Research On Crop Phenotypic Parameters Acquisition And Dynamic Quantification Method Based On Multi-temporal Point Cloud Data

Genome sequencing of plant species reveals their genetic information,while plant phenotypic studies reveal the effects of interaction between genome and environment on plant traits.Phenotypic analysis and genotypic analysis are equally important in establishing the relationship between genes and plant traits.Plant characteristics can be determined by phenotypic analysis,however,many current plant phenotypic studies need a large amount of artificial participation,and only at a specific stage of plant growth can not be carried out to study the continuous growth changes of the plant,and there are deficiencies in the analysis process and result evaluation.At the same time,in order to cope with the challenges brought by the rapid growing world population and global environmental change,it is necessary to select crop varieties that are more suitable for limited resource environment and soil conditions through genotypic and phenotypic analysis,so as to improve the yield and quality of crops.Therefore,efficient,accurate and non-destructive measurement of crop phenotypic parameters and dynamic quantification of crop continuous growth and phenotypic parameters are urgent problems to be solved in crop phenotypic research and breeding,as well as the trend of crop phenotypic research and modern agricultural development.With the development of crop phenomics and phenotypic technology,threedimensional laser scanning technology has gradually attracted the attention and interest of high-throughput phenotypic measurement researchers.Therefore,rapeseed and cotton were taken as research objects in this paper,and the multi-temporal point cloud data of rapeseed and cotton plants were acquired by three-dimensional laser scanning technology,the main stem extraction,leaf segmentation and related phenotypic parameters measurement were realized,completes the point cloud data alignment of crop in adjacent periods,leaf organ correspondence and dynamic quantification of phenotypic paramaters in adjacent periods were completed.The research contents in this paper include:(1)point cloud data acquisition: comparison of rapeseed point clouds collected based on multi-image reconstruction and three-dimensional laser scanning technology;in the later study,three-dimensional laser scanning technology was used to obtain multi-temporal original point cloud data of rape and cotton plants,and pre-process point cloud data;(2)crop organ segmentation:(1)stem extraction,the stem of rapeseed in budding stage and cotton tends to be straight,so RANSAC algorithm combined with linear model is used to extract the main stem of crops;(2)Leaf segmentation,in the seedling stage,the rapeseed did not grow out of the main stem and the plant was in the shape of a rosette,the leaf segmentation of the rapeseed point cloud in seedling stage was carried out based on the method of concavity and convexity,the point cloud after extracting the main stem of cotton and rapeseed in budding stage uses Euclidean Cluster Extraction algorithm to comlete the leaf clustering;(3)Establishment of multitemporal point cloud data alignment and leaf correspondence:use sample consensus the initial alignment algorithm to complete the point cloud alignment in the adjacent period of the same crop,and establish the corresponding relationship between the adjacent period of crop leaf organs on the basis of alignment;(4)Quantification of phenotypic parameters includes:(1)plant height,the relative error between the plant height of point cloud computing and the manual measurement results does not exceed 2.5%.;(2)the number of leaves,the number of statistical leaves after crop leaf segmentation is consistent with that of manual statistics;(3)length and width of leaves,the length and width of leaf in point cloud computing are highly correlated with the length and width of leaf measured manually,and their correlations are more than 0.99;(4)leaf area,leaf area is obtained by surface fitting of the encapsulated blade;(5)the volume of plants,for the plant point cloud of each period,the plant is orthographically projected from the top of the plant and the projected area is calculated,and the plant volume measurement is completed in combination with the plant height.The phenotypic parameters of adjacent crop periods were compared to quantify the dynamic changes of crop phenotypic parameters.Through a series of experiments,it has been proved that the relevant algorithms in this paper can extract crop phenotypic parameters nondestructively and accurately,and realize dynamic quantification of crop phenotypic parameters,so as to provide methods and more accurate and large-scale data support for phenotypic studies of rapeseed,cotton and other crops,and guide actual crop production.