Research On 3D Reconstruction System Of Potato Based On Triangular Image

With the gradual combination of computer technology and agricultural knowledge,the research of crop morphological structure and physiological function has entered the stage of digitalization and visualization.By using advanced sensor technology and modern information technology on computer,the shape and structure of crops can be visualized in three dimensions,and the dynamic process of plant growth and plant environment interaction can be analyzed,simulated and predicted.It is a hot and difficult point in the research of digital plant and crop breeding to acquire data information quickly,reconstruct the three-dimensional structure of plants,analyze and study the morphological structure and growth process of crops in the way of three-dimensional visualization,and measure their phenotype.This is of great significance to the extraction of plant phenotype characteristics and the study of 3D visualization of plants.The 3D reconstruction method based on triangulation and multi image can generate dense point cloud and texture data of plant morphology and structure by photographing 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.The acquired 3D information of point cloud not only has high density,but also contains the true color texture information of the image,which has a strong sense of reality.In recent years,this method has attracted the attention of many researchers and has become one of the main methods of plant 3D modeling.The 3D reconstruction of plants based on multi perspective image technology is affected by environment,camera point design,acquisition frequency and other factors,as well as the processing after point cloud synthesis,which is the focus and difficulty of the research.The reasonable design of the data acquisition device and the post-processing algorithm are the key points of this technique.(1)The basic flow and optimization methods of 3d reconstruction algorithm are established:based on SIFT feature extraction and matching,calculation and integration of each grouping projection structure,camera self-calibration and 3d point cloud reconstruction.finally,the BA method is used to optimize the preliminary estimation results.(2)Combined with the general process of SIFT algorithm,including four steps:local extremum detection,feature point selection,feature point direction determination and feature descriptor generation,the algorithm steps are given,and the method of feature extraction and matching in image processing is clarified.(3)The needed texture model is obtained by using the three-dimensional reconstruction software based on motion recovery structure(SfM)and multi-angle stereo vision(MVS)Photo Scan and the point cloud processing is carried out for the deficiency that can not be measured in point cloud environment.(4)Through the point cloud preprocessing method,the average point distance of threshold setting in the process of neighborhood search,plane segmentation and contour extraction is calculated,the k-d tree topology structure of neighborhood search is established,and the normal vector is estimated to facilitate plane segmentation.Through RANSAC algorithm,the plane of the segmented standard point cloud is determined,and the parameters of the projected point cloud are obtained by projecting the segmented point cloud onto the plane.In order to solve the problem of error fitting when the inner and outer contour directly fit the straight line,the connectivity judgment of contour is introduced.The plane contour is divided into a set of connectable contours,and the connectable contours in the set are successively projected with straight lines.Finally,the straight line contour features of potato model are obtained.(5)Using Ba method to calculate the re projection error of the above results,determine the final sparse 3D point cloud structure,design and test the data according to the research objectives and the content of the calculation process.The experimental results show that the algorithm in this paper has good accuracy.(6)The system consists of five modules:data input,data output,visual interaction,3D reconstruction and point cloud processing.Crop 3D reconstruction based on triangulation has the advantages of low equipment cost,convenient data acquisition and flexible use.At the same time,the collected point cloud data is not only high-density,but also high-density,with a strong sense of reality,which can better realize the virtualization of real things.