3/28/24, 1:28 PM Detection of maize seed germination rate based on improved locally linear embedding - ScienceDirect

Computers and Electronics in Agriculture

Volume 204, January 2023, 107514

Detection of maize seed germination rate based on

improved locally linear embedding
Shu Liu a, Zhengguang Chen a , Feng Jiao b

Show more

Share Cite

https://doi.org/10.1016/j.compag.2022.107514
Get rights and content

Highlights

• The LLE algorithm was improved and applied to maize germination rate
detection.

• The model results of different improvement strategies of LLE are studied and
compared.

• The cosine similarity is suitable as a distance metric for NIR spectral data.

• This study can provide a reference method for other agricultural products
quality testing.

Abstract

Locally linear embedding (LLE) is a nonlinear dimensionality reduction method, which has great
advantages over linear dimensionality reduction methods. However, the traditional LLE takes the
Euclidean distance as the distance measure, which is difficult to accurately reflect the spatial position
relationship between the high-dimensional data of the near-infrared spectrum, resulting in a poor
modeling effect. This paper tries to improve the LLE with different distance metric methods and
proposes a rapid detection method for maize seed germination rate based on improved local linear
embedding and near-infrared spectroscopy. In this paper, a total of 315 samples from 7 different types
of maize seeds, purchased from the seed market, were adopted as the research object. We performed

https://www.sciencedirect.com/science/article/abs/pii/S0168169922008225 1/7
3/28/24, 1:28 PM Detection of maize seed germination rate based on improved locally linear embedding - ScienceDirect

artificial aging tests with 8 different gradients (from 0d to 7d with interval of 1d) on them and
completed the germination rate test after collecting the near-infrared spectral data for each sample.
The Monte Carlo cross-validation (MCCV) algorithm, combined with PLSR and SVM, was used to
remove abnormal samples from the spectral and germination rate data. And then, for comparison, we
used several different improvement strategies for LLE (traditional Euclidean distance, Manhattan
distance, Chebyshev distance, Correlation coefficient, and Cosine similarity) to reduce the spectral
data dimension and established PLS and SVM germination rate prediction models. We compared the
prediction effects of different models to explore the optimal improvement strategy of LLE dimension
reduction distance measurement. The results showed that the cosine similarity was the best
improvement strategy under the same modeling method. The R2 of the LLE_ cos-PLS model's test set
can reach 0.8384, and the R2 of the LLE_cos-SVM model's test set can reach 0.8765. The results showed
that the cosine similarity could better reflect the spatial distribution in the spectral data of aged maize
seeds, and the precision of the model was higher after LLE_cos dimensionality reduction. Compared
with the linear modeling method PLS, the nonlinear modeling method SVM is more suitable for
predicting the germination rate of maize seeds. This study can provide a reference method for the
quality inspection of other agricultural products.

Introduction

Maize is one of China's principal crops and an important food crop globally. Its seed quality is one of
the critical factors in ensuring the yield of maize. In agricultural production, the reduction of grain
production due to the decrease of seed vigor often occurs, causing heavy economic losses to farmers.
Therefore, the detection of maize seed activity is vital to ensuring maize yield. The traditional seed
vigor testing methods include seedling growth testing (Chen and Yuan, 2019), tetrazolium staining
method (Peng et al., 2017), cold resistance testing (Pan et al., 2020), electrical conductivity testing
(Marin et al., 2018), etc. These testing methods are time-consuming, require high professional
requirements for laboratory personnel, and are not easy to popularize. Therefore, how to detect maize
seed vigor quickly, non-destructively, and accurately is an urgent problem for researchers.

Near-infrared spectroscopy has the advantages of being non-destructive, fast, and low-cost. In recent
years, some scholars have applied near-infrared spectroscopy to the field of seed vigor detection.
Ambrose (Ambrose et al., 2016) et al. carried out an aging treatment on three varieties of maize seeds
and established a discriminant model. The classification accuracy of live and non-live corn seeds could
reach 100 %. Santosh (Shrestha et al., 2017) et al. conducted a classification study on the vigor and non-
vigorous tomato seeds of two varieties, and the identification accuracy rate reached 94 %, which could
effectively identify whether the tomato seeds have vitality. Flores (Flores et al., 2020) et al. predicted
the germination rate and vigor of hybrid rice using near-infrared spectroscopy, proving that near-
infrared spectroscopy is a non-destructive method used for seed quality parameter prediction and
seed vigor classification. The above studies have proved the feasibility of near-infrared spectroscopy
analysis technology in detecting seed vigor. However, most of them are qualitative analyses, and there
are few quantitative analysis studies.

