Flowers in nature do not have a similar tree like structure, so for the 3D reconstruction of flowering plants, this article will focus on the 3D reconstruction of the natural structure of the plants themselves. Through observation, it is known that the petals of the vast majority of flowering plants in nature have similar or even identical shapes, known as self similarity. Therefore, for the modeling of flowering plants, this article will follow the general idea of tree modeling and use fractal geometry to generate complex flowers with self similar shapes through a simple iterative process. The lightweight reconstruction algorithm proposed in this article for flowering plants fully utilizes the self similarity characteristics of plants. While ensuring the quality of plant models, it minimizes the data volume of individual models as much as possible, achieving lightweight modeling and facilitating the subsequent construction of large-scale botanical garden scenes to accommodate as many plant models as possible.