Remote sensing estimation of chlorophyll content in rape leaves in Weibei dryland region of China

Abstract

To explore the Hyperspectral Estimation Method for estimating the chlorophyll content of rape leaves, so as to provide a scientific basis for rapid and nondestructive monitoring of the chlorophyll content of rape crops in Northwest China.Taking the rapeseed crops in the northwest region as the research object, through the correlation analysis of the SPAD value and the spectral parameters of the rape leaves, the spectral parameters sensitive to SPAD were screened, and the single factor model,the partial least square regression model (PLSR) and BP neural network model optimized by genetic algorithm based on multiple linear stepwise regression based on the spectral parameters were constructed respectively and were compared.The results showed that: 1) The general trend of the spectral curve of rape leaves was the same, and the spectral reflectance decreased with the increase of chlorophyll content; 2) The correlation of seven spectral parameters involved in the modeling was above 0.770, all of which reached significant correlation at 0.01 level; 3) In each growth period, the BP neural network model optimized by genetic algorithm based on multiple linear stepwise regression is the optimal model. The modeling R 2 is above 0.77, and the maximum can reach 0.91. It is verified that R 2 is above 0.73, the maximum can reach 0.92, RMSE is between 1.32–3.22, RE is between 2.50% − 4.49%. BP neural network model optimized by genetic algorithm based on multiple linear stepwise regression is an inversion method which can estimate the SPAD value of rape leaves accurately and quickly.

Plant phenotyping relevance

ラップ葉のクロロフィル含量という植物形質を、ハイパースペクトル計測と回帰・ニューラルネットワークで推定する手法の構築および検証が研究の中心である。

Code and data availability

The paper reports SPAD and hyperspectral measurements of rape leaves and MLSR-GA-BP/PLSR models, but explicitly states the datasets are not publicly available and only available from the corresponding author on reasonable request. No public data, code, model, or supplement URLs are provided.

No evidence-backed public reproduction asset is currently recorded.