A lettuce moisture detection method based on terahertz time-domain spectroscopy

Xiaodong Zhang,Hongyan Gao,Zhaohui Duan,Zhiyu Zuo,Hanping Mao

doi:10.1590/0103-8478cr20210002

Xiaodong Zhang, Hongyan Gao + Show 3 more

Open Access

https://doi.org/10.1590/0103-8478cr20210002

Copy DOI

Journal: Ciência Rural	Publication Date: Jan 1, 2022
Citations: 1	License type: CC BY 4.0

Affiliation: Jiangsu University

Abstract

ABSTRACT: For non-destructive detection of water stress in lettuce, terahertz time-domain spectroscopy (THz-TDS) was used to quantitatively analyze water content in lettuce. Four gradient lettuce water contents were used . Spectral data of lettuce were collected by a THz-TDS system, and denoised using the S-G derivative, Savitzky-Golay (S-G) smoothing and normalization filtering. The fitting effect of the pretreatment method was better than that of regression fitting, and the S-G derivative fitting effect was obtained. Then a calibration set and a verification set were divided by the Kennan-Stone algorithm, sample set partitioning based on joint X-Y distance (SPXY) algorithm, and the random sampling (RS) algorithm, and the parameters of RS were optimized by regression fitting. The stability competitive adaptive reweighted sampling, iteratively retained information variables and interval combination optimization were used to select characteristic wavelengths, and then continuous projection was used on basis of the three algorithms above. After the successive projection algorithm was re-screened, partial least squares regression was used into modeling. The regression coefficients Rc 2 and RMSEC reach 0.8962 and 412.5% respectively, and Rp 2 and RMSEP of the verification set are 0.8757 and 528.9% respectively.

Highlights

China is the largest producer of lettuce, accounting for approximately 50% of the world’s output
A calibration set and a verification set were divided by the Kennan-Stone algorithm, sample set partitioning based on joint X-Y distance (SPXY) algorithm, and the random sampling (RS) algorithm, and the parameters of RS were optimized by regression fitting
The window widths were either 5, 7, or 9 points, and the optimal window width was selected by comparing the determination coefficient Rc2 obtained from partial least squares (PLS) linear fitting with root mean square errors of calibration (RMSEC)

Summary

Introduction

China is the largest producer of lettuce, accounting for approximately 50% of the world’s output. Supplying appropriate amounts of water at different growth stages, as well as rapidly and accurately detecting water stress, is essential for maximizing the yield, quality, and taste of lettuce (WANG et al, 2016). Crop canopy temperature is an indicator of crop water stress. Infrared thermal imaging is a well-studied technique that has been widely used to monitor crop water stress because it permits the water status of crops to be determined rapidly and in a non-destructive manner (O’SHAUGHNESSY et al, 2011). A previous study using thermal imaging to explore the relationship between water stress in papaya and three physiological indexes (stomatal conductance, transpiration rate, and net photosynthesis) under different irrigation conditions demonstrated that thermal imaging is a promising technology for monitoring the physiological state of papaya under drought conditions (LIMA et al, 1999)

Methods

Results

Conclusion