In the present research, a model of three-dimensional structure of rice was developed. Three-dimensional position of rice leaves were measured along the edges of the leaves, and the position of stems were measured along the center line. The measured data were contaminated by systematic error due to the device and random error due to the wind and device. The proposed methodology reduces such errors through the following steps. First, the positional data of the center line of leaf was estimated from a pair of data measured at both leaf edges. Next, in
xy-plane, principal component analysis (PCA) produced the first principal component, named as
u-axis. Then, in
uz-plane, a quadratic curve of the leaf was estimated by minimizing the sum of squared minimum distances between the points and the curve. The leaf width was modeled as a function of the leaf relative length to the full length. Finally, three-dimensional structure of rice was modeled. The estimated length and area of rice leaves were compared with the measured data. It was revealed that root mean square of errors (RMSE) of the estimated center length of all leaves was 7.4%, the relative error of the estimated areas of all leaves per rice individual was -6.8%, which is acceptable for the analysis of microwave scattering.
抄録全体を表示