Research On 3D Reconstruction Technology Based On Single Leaf Image

The 3D modeling method of plants has been the focus of research in fields such as computer graphics and visualization modeling.As an important part of the individual plant,the modeling method of plant leaves has always been one of the important aspects of plant digitization technology as well.The leaf models can not only provide more information and data for botanical research but can also be widely used in virtual reality,film and video games,animation,and other fields.Most existing methods of modelling leaf models use scanning equipment,process-based methods or multi-view geometry,which are less efficient and more expensive to model and have limitations that limit the development of leaf modelling techniques.With the development of single-image 3D reconstruction technology,it has become possible to recover the surface model of an object from a single-view image.Due to the convenience of single-image acquisition,if a 3D model of a leaf can be recovered using a single image,it will not only greatly improve the efficiency of modelling but also promote the development of plant digitization technology.This paper will therefore investigate leaf modelling techniques,propose a new single-image-based leaf model reconstruction technique to easily and quickly generate leaf model results in 3D form,and also develop a corresponding plant leaf modelling system.The main elements include:(1)This paper presents a technique for digitally modelling plant leaves based on a single image,where the corresponding 3D model results are recovered by inputting a single view of the leaf image.Firstly,the FCM algorithm is used to generate a mask map from the original image,extract the corresponding plant leaf body,and enhance the image with grayscale?then,the gradient field information in the image is calculated by the gradient field operator and converted into normal vector map results? the 3D reconstruction task of the leaf is divided into two parts: surface model reconstruction and macrostructure optimization.In the surface model reconstruction process,the surface model of the plant leaf is recovered from the gradient information based on the extracted gradient field information and combined with discrete geometry processing operations? in the macro-structure optimization process,the leaf model is optimized using the leaf vein median curve function based on the structural characteristics of the plant leaf to obtain the final 3D model of the plant leaf.The results show that this method can generate leaf model results with rich details and complete macro-structure.(2)Multiple sets of experiments are designed to verify the effectiveness of the algorithms in this paper.Firstly,experiments are set up to model plant leaves in images under ideal lighting conditions and compare them with the classical SFS-based method to verify the effectiveness of this paper’s method.Experiments were also set up to model directly acquired real leaf image data to verify the applicability of the method in this paper.The results show that the method is able to obtain structurally complete and detailed texture-rich leaf model results for different forms of individual plant leaves with only a single image input.Finally,experiments are set up to model common objects such as coins and tea bags and show the corresponding results to illustrate the scalability of the method in this paper.(3)Development of software for modelling plant leaves.In order to extend the application of plant leaf 3D modelling techniques described in this paper and to enhance the practicality of the method proposed in this paper,a plant leaf modelling software has been developed based on the method proposed in this paper.The software is based on the Windows platform and is able to recover the corresponding leaf model results using the input single image and make a 3D visual display while integrating a number of functions such as mask map calculation and gradient field calculation to meet the different needs of users.From the results,the blade modelling software uses a single blade image combined with simple user interaction to produce high-quality blade model results with low cost,high applicability,and high efficiency.