Research On Calculation Method Of Soybean Canopy Structural Parameters Based On Three-dimensional Point Clou

China will vigorously implement the soybean and oil productivity upgrading project to promote the improvement of soybean self-sufficiency rate("the 14th Five-Year"Plan).Soybean affects food security from the whole industrial chain,increasing soybean output can effectively alleviate food security in the future.At present,the primary goal of breeding and actual production is to select and cultivate crop varieties with excellent characters according to the relevant phenotypes of crops.Due to the problems of great subjectivity,difficulty and error prone in the traditional measurement of plant phenotypic parameters,there is an urgent need for an efficient,low-cost and nondestructive phenotypic extraction method.With the soybean in cold regions as the research object,a method of measuring soybean canopy structure parameters was proposed based on three-dimensional point clouds.The main research work and results were as follows:(1)For soybean canopy point cloud extraction,the combination of grid method,depth threshold filtering and statistical filtering were used to finish the extraction without information about the color.The vegetation point cloud was extracted by the combination of excess green and the otsu method.After rasterizing it,the soybean canopy pixels were extracted by two-dimensional image processing technologies including depth,area and spatial information,and projected back to three-dimensional space to realize the extraction of soybean canopy point cloud with color information.Against the complex background,the canopy segmentation performance of the two methods was evaluated from three aspects:extraction speed,segmentation accuracy and parameter calculation accuracy of soybean canopy point cloud.Among them,the extraction with color information had a better result,with its extraction speed up by 91%,the segmentation ratio of clean canopy up by 9.38%and the mean absolute error down by 0.005 m.(2)For measuring soybean leaf inclination parameter,the overall least square method was used to fit the plane of leaf point cloud and finally the measurement of single leaf inclination was realized by calculating the angle between its normal vector and zenith axis.The mean absolute error of single leaf inclination was 16.54°,and the determination coefficient R~2was 0.6746.K-means clustering algorithm was used to segment the single leaf point cloud of soybean canopy and Delaunay triangulation algorithm was used to reconstruct the leaf surface mesh model and calculate its area and projected area to finally measure the group leaf inclination.The mean absolute error was 0.06 and the determination coefficients R~2 were 0.8317,0.9075 and 0.9186respectively.The calculation method of group leaf inclination could better reflect the real situation of leaves in the process of soybean growth.(3)For measuring the size parameters of soybean canopy,through projecting the three-dimensional point cloud of soybean canopy,the extreme points were identified and the height of soybean canopy was calculated.Also the canopy width was calculated by establishing the bounding box of projection point cloud and the projected area and volume of soybean canopy were measured by using Concave hull,Alpha shape and Convex hull algorithms.The determination coefficient R~2 was more than 0.88.At the same time,the change rule of soybean canopy size parameters with time in different growth periods was analyzed to verify the effectiveness of the method.Finally,the relevant algorithms were encapsulated to realize the development of soybean canopy structure parameter calculation system.