Three-dimensional Visual Modeling Of Seedling Rapeseed Based On MVS Technology

Precision application technology has a significant advantage in improving pesticide utilization,reducing environmental pollution and enhancing food safety,and has become the main trend of current agricultural development.The core of precision application technology is to obtain information on the growth status of field plants,such as morphology,density,pests and diseases,and to develop different operational measures based on the differences in growth between plants.The key step in obtaining plant growth status information and its growth variability is the three-dimensional visualization of plant morphological structure.Research on 3D visualization of plants,combining organic technologies such as sensors,artificial intelligence,computers,and visual images.The three-dimensional visualization of plants can accurately describe the spatial structure distribution characteristics of plants,To a certain extent,it can reflect the growth state of plants,It greatly promotes the research and application of plant three-dimensional structure model,and provides a reliable technical means for the visualization of plant morphological characteristics and physiological structure functions.In this paper,the simulation of green plants(green radish,evergreen)and seedling rapeseed(three-leaf stage,four-leaf stage,seven-leaf stage,insect pests)as the research object.Aiming at the three-dimensional morphological model construction and growth parameter extraction of plants,a three-dimensional morphological model construction method based on MVS sequence image technology was proposed.Calculate and extract feature points using the SIFT algorithm.The SFM and PMVS are calculated and adjusted for the feature points,and the original three-dimensional point cloud data of the whole plant is obtained,and the point cloud data is filtered,streamlined,reconstructed,and morphological parameters are calculated.This study provides basic technology for the realization of precision pesticide application technology in rapeseed field.The specific research work is as follows:(1)Build a multi-view stereo vision system platform.Based on the analysis and research of the working principle of multi-view data acquisition and the related camera calibration work,combined with the adjustable rotation platform,a multi-view stereo vision system platform with a certain speed and angle rotation is designed.(2)Acquisition of the original plant point cloud.Based on the comprehensive consideration of the MVS sequence image method and the structural characteristics of the plant,the existing image matching and spatial point cloud data acquisition methods are discussed,including the theory of SIFT algorithm,the basic concept and application of SFM algorithm,research on PMVS algorithm model.A three-dimensional morphological model construction method and a growth parameter extraction scheme for whole rapeseed were developed.(3)Study on 3D reconstruction technology of rape plant.According to the obtained original plant point cloud data,a filtering method combining manual and algorithm is adopted.According to the morphological characteristics of rape leaves,an adaptive point reduction algorithm based on local curvature was proposed to deal with the removal of noise points and streamlined point cloud data from rapeseed plants.For the streamlined point cloud model,the triangle,the leaves and stems of the rapeseed plants were obtained,and the complete rapeseed plant surface model was obtained.(4)Growth parameter calculation and measurement of plant 3D surface model.Based on the plant 3D surface model,the plant growth parameters(plant height,leaf area,leaf length,leaf width,leaf dip angle)were calculated using Euclidean distance,Hausdorff distance,and triangular mesh accumulation.According to the statistical analysis,the parameters calculated by the model and the manually measured data are accurately evaluated.To verify the accuracy of the surface model.