Soybean Leaves And Seeds Phenotype Parameters Measurement Platform For Soybean Yield And Quality Detection

As the population continues to increase,the demand for food is increasing.Soybean is one of the most important economic crops and the main source of human edible oil and protein.Studying the geometric dimensions of soybean leaves is of great guiding significance for improving soybean planting.Studying leaf pest conditions and color parameters can increase the detection index of soybean seeds quality,the number of seeds per plant,seeds length,seeds width and seeds area,perimeter,etc.The parameters are often used to evaluate the yield and quality of soybeans.This paper presents a method to study the geometric dimensions,pest parameters,and color parameters of isolated soybean leaves and the non-destructive measurement methods of the geometric dimensions and leaf inclination of living leaves,as well as soybean seeds count,area,perimeter,seeds length,and seeds width measurement method.In order to promote the efficiency of soybean production,soybean quality,soybean breeding etc.This paper developed a soybean leaves and seeds phenotype parameter measurement platform for soybean yield and quality detection.For the measurement of phenotype parameters of isolated soybean leaves,a flat-panel scan was used to obtain color images of the leaves.For pests on the leaf boundaries,the Graham convex hull algorithm was used to reconstruct the leaf boundaries.Based on the reconstructed binary image of soybean leaves,the geometrical size parameters such as leaf area,perimeter,length,width,pest area,and leaf loss rate are calculated using statistical moments of the image.The absolute error values of leaf perimeter,length,and width are 0.854cm,0.0505cm,and 0.052cm respectively,and the relative error of wormhole area is less than 5 percent.Using the RGB color values in the leaf scan diagram to determine whether the leaves are healthy.The leaves with a standard deviation of less than 20 are pure color leaves.When the standard deviation of the R value is about 40,they are pure color leaves.The color parameters of the pure color leaves are the average value of R and G color values.The color parameter of impure color leaves is the proportion of yellow pixels.For non-destructive measurement of soybean leaf phenotype parameters,SR300 camera was used to obtain leaf color and depth image sequences.The interference background was removed based on converting RGB image to LAB image.Depth threshold was automatically set to obtain canopy leaf binary images.The watershed transform method was used to segment the sticky leaves in the binary image of the leaves.For the connected domain of a single leaf,the area of which was calculated using the surface integration algorithm based on triangulation.Converting the pixel coordinates of the leave pixel points into world coordinates.The leaf perimeter was calculated based on the depth image of the leaf boundary contour,the three-dimensional coordinates of the border pixels and the spatial distance formula.The leaf length and width was calculated based on the statistical moments and the three-dimensional coordinates of the leaf endpoints.The angle between the leaf and the vertical direction was calculated based on the leaf imaging principle.Using potted plants and three soybean varieties in the field to verify the applicability of the measurement software.The results showed that the measurement errors of leaf area,perimeter,leaf length and leaf width were 1.020cm2,1.540cm,0.810cm and 0.820cm respectively.For soybean seeds counting and morphological parameters measurement,color images of soybean seeds and calibration plates were collected by the Liangtian camera.The image background was removed based on the "Otsu" threshold method to obtain a binary image of soybean seeds.The watershed transformation method was used to segment adhesion soybean seeds to obtained a series of single-seeded,multi-seeded soybean connected domains.Correcting the seed count for a few multi-seeded connected domains.The soybean seeds area was calculated by using the sum of pixels of single-seed connected domains.Based on freeman chain code algorithm,the correction formula was used to calculate the perimeter of soybean seeds.Using the second-order statistical moment and the difference between the horizontal and vertical coordinates of the soybean boundary points to obtain the seed length and seed width.The accuracy of the software was verified with three copies of soybean materials of different sizes.The results show that the accuracy of soybean seed counting is 100 percent.The average error of seed length and average width of the soybean seeds measured by the software and the manual measurement is generally 0.01cm to 0.04cm.The average relative error is 5.6 percent.The measurement platform can effectively meet the accuracy requirements of agricultural scientific research work and low cost.Compared with manual measurement,which takes time and effort,the platform greatly improves work efficiency.