Abstract: In recent years, great progress has been achieved in the domain of image-based high-throughput rice phenotype. However, only a few researches considered 3D plant models. Generally, more comprehensive information can be extracted from 3D models than images. The reconstruction of 3D models is the foundation of 3D crop phenotyping. This paper proposed a method to reconstruct 3D point clouds for pot rice plants. In this method, a series of rice images were taken by a camera in fixed place, then the point cloud representations of reconstructed visual hulls of rice plants were acquired by shape from silhouette according to camera parameters and binary rice silhouette images. Finally, the reconstructed models were colored by back projection method. The experiment results showed that this method was well suitable for rice plants of different types and stages.

Key words: rice phenotype, 3D reconstruction, visual hull

摘要： 近几年，基于图像的高通量水稻表型研究取得了极大进展，但从三维层面进行研究的工作则相对较少。一般而言，相对于二维图像，从三维模型中能提取更为全面的性状参数。三维模型重建是作物三维表型研究的基础，提出一种适用于盆栽水稻三维点云的重建方法。该方法在相机固定、样本旋转的拍摄模式下获取水稻多视角图像，根据相机标定参数以及水稻轮廓二值图，通过轮廓投影方法重建水稻三维可视外壳点云模型，并通过反投影方法进行点云着色。结果表明，该方法对于不同时期及不同品种的水稻样本均能取得较好重建效果。

关键词: 水稻表型, 三维重建, 可视外壳