The Online Journal of Distance Education and e-Learning, January 2023 Volume 11, Issue 1

A LITERATURE REVIEW STUDY ON: DEVELOPMENT OF MAIZE SEED

CLASSIFICATION AND RECOGNITION PREDICTION SYSTEM USING
MACHINE LEARNING FOR HIGH YIELD MAIZE PRODUCTION
Mrs. Pallavi Ladkat-Gavali
Research Scholar, Assistant Professor,
Bharati Vidyapeeth (IMED)Kothrud, Pune
pallavigavali1@gmail.com

Dr. Kavita Suryawanshi

Research Guide,Professor
D. Y Patil Institute of MCA and Management, Akurdi, Pune
Savitribai Phule Pune University,Pune
kavita1104@yahoo.com

ABSTRACT
In the current research, the importance of maize seed identification and its varieties are appropriately studied.
The machine vision was used to quickly employ methods for checking the seed quality, which helps for better
yield. The proposed methods central aspect helps integrate the machine vision with ML methods to analyse
different parameters such as color, shape, and texture. Typically, ML techniques are used to capture the various
details such as seed quality, pruning, fertilizer application, and environmental and genetic analysis to increase
the yield of maize in agriculture. Eventually, it is important to find an effective integration of ML methods in the
identification of maize seeds in the agriculture field, which helps the farmers to procure better yield and
minimize the loss. Healthy maize seeds are essential for better yield, and various detection methods are utilized
in research to find the best one in agriculture. However, most of the techniques are manual, expensive, and time-
consuming, which makes the process difficult. The proposed techniques that have been utilized in the study
have the possibility to present accurate results that help in identifying similar maize varieties.
Therefore, different parameters are used in the process in order to identify the maize seeds and their varieties to
procure efficient results compared to the existing techniques.
Keywords: Maize seed, Deep convolutional neural network, Convolutional neural network, Long short-term
memory, Machine learning

Introduction
One of the primary crops in the food business is maize, which is subject to abiotic stress such as drought stress,
which has a considerable impact on development and reduces production at significant development levels. As a
result, different stress levels have different negative impacts, particularly on the production and growth of
maize. The levels of drought stress in maize require varying amounts of watering in the interim. In order to
minimize water use, provide a steady and high-yield crop like maize, early prediction and monitoring of drought
stress accuracy are therefore regarded crucial. The traditional approaches to measuring the intensity of a
system's drought stress rely solely on soil moisture sensors, which are inefficient and only effective over small,
indirect areas. Additionally, maize plants develop a number of physiological defense mechanisms to lessen the
effects of drought stress on a phenotypic series of alterations, including leaf color and texture, plant dwarfing,
and a decrease in leaf extension rate. Deep convolutional neural networks (DCNN) have also been used to
categories and identify maize drought stress.

Additionally, the investigation made use of the most recent commercial and scholarly functional breakthroughs
in the FMIS - Farm Management Information System. The studies that have hitherto been conducted mostly
focused on crop output and farm managers who had decision-making authority. The majority of study focuses
on supply chain digitalization combined with smart agricultural practices at the farm level. In fact, farmers have
historically been hesitant to launch, especially in medium and small agri-food firms. Additionally, a variety of
DL - Deep Learning techniques have been used in the studies, including CNN - Convolutional Neural Network,
which has improved adaptability, generalization, and self-learning capabilities and produces better results in
face recognition, object detection, and image classification.

Additionally, a number of techniques have been used to forecast variety, and morphological identification was
limited by morphological traits, environmental variables, and human interference, which has an impact on the
cost or testing time that limits the accuracy of identification. As a result, biochemical identification enables seed
recognition with a variety of genetic traits; yet, it can occasionally be difficult to forecast the types. On the other

www.tojdel.net Copyright © The Online Journal of Distance Education and e-Learning 421
 The Online Journal of Distance Education and e-Learning, January 2023 Volume 11, Issue 1

hand, molecular identification by DNA markers provides advantages for genetic stability and is suited to many
environmental circumstances.

However, it is difficult to apply the majority of current detection techniques in the seed processing industry.
Therefore, it is important to develop quick, effective, and non-destructive techniques for classifying and
identifying the many varieties of maize seeds. Due to its dependability, speed, and accuracy, DL and machine
vision have been merged by numerous researchers. But as seed consumption and variety increase, there are
more demands placed on these studies, and the use of earlier studies has decreased. As a result, many seed
variants are used in deep CNN classifications of various agricultural goods.

Objectives of the study

A literature review is a comprehensive overview of all the knowledge available on a specific topic, ranging from
early research to current developments. It serves several purposes:

• To study literature review related to the development of maize seed classification and recognition
prediction systems using machine learning techniques for high-yield maize production.
• To study literature review on identify the different machine learning techniques and algorithms utilized
for maize seed classification and recognition prediction in the literature.
• To study literature review on explore the various features extracted from maize seeds, such as texture,
color, and shape, to aid in the classification and recognition prediction process.
• To provide insights and recommendations for future research on the development of more accurate and
efficient maize seed classification and recognition prediction systems for high-yield maize production.

Literature Review
A literature review is a comprehensive overview of all the knowledge available on a specific topic, ranging
fromearly research to current developments. It serves several purposes-

Huang, He, Zhu and Qin (2016) Multidimensional scaling and PCA - Principle Component Analysis were
utilized in order to minimize and transform the classification features for solving dimension disaster risk and
formulate the classification approach for maize seeds. The LS-SVM (least squares-support vector machine)
features has been integrated and utilized.

Huang ,Fan, Sun , Shen, and Suo (2019) the main focus of the study was to utilize effective transfer learning and
CNN - convolutional neural network techniques to predict the maize seeds defect appearance. The transfer
learning and CNN has been utilized for quality classification of maize seeds. The visualization technology has
been used to acquire the feature map of CNN network with the heat map to describe the inference results
probability distribution.

Altuntaş, Cömert , Kocamaz (2019) Doubