Three Dimensional Phenotypic Measurement Method And System Design Of Maize Plants In The Field

Maize plant height,leaf width and leaf angle can reflect the information of maize plant growth,resistance to overturning and yield.The acquisition of maize plant parameters is an important part of maize precision breeding and cultivation research.At present,the acquisition of crop 3D information in complex environment is mainly done by three ways:multi-vision camera,depth camera and lidar.Multi-eye cameras can simulate the human eye to observe plant phenotypes.Multi-eye cameras can also lead to a dramatic degradation of algorithm effect under strong and dark lighting conditions,while depth cameras can better solve such problems.Depth cameras have problems such as narrow measurement range,high noise,small field of view,susceptibility to daylight interference,and inability to measure transmissive materials.Lidar uses laser signal as transmission signal,which has the advantages of high resolution,high sampling rate,good concealment,strong anti-interference ability,good low-altitude detection performance,and portability.In this thesis,we propose a 3D point cloud-based measurement method for maize plant height,leaf width and leaf angle in the field using lidar as the measurement sensor to address the problems of low automation of existing measurement methods,long measurement time and environmental interference in leaf measurement.A fast,automatic and accurate maize phenotype information is provided for breeders.The main research work and results accomplished are summarized as follows1)In this thesis,a "maize phenotype reconstruction method based on the plant protection vehicle platform".First,through the sixteen-line laser radar installed in front of the platform,the three-dimensional point cloud data of maize plants is collected.Secondly,multi-frame point cloud stitching is performed on the collected point cloud data to enhance the three-dimensional phenotype information of plants.Then,the ground point cloud is removed using a random sampling consensus method.Finally,noise points are removed by the filter method of nearest neighbor search,so that the 3D point cloud reconstruction of field maize is carried out.2)In this thesis,a "maize phenotype segmentation method based on spatial projection and European clustering".First,after completing the 3D point cloud reconstruction of maize,use the point cloud processing tool to extract the single maize point cloud.Secondly,through the maize stalk axial correction method,the stalk direction of a single maize plant is corrected.Thirdly,the method of spatial projection,density analysis,normal distribution,confidence interval and cylinder fitting is used to extract the point cloud of maize stalks.Then,use the Euclidean clustering method to realize the segmentation of maize leaves.Finally,the data of plant height,leaf width and leaf angle of maize were obtained by spatial projection and linear fitting.3)In this thesis,a "Design and research method of maize phenotype measurement system based on.Net platform".First,analyze the system users to obtain the functional and non-functional requirements of the system.Secondly,carry out the work of system architecture design,function module design and database design.Then,carry out function development and page design for different modules.Finally,in order to verify the feasibility of the system,the automatic measurement of the plant height,leaf width and leaf angle through the maize point cloud was compared with the manual measurement,and the average absolute error percentages were 1.34%,10.32% and 13.5%,respectively.The results show that the maize point cloud measurement system has less error than the traditional manual measurement method,and can improve the measurement efficiency and accuracy,so it has a wide application prospect in agricultural production.