Most plants in nature have the characteristic of self similarity, which refers to the similarity between any local shape and the overall shape, and these self similar shapes are mathematically known as fractals. The self similarity of plants provides a powerful tool for lightweight modeling of plants. Through techniques such as recursive generation, fractal geometry, texture mapping, detail hierarchy, spline curves, and real-time generation, the storage and computational requirements of the model can be significantly reduced, and rendering efficiency can be improved. By applying these methods, efficient rendering of large-scale plant scenes can be achieved while ensuring visual effects. The lightweight reconstruction algorithm proposed in this article for tree like plants fully utilizes the self similarity characteristics of plants. While ensuring the quality of plant models, it minimizes the data volume of individual models, achieves lightweight modeling, and facilitates the subsequent construction of large-scale botanical garden scenes to accommodate as many plant models as possible.