摘要

冠层散射系数（CSC）是双向反射系数（BRF）与定向面积散射系数（DASF）的比值，已成功应用于校正冠层结构的影响。计算CSC的关键是计算DASF，DASF由BRF和BRF之间的线性关系的截距b和斜率kλ（Ω）ωλ（特定视角方向上的高光谱BRF与叶片反照率（ωλ）的比率）决定。然而，由于多光谱波段的限制，如何利用多光谱无人机数据准确计算整个生长期的作物DASF仍然是一个亟待解决的问题。在本研究中，连续三年获得了小麦冠层多角度（0°，−30°，−45°）数据集，包括近地高光谱数据、无人机多光谱数据和PROSAIL模拟数据。提出了基于多光谱传感器的DASF k-b模型，通过相对累积生长度天数（RAGDD）和BRF来估计b和k。将先前开发的DASF g-NIR模型和植被指数（VI）模型与DASF k-b模型在不同视角（VZAs）下进行比较，以评估它们在校正冠层结构效应、基于广义加性模型（GAM）估计叶片氮含量（LNC）和叶绿素含量（LCC）方面的性能。结果表明，适用于估算小麦DASF的高光谱波段范围为710-760nm。参数b随着施氮量的增加而降低，并随着RAGDD的增加而先激活后抑制；k、DASF Hy的趋势相反。与CSC g-NIR（根据DASF g-NIR模型计算）和VIs相比，CSC k-b（根据DACF k-b模型计算）在不同VZA下对冠层结构的校正效果最好。与CSC g-NIR和VIs相比，使用CSC k-b的LNC和LCC的估计精度有所提高，RRMSE值分别为9.2%和7.0%，两种模型的推荐VZA均为−45°。

Estimating leaf nitrogen and chlorophyll content in wheat by correcting canopy structure effect through multi-angular remote sensing 

Abstract

Canopy scattering coefficient (CSC) is the ratio of bidirectional reflectance factor (BRF) to directional area scattering coefficient (DASF), and has been successfully applied to correct the effect of canopy structure. The key to calculate CSC is to calculate DASF, which is determined by the intercept b and slope k of the linear relationship between BRF and BRFλ(Ω) ωλ (the ratio of hyperspectral BRF in certain view direction to leaf albedo (ω λ). However, due to the limitation of multispectral bands, how to accurately calculate crop DASF during the whole growth period using multispectral UAV data is still an urgent problem to be solved. In this study, wheat canopy multi-angular (0 • , − 30 • , − 45 •) datasets including near-ground hyperspectral data, UAV multispectral data, and PROSAIL simulation data were obtained for three consecutive years. The DASF k-b model was proposed to estimate b and k based on multispectral sensors by relative accumulated growing degree days (RAGDD) and BRF. The previously developed DASF g-NIR model and vegetation index (VI) model were compared with DASF k-b model under different view angles (VZAs), to evaluate their performances in correcting canopy structural effect, estimating leaf nitrogen content (LNC) and leaf chlorophyll content (LCC) based on generalized additive model (GAM). The results showed that the hyperspectral band range suitable for estimating wheat DASF was 710-760 nm. Parameter b decreased with increased N application rates, and activated first and then inhibited with increased RAGDD; the tendency of k, DASF Hy were opposite. Compared with CSC g-NIR (calculated from DASF g-NIR model) and VIs, CSC k-b (calculated from DASF k-b model) had the best correction effect on canopy structure under different VZAs. The estimation accuracy of LNC and LCC using CSC k-b was improved compared with CSC g-NIR and VIs, with RRMSE values of 9.2% and 7.0%, respectively, and the recommended VZA was − 45 • for both models. 

Imagery collection: UAV imagery was collected by M600 Pro (DJI technology company, Shenzhen, China), equipped with AIRPHEN multispectral camera (https://www.hiphen-plant.com/our-solutions/airphen/). The camera contains six spectral wavebands with 10 nm full width of half peak. The focal length and field-angle (FOV) of 450 nm, 530 nm, 675 nm, 730 nm and 850 nm wavelengths is 8 mm and 33◦ ×25◦, respectively. While 570 nm wavelength is 4.2 nm and 60◦ × 46◦, respectively. The UAV data collection was conducted at 10:00–14:00 a. m. on clear and cloudless days during the key growth period of wheat. The flight height, speed and VZAs were 10 m, 2.5 m/s, and ⊖ 45◦, ⊖ 30◦ and 0◦, respectively. The sideways and forward overlaps were 94% and 85%, respectively. UAV images were geometric corrected and mosaicked by professional software Agsoft PhotoScan (Version 1.2.4.2399, Agsoft LLC., Russia) (Li et al., 2021a), then BRF was extracted through ENVI 5.3 (Exelis Visual Information Solutions, Boulder, Colorado, USA). The ground sampling data were synchronized with the flight day.

相关阅读

便携式田间表型成像系统LITERAL

HI-phen田间高通量植物表型分析系统