3D Phenotypic Information Extraction Method of Maize Seedlings at Leaf Scale

In biological breeding and genomic research, the three-dimensional phenotypic structure information of plants is especially crucial. To extract the three-dimensional phenotypic information of plants efficiently, quickly, and nondestructively, taking corn as an example, a method for extracting the three-dimensional phenotypic structure information of maize seedling at leaf-scale from a three-dimensional point cloud produced from an image is proposed in this study. First, using a motion recovery structure algorithm, the image obtained from a mobile phone is rebuilt to produce a three-dimensional point cloud and then integrated with the ExGR index and conditional Euclidean clustering algorithm to automatically extract the corn seedlings from the surrounding environment. We employ the regional growth algorithm to segment the leaves. Finally, the three-dimensional phenotypic structure information of corn seedlings, including height, three-dimensional volume, leaf area, and leaf perimeter, are computed, and the dynamic changes of phenotypic information over time are examined. The findings demonstrate that the method in this study compares with the real value; the root mean square error (RMSE) of plant height, leaf area, and leaf circumference is 0. 77 cm, 1. 62 cm(2), and 1. 21 cm, respectively; the mean absolute percentage error (MAPE) is 3. 23%, 8. 27%, and 4. 75% respectively; and the determination coefficient R-2 reaches above 0. 98. The proposed method can efficiently and nondestructively extract the three-dimensional phenotypic structure information of corn seedlings and can be extended to the extraction of other columnar structure plant phenotypic information.