| Plant three-dimensional morphological structure is a true reflection of plant growth and development state.It is able to establish a rapid acquisition of data,and then accurately build a three-dimensional plant model.The three-dimensional plant model has always been a research hotspot in botany,computer graphics and other fields.However,plants are complex living organisms,and the modeling of their morphological structure is much more difficult than that of industrial machinery parts,buildings and other structural modeling,mainly manifested by their non-rigid,dynamic growth,different sizes,structural similarities and differences,and so on.In addition,the morphological structure of crops keeps changing throughout the whole growth period.It requires a lot of complicated and tedious modeling work to describe the evolution of data collection and modeling of the entire morphological structure process of plants from realizing the germination,emergence,growth and development,morphological formation,maturity and aging.How to obtain plant morphological and structural data comprehensively,accurately,nondestructively and continuously in a relatively low-cost way,analyze and study the morphological structure and growth process of crops in a three-dimensional visual way,and conduct phenotypic measurement are the key links in the current research on digital plants and phenotypic omics.The 3d reconstruction method based on multi-view image can generate dense point cloud and texture data of plant morphology and structure by shooting multi-view 2d image sequence.Compared with other 3d reconstruction methods,this method has the advantages of low equipment cost,convenient data acquisition and flexible use,etc.The obtained 3d point cloud information not only has a high density,but also contains the image true color texture and other information,with a strong sense of reality.This method got the attention of the researchers in recent years,become a plant one of the main methods of 3d modeling research,however,based on the perspectives of image technology for 3d reconstruction of plant,influenced by the environment,the camera point design,sampling frequency and the influence of such factors as the restriction of point cloud synthesis post-processing,so the reasonable design of data acquisition devices and the late algorithm processing is the emphasis and difficulty in the technology of the current study.This study takes corn as an example and tries to find a fast and accurate method to obtain the 3d model of corn plant.In this paper,the different growth period of maize plants into the 3d reconstruction research,aiming at the condition of the plant is too high at the same time,point cloud registration and splicing technology,implementation of tall high maize plants(taller than 3m)3d reconstruction,through this article research,to the development of plant 3d reconstruction technology,plant phenotypic omics research and the development of low cost and high flux equipment provides the beneficial reference.This paper mainly includes the following aspects:(1)Based Based on the technical principle of motion recovery structure,the corresponding image acquisition and 3d reconstruction of multi-perspective images were carried out for maize plants with different height in different growth periods.The multi-angle image acquisition device is designed by adopting two different forms,namely"phase maneuver,plant immobility",namely rotating orbit type and"plant immobility,camera immobility",namely electric rotary table,denoted as"mode 1"and"mode 2"respectively.Study revolving type structure device and electric turntable structure device can realize the three-dimensional reconstruction of corn plant,higher quality,texture and the reconstruction of point cloud information more real,rebuild the success rate and higher robustness,by extracting 3d scanner FARO,multiple points of view and high resolution reconstruction of image sequence of the two groups of small maize plants from the leaf sheath and blade width phenotypic parameters,get the root mean square error is respectively:RMSE=0.098 cm and 0.028 cm,the correlation coefficient is respectively:R~2=0.83 and 0.84,which can achieve better reconstruction quality than the 3d scanner.While the latter is not very effective in reconstructing the quality of point cloud,but it is more portable in carrying devices and has great research potential.(2)Aiming at the problem that the reconstructed point cloud in"mode 2"is noisy and affects the reconstruction quality,and aiming at the characteristic that corn plants are green,a background segmentation algorithm based on ultra-green algorithm is proposed to reduce the background interference irrelevant to the corn plants to be reconstructed in the image.In the image acquisition process structures,on the background of the blue curtain shooting environment,super green(EXG)operator is then used to obtain the image segmentation in the green plants,the image sequence into multiple points of view contains only the green plant rebuilding image sequence,solved the complex image background environment and shortcomings affecting the quality of reconstruction,to avoid the background environment disturbance to the image feature point extraction,greatly reduces the number of reconstruction noise,improve the reconstruction results.(3)In order to solve the problem that high plant height affects the integrity and precision of reconstruction,this paper proposes a multi-perspective point cloud splicing method for maize plants based on ICP algorithm.The method based on truncation processing plants,segmented perspectives of reconstruction,and then using ICP algorithm,to obtain the truncation point cloud plant for Mosaic,the splicing overlap part of the point cloud,small errors error range between 0-0.12 cm,the distance is small,the larger problem stitching accuracy is higher,can satisfy 3d reconstruction of tall plant,solved the plants,plant is much too high perspective image to acquire the inconvenience shade and get the poor image quality. |
PDF Full Text Request