Introducing partial pixel integration of UAV imagery to estimate maize (Zea mays L.) aboveground biomass

Abstract

Accurate estimation of aboveground biomass (AGB) helps to monitor maize growth and yield prediction, and unmanned aerial vehicles (UAVs) have become one of the most significant technological tools in precision agriculture. However, previous studies have mainly focused on utilizing spectral indices, texture metrics and structural features derived from UAV multispectral imagery. These methods often involve significant uncertainties and ignore overall maize morphological characteristics. In this study, an innovative partial pixel integration (PPI) parameter is introduced to characterize both horizontal and vertical structural features of maize (Zea mays L.) at the plot scale. Field experiments were conducted in Dafeng District, Yancheng City, Jiangsu Province, China. Multispectral UAV imagery was captured at five flight altitudes (10 m, 20 m, 30 m, 50 m, and 80 m). Five structural features—fractional vegetation cover (FVC), plant height (PH), FVC × PH, pixel integration (PI), and PPI—were extracted to develop Fresh and Dry AGB estimation models based on linear, exponential, and power functions. The models were verified with the method of five-fold cross-validation to evaluate the predictive performance of different parameters. The results revealed that: (1) The models with PPI parameter outperformed that with all other metrics (FVC, PH, FVC×PH, PI), achieving the highest R² values of 0.968 for Fresh AGB (at 20 m flight altitude) and 0.948 for Dry AGB (at 10 m flight altitude); (2) Fresh AGB estimation models were generally more accurate than Dry AGB estimation models; (3) Contrary to expectations, increasing UAV flight altitude did not necessarily reduce AGB prediction accuracy. These findings demonstrate that the PPI parameter delivers high accuracy and robustness, presenting a novel and reliable approach for in-field maize AGB estimation.

Plant phenotyping relevance

UAV画像からトウモロコシの形態特徴を抽出し、PPIという新規パラメータで地上部バイオマスを推定・検証することが研究の中心であり、植物表現型取得手法として適格。

Code and data availability

公開状態または取得可能な本文経路を確認できませんでした。

No evidence-backed public reproduction asset is currently recorded.