Study On Automatic Measurement Method Of Soybean Leaf Area Based On Three-dimensional Reconstruction

Cultivation is the first priority,and good variety is the premise of high yield.Phenotypic studies of soybean crops are crucial for the development of soybean breeding,precision cultivation and climate change monitoring technologies.Soybean needs to continuously produce organic matter through photosynthesis to ensure its growth.As the main organ of the aboveground part of soybean,the growth status of leaf has a direct impact on the photosynthesis of soybean plant.Therefore,the accurate measurement of soybean leaf area is of great help to the related research of soybean breeding.This study took soybean,an annual herb crop,as the research object,and carried out soybean planting and sample collection in the experimental field of the School of Software of Shanxi Agricultural University in Wujiapu Village,Taigu County,Jinzhong City,Shanxi Province.The method of automatic leaf area measurement of soybean plant point cloud data was studied by combining theory with practice,and the method based on image processing technology was used to measure the parameters of soybean leaf area,which verified the accuracy and effectiveness of this research algorithm.The main research contents are as follows:(1)An image acquisition platform of soybean plants based on Motion 3D reconstruction technology was built,and the point cloud data of soybean plants were obtained by using SFM(Structure From Motion)and MVS(Multi View Stereo)algorithms.(2)the first to use the method of data consistency ensure removal of soybean plant has nothing to do with the object of study in three-dimensional point cloud data of point cloud,then adopted the direct filter and statistical filter of soybean plant point cloud data to deal with the noise,finally using the based on voxel grid sampling method of soybean plant under simplified point cloud data.(3)Firstly,the point cloud segmentation method based on the region growth algorithm of normal vector was adopted to realize the rapid and automatic segmentation of soybean leaves.Then,the color space conversion based on image processing technology and the threshold segmentation method were adopted to segment soybean leaves.In the end,the hole filling method based on digital morphology was studied to fill the holes in the segmented soybean leaf image.It laid a foundation for the implementation and verification of leaf area measurement methods in the future.(4)Firstly,the resampling method based on the moving least square method was adopted to smooth the point cloud data of soybean leaves.Since it is impossible to directly measure the area of soybean leaf point cloud data,the greedy projection triangulation method is adopted to reconstruct the surface of soybean leaf point cloud data,and then the area of soybean leaf is measured by measuring the size of the grid model reconstructed by the surface.At the same time,the measurement method of soybean leaf area based on image processing technology was used to measure the soybean leaf area,and it was used as the reference value to compare the values obtained in the experiment.The experimental results show that the point cloud segmentation method based on regional growth in this paper can effectively segment soybean leaves,and the average correct segmentation rate can reach95.28%.The measurement method of soybean leaf area parameters based on soybean three-dimensional point cloud data used in this paper can automatically and accurately calculate the soybean leaf area parameters with an average error of 12%.Therefore,it provided a feasible way for breeding excellent soybean germplasm.