Near-infrared spectral data generally have high dimensions and redundant information. Full-spectral
modeling affects the accuracy and calculation speed of the model. To expedite modeling and enhance
the model precision, it is often necessary to reduce the dimension of spectral data prior to modeling.
Amirvaresi (Amirvaresi et al., 2021) et al. established a sample source identification model after

https://www.sciencedirect.com/science/article/abs/pii/S0168169922008225 2/7
3/28/24, 1:28 PM Detection of maize seed germination rate based on improved locally linear embedding - ScienceDirect

principal component analysis (PCA) dimensionality reduction on the spectral data of saffron samples
from Iran. The method can well identify the geographic sources of samples from two regions near Iran.
Pudełko (Pudełko and Chodak, 2020) et al. established partial least squares regression (PLSR), principal
component regression (PCR), and artificial neural network (ANN) based on the whole spectral data,
principal components (PCA-ANN) and latent variables (PLS-ANN) artificial neural network model to
predict the content of total nitrogen and organic carbon in mine soil. Their study showed that ANN
models based on reduced spectral data (PCA-ANN and PLS-ANN) performed better.

In the existing studies, linear dimensionality reduction methods, such as principal component analysis
(PCA) and partial least squares (PLS), are often used to reduce the dimensionality of spectral data.
However, due to the overlapping effect of absorption peaks of different substances in the near-infrared
spectrum, there is a complex nonlinear relationship in the near-infrared spectrum data. The linear
dimensionality reduction method can not effectively account for these non-linear relationships.
Locally Linear Embedding (Roweis and Saul, 2000) (LLE) is a nonlinear dimensionality reduction
method. It can ensure that non-linear information is not lost with dimensionality reduction. LLE
combines the advantages of linear dimension reduction and nonlinear dimension reduction to a
certain extent. In recent years, researchers have conducted many studies using the LLE method.
Boschetti (Boschetti, 2005) used LLE to reduce the dimensionality of different geoscience image
datasets, enabling fast classification of images. Wang (Wang et al., 2015) et al. proposed a mechanical
fault diagnosis method based on a statistical locally linear embedding algorithm, which significantly
improved the classification performance of fault pattern recognition. Duan Yufei (Duan et al., 2016) et
al. developed an egg freshness prediction model using principal component analysis (PCA) and locally
linear embedding after pre-processing the egg NIR spectral data. The results showed that the model
base on the LLE dimension reduction outperformed PCA. Zaman (Zaman, 2020) tested recognition on
multiple public face datasets using an algorithm based on Gabor transformations and locally linear
embedding and showed that the method could significantly improve the accuracy and speed of face
recognition. Zhang (Zhang et al., 2021b) et al. proposed a semi-supervised LLE (SSLLE) method and
verified it on the actual data of the ship's main engine. The results showed that its fault diagnosis
performance was better than the traditional LLE method.

Although LLE has certain advantages over linear dimensionality reduction methods, LLE needs to find
the nearest neighbor points first before reducing the dimension. The original LLE algorithm uses
Euclidean distance as a measure to calculate the nearest neighbor points. Although Euclidean distance
is the most widely used distance measure, it represents the straight line distance between two points,
which does not necessarily reflect the actual spatial distribution relationship between data in high-
dimensional space (Carvalho Júnior et al., 2011).

Therefore, the aim of our study is to improves the distance measurement when the LLE algorithm
selects its K-nearest neighbor and explores a distance measurement method suitable for aging maize
seed spectral data. In this paper, we carried out the artificial aging test of maize seeds, collected the
near-infrared spectral data of seeds with different aging times, and then carried out the germination
test. We used the improved locally linear embedding algorithm to reduce the dimension of the
spectral data and then established the germination rate prediction model to realize the rapid and non-
destructive prediction of the germination rate of maize seeds, which provided a new idea for the
evaluation of seed quality.

https://www.sciencedirect.com/science/article/abs/pii/S0168169922008225 3/7
3/28/24, 1:28 PM Detection of maize seed germination rate based on improved locally linear embedding - ScienceDirect

Section snippets

Materials

The test samples were purchased from the seed market and consisted of seven varieties of maize
seeds, namely Zhengdan, Xianyu, Meiyu, Jingke, Denghai, Jinyu and Demeiya. After removing damaged
and deflated maize seeds, every 200 maize seeds were taken as one sample, making a total of 315
maize samples, including 21 samples of Zhengdan, 9 samples of Xianyu, 22 samples of Meiyu, 18
samples of Jingke, 62 samples of Denghai, 84 samples of Jinyu and 99 samples of Demeiya.…

Seed aging experiments

Germination data of maize…

Eliminating abnormal samples

During the spectral collection and germination rate experiments, there may be some abnormal
samples, also called outliers, whose spectral data or germination rates significantly deviate the true
values. In order to eliminate the influence of outliers samples on the modeling effect, the Monte Carlo
cross-validation (MCCV) (Jiao et al., 2016) algorithm was used in this study to locate the abnormal
samples. The traditional MCCV algorithm uses PLS as the modeling method internally, whose results…

Conclusions

LLE is a manifold learning method that can reduce the dimension of high-dimensional data. However,
the traditional LLE algorithm uses Euclidean distance to select K-nearest neighbors, which cannot
accurately express the spatial distribution relationship between high-dimensional near-infrared
spectral data. In this paper, taking different aged time maize seeds as the research object, we try to
improve the LLE algorithm by using different distance measurement methods instead of Euclidean
distance …

CRediT authorship contribution statement

Shu Liu: Conceptualization, Methodology, Software, Investigation, Writing – original draft.

