Abstract

Hyperspectral remote sensing is frequently used to monitor chlorophyll content, an important characteristic for assessing photosynthetic ability, health and defence against a variety of degenerative diseases. To obtain hyperspectral data, field portable spectroradiometers, such as Ocean Optics Hyperspectral Vis-NIR spectroradiometers and Analytical Spectral Devices FieldSpec series, have been widely used. The development of an affordable hyperspectral remote sensing system would be advantageous. Highly sensitive, affordable and cost-effective finger-tip size spectrometers have recently been released. In this study we investigate the potential of hyperspectral data obtained from such a compact spectrometer (C12880MA-10, Hamamatsu Photonics) for estimating chlorophyll content in Zizania latifolia. We also tested the efficacy of five pre-processing techniques (first derivative reflectance, continuum-removal transformation, de-trending, multiplicative scatter correction and standard normal variate) in conjunction with five machine learning algorithms.

摘要

高光谱遥感经常用于监测叶绿素含量，这是评估光合能力、健康和防御各种退行性疾病的重要特征。为了获得高光谱数据，现场便携式光谱辐射计，如海洋光学高光谱可见-近红外光谱辐射计和分析光谱设备 FieldSpec 系列，已被广泛使用。开发负担得起的高光谱遥感系统将是有利的。最近发布了高度灵敏、经济实惠且具有成本效益的指尖尺寸光谱仪。在这项研究中，我们研究了从这种紧凑型光谱仪（ C12880MA -10，Hamamatsu Photonics）获得的高光谱数据在估计阔叶茯苓中叶绿素含量的潜力. 我们还结合五种机器学习算法测试了五种预处理技术（一阶导数反射率、连续介质去除变换、去趋势、乘法散射校正和标准正态变量）的功效。