Abstract: We present an approach to modeling two different types of natural plants with sensor data．The approach first constructs the geometric models of the plants．The models are composed of billboard clusters,each with billboards corresponding to all input views．The plants are then rendered by using view-dependent billboards．For plants with relatively large leaves,the geometric clusters are constructed by segmenting volumes recovered from images．The clusters for thin-leaf plants are built by subdividing the bounding boxes of 3D laser data．Results in our experiments demonstrate that the proposed method can capture the reality of natural plants．