Plants Dimensional Image-based Morphometry And Reconstruction System

Image-based measurement and modeling is a hot research topic in recent years,with a goal for reconstructing accurate and realistic model of a specific object.Plant plays an important role in real world,the 3-D reconstruction of plant has significance in plant growth detection,plant shape analysis,game development and natural scenes simulation,etc.This thesis,taking corn and cucumber as examples,discusses the problem of image-based plant modeling and shape measurement,and proposes methods for reconstructing the leaf and skeleton of plant,and also develops the corresponding software system.Concretely,the primary work and research results can be listed as follows:a)Establishing a stereo vision system.We design a simple and effective image acquisition system base on the study of the existing image acquisition system.This system is not only easy to operate,but also reduces the false matching probability.Results showed that the captured images by this system can be well used for the reconstruction of plants.b)Proposing a feature matching algorithm based on SIFT descriptor.Stereo matching is the key and difficult point in the field of image-base modeling.The accuracy of matching has a direct impact on the measurement and reconstruction.SIFT is a better method of matching,on the basis of SITF,this article gives a description of the characteristics point,and puts forward a more robust match approach.c)Realizing multiple images-based 3D reconstruction of crop leaf.Firstly,we capture leaves images of watermelon and cucumber from different angles,and use corner detection to extract features of the leaf;Second,after matching characteristics of the feature points we obtain the three-dimensional coordinates of these feature points,and then use the spline interpolation curve, re-sampling and grid,we reconstruct a grid model;At last,we product a more real plant leaves model through texture mapping.d)Realizing skeleton reconstruction based on two images.We take the corn as study object. First of all,we extract skeleton of each single plant image;then match the skeleton feature points and gained the three-dimensional data,generate three-dimensional cloud point model of the skeleton.Based on the model,we can carry out plant height,leaf length,angle between stem and leaf.Lastly,a curve model for the plant skeleton is reconstructed by using curve interpolation.