Plant leaves visualization based on leaf vein extraction

Siyuan Zhu, Meili Wang

Abstract: Three-dimensional (3D) reconstruction of plant leaves plays a key role in virtual-plants modelling. To overcome the inefficiency in 3D modelling of plant leaves, we propose a method based on leaf vein extraction, which takes the images of leaves as inputs. Firstly, we convert the color space from RGB to HSI. By decomposing HSI we extract the component-I next. Then, we use the pixel values of component-I as the inputs of K-means clustering to obtain the accurate margin and leaf vein. Next, we triangulate images to meshes and build the texture for them. Finally, we achieve leaves deformation using spring-mass system. The experimental results show that our proposed method can reconstruct leaves effectively and accurately with high adaptability.