Zhengguang Chen: Resources, Supervision, Project administration, Writing – review & editing. Feng
Jiao: Supervision, Funding acquisition.…

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships
that could have appeared to influence the work reported in this paper.…

Acknowledgments
This work was supported by the National Natural Science Foundation of China (grant number
41977202) and the Basic Research in "sanzong" Program of Byau (grant number ZRCPY202214).…

https://www.sciencedirect.com/science/article/abs/pii/S0168169922008225 4/7
3/28/24, 1:28 PM Detection of maize seed germination rate based on improved locally linear embedding - ScienceDirect

Recommended articles

References (40)

A. Ambrose et al.
Comparative nondestructive measurement of corn seed viability using Fourier transform
near-infrared (FT-NIR) and Raman spectroscopy
Sens. Actuators B (2016)

A. Amirvaresi et al.
Comparison of near-infrared (NIR) and mid-infrared (MIR) spectroscopy based on
chemometrics for saffron authentication and adulteration detection
Food Chem. (2021)

F. Boschetti
Dimensionality reduction and visualization of geoscientific images via locally linear
embedding
Comput. Geosci. (2005)

L. Deiss et al.
Tuning support vector machines regression models improves prediction accuracy of soil
properties in MIR spectroscopy
Geoderma (2020)

L. Jiao et al.
QSPR study on the flash point of organic binary mixtures by using electrotopological state
index
Chemom. Intel. Lab. Syst. (2016)

T. Ma et al.
Rapid and non-destructive seed viability prediction using near-infrared hyperspectral
imaging coupled with a deep learning approach
Comput. Electron. Agric. (2020)

S. Mirjalili
The ant lion optimizer
Adv. Eng. Softw. (2015)

A. Pudełko et al.
Estimation of total nitrogen and organic carbon contents in mine soils with NIR reflectance
spectroscopy and various chemometric methods
Geoderma (2020)

S. Shrestha et al.

https://www.sciencedirect.com/science/article/abs/pii/S0168169922008225 5/7
3/28/24, 1:28 PM Detection of maize seed germination rate based on improved locally linear embedding - ScienceDirect

Separation of viable and non-viable tomato (Solanum lycopersicum L.) seeds using single
seed near-infrared spectroscopy
Comput. Electron. Agric. (2017)

Y. Wang et al.
Discriminant analysis and comparison of corn seed vigor based on multiband spectrum
Comput. Electron. Agric. (2021)

N. Wu et al.
Deep convolution neural network with weighted loss to detect rice seeds vigor based on
hyperspectral imaging under the sample-imbalanced condition
Comput. Electron. Agric. (2022)

J.D. Bewley et al.

Seeds: Physiology of Development and Germination
(2013)

O.A. Carvalho Júnior et al.

A new approach to change vector analysis using distance and similarity measures
Remote Sens. (Basel) (2011)

Z. Chen et al.
Research progress of Seed Vigor Determination Methods
Seed Sci. Technol. (2019)

C. Cortes et al.
Support-vector networks
Mach. Learn. (1995)

H. Cui et al.
Prediction of maize seed vigor based on first-order difference characteristics of
hyperspectral data
Agronomy (2022)

J. Deng et al.
Effects of aging treatment on germination and seedling growth of different vigor maize
seed
Seed (2021)

Y. Duan et al.
Study on non-destructive detection method for egg freshness based on LLE-SVR and
visible/ near-infrared spectrum
Spectrosc. Spectr. Anal. (2016)

R. Feng et al.
Identification of corn varieties based on Bayesian optimization SVM
Spectrosc. Spectr. Anal. (2022)

https://www.sciencedirect.com/science/article/abs/pii/S0168169922008225 6/7
3/28/24, 1:28 PM Detection of maize seed germination rate based on improved locally linear embedding - ScienceDirect

L. Feng et al.
Identification of maize kernel vigor under different accelerated aging times using
hyperspectral imaging
Molecules (2018)
There are more references available in the full text version of this article.

Cited by (0)

View full text

© 2022 Elsevier B.V. All rights reserved.

All content on this site: Copyright © 2024 Elsevier B.V., its licensors, and contributors. All rights are reserved, including those for text and data mining, AI training, and similar
technologies. For all open access content, the Creative Commons licensing terms apply.

https://www.sciencedirect.com/science/article/abs/pii/S0168169922008225 7/7