基于水稻三维模型的表型参数提取及生物量估测

Rice Phenotypic Parameters Extraction and Biomass Estimation Based on Three-Dimensional Model

基于二维图像的水稻表型检测方法缺乏空间维度信息,难以提取全面的水稻表型参数,因此建立水稻三维模型进而提取全面的水稻表型参数具有重要意义。体积是指示水稻生长状况的重要参数,如何在不破坏水稻植株的前提下获取水稻体积参数是当下研究中仍有待解决的问题。基于上述考虑,提出一种基于水稻三维模型的表型参数提取及生物量估测方法,实现在不破坏水稻植株的前提下提取水稻体积参数并估测水稻生物量。以盆栽水稻为对象,首先利用Alpha-shape算法重建其三维模型,然后提取并评估水稻株高、茎粗、植被覆盖度、体积参数,并基于体积参数估测水稻生物量。结果表明,空间雕刻方法可以重建高精度的水稻三维模型,并准确测算水稻表型参数及生物量,其中水稻株高、茎粗、单株植被覆盖度均方根误差(RMSE)和平均绝对百分比误差(MAPE)分别为63.27 mm、4.01 mm、5.04%和7.15%、14.91%、12.59%,水稻生物量RMSE和MAPE分别为62.44 g、19.25%,均优于现有三维重建软件的测算结果。

The phenotype detection methods based on two-dimensional images lack spatial dimension information and have difficulty in extracting comprehensive rice phenotype parameters.Therefore,it is of great significance to establish a three-dimensional model of rice to extract comprehensive rice phenotype parameters.The volume is an important parameter indicating the growth status of rice.How to obtain rice volume parameters without damaging the rice plant is still a problem to be solved in current research.Based on the above considerations,a method of rice phenotypic parameters extraction and biomass estimation based on three-dimensional model is proposed in this paper,which can extract rice volume parameters and estimate rice biomass without damaging the rice plant.This study focused on potted rice,first reconstructed its three-dimensional model using the Alpha-shape algorithm,and then extracted and evaluated rice height,stem thickness,vegetation coverage,volume parameters,and estimated rice biomass based on the volume parameter.The experimental results showed that space carving could reconstruct a high-precision three-dimensional model of rice and accurately measure rice phenotype parameters and biomass,the RMSE and MAPE of rice height,stem thickness,and single plant vegetation coverage were 63.27 mm,4.01 mm,5.04% and 7.15%,14.91%,12.59%,and the RMSE and MAPE of rice biomass were 62.44 g and 19.25%,which were better than the results obtained by existing three-dimensional reconstruction software.