point cloud registration method based on the target ball realized the registration between different station clouds.The registration error was between 1.81 mm to 3.70 mm.The point cloud data was simplified to about 20% of the original data by the curvature sampling method,and the geometric structure characteristics of crop population were retained.(2)Maize population point cloud data segmentation: based on the ultra-green index method,the separation accuracy of maize plants and soil background was achieved,and the separation accuracy was between 88.87% to 98.79%.A single plant segmentation method combined with cylinder space and connected region analysis was proposed to realize the segmentation of maize single point cloud data in the maize population point cloud.On this basis,a method of leaf and stem segmentation based on RANSAC cylindrical fitting was proposed to realize the separation of maize leaf and stem.(3)Measurement of plant morphological parameters of maize population: the morphological parameters of maize plant height,stalk height and width,leaf area density and leaf angle at different periods were measured based on the segmentation results of the maize population point cloud.RANSAC algorithm was used to fit the soil plane and calculate the vertical distance between maize plant apex and soil plane,so as to realize the measurement of maize plant height.The measurement of stem height was realized based on the RANSAC cylindrical fitting results.The ellipse fitting method based on the least squares method can measure the stalk width.The leaf area density can be measured based on the three-dimensional voxel method.Leaves were extracted based on RANSAC cylindrical fitting and Euclidean distance clustering to measure the leaf angle.The accuracy of the measured morphological parameters was evaluated,and the linear regression analysis was conducted with the measured data.The determination coefficient of the four parameters of plant height,stem height,stem width and leaf area was above 0.9873,RMSE was within 0.0617,and the leaf angle was 0.8721,RMSE was 2.2734,which showed strong correlation and small error.(4)Dynamic simulation of maize population growth: Through the measurement of various morphological parameters of maize in the field during the whole growth period,the growth process of maize was studied,the dynamic changes of maize morphological parameters and the dynamic changes of leaf area in vertical structure were analyzed,and the dynamic quantification of maize morphological parameters were completed.A dynamic simulation model of field maize morphological parameters was established to realize the full cycle growth dynamic simulation of field maize.The established growth dynamic simulation model was tested,and the measured values were analyzed by linear regression analysis.The of the dynamic simulation model of maize morphological parameters was above 0.9568,the root mean square error(RMSE)was within 0.0078,and the model accuracy EF was above 0.9781.The fitting effect was good,and the overall accuracy was high. |
PDF Full Text Request