| Kidney beans have high economic and nutritional value,and are also important crops for intercroppingand rotation.Phenotypic traits such as plant height,stem diameter,and leaf area change with the progress of kidney bean growth,exhibiting different changing characteristics at different sowing times and environments.These phenotypic traits are of great significance for the study of kidney bean cultivation models,selection of excellent varieties,growth process management,and yield prediction.However,manual measurement of phenotype has issues such as strong subjectivity,high labor intensity,and damage.The three-dimensional morphological structure of plants is a true reflection of their growth and development status.Accurately establishing a three-dimensional model of plants and segmenting plant organs based on the three-dimensional model to obtain phenotype parameters is currently a hot topic in plant phenotype research.This article focuses on two types of kidney bean plants(cloud kidney bean 1 and cloud flower kidney bean 2)at the young seedling stage.Based on the Structure from Motion(Sf M)algorithm,a three-dimensional model of the plant is reconstructed,and phenotypic parameters such as plant height,stem diameter,leaf area,leaf projection area,and minimum bounding box volume per plant are calculated based on the three-dimensional model,achieving non-destructive and accurate measurement of phenotypic parameters.This study provides a scientific and convenient method for low-cost and efficient acquisition of kidney bean phenotypes,and provides a reference for crop 3D reconstruction research and plant phenotypeomics research.The main research content and results of this article are as follows:(1)Use a regular RGB camera to capture images of kidney bean plants at three different heights,and manually surround the plants at set intervals at each height to capture multi view images.Shoot 72 images at a 15°surround interval,108 images at a 10°interval,and 216 images twice at a 10°interval to establish different multi view image sequences.Using Visual SFM to reconstruct a three-dimensional model of kidney bean plants,evaluate the reconstruction effect of different image numbers and photo intervals based on the model’s integrity,number of point clouds,and reconstruction time as evaluation indicators.The results show that the reconstructed model with an interval of10°has high integrity,and 108 images can reconstruct a 3D model that is close to the real object and has low noise in less time.Therefore,this article chooses to capture a multi view image sequence with a 10°photography interval and 108 images to complete the subsequent reconstruction task.(2)Processing point clouds through PCL and Cloudcompare software provides a foundation for phenotype measurement.Non destructive measurement of parameters such as plant height,stem diameter,leaf area,and leaf projection area was achieved through algorithms such as maximum traversal,triangulation,and cylindrical fitting.The results were compared with corresponding manual measurements,and the corresponding determination coefficients(R~2)were 0.99,0.548,0.938,0.728,root mean square error(RMSE)were 1.947cm,0.713mm,6.584cm~2,47.773cm~2,and mean absolute percentage error(MAPE)were 6.598%,16.369%,11.169%,85.485%.The research results indicate that the plant height and leaf area parameters calculated based on the three-dimensional model maintain good consistency with manual measurement values,which can meet the needs of phenotype research.(3)We reconstructed a three-dimensional model of kidney bean seedlings during their growth process and analyzed the changes in plant height and minimum bounding box volume per plant of two types of kidney beans during the seedling stage.Through experiments,it was found that the absolute growth rates of plant height and minimum bounding box volume per plant of cloud kidney bean 1 were 2.63cm and 847.49cm~3,and the relative growth rates were 12.38%and 86.43%;The absolute growth rates of plant height and minimum bounding box volume per plant of cloud flower kidney bean 2 were3.01cm and 724.42cm~3,and the relative growth rates were 14.35%and 79.55%.The results showed that the plant height growth rate of cloud flower kidney bean 2 was faster than that of cloud kidney bean 1,while the minimum bounding box volume growth rate of cloud kidney bean 1 was faster than that of cloud flower kidney bean 2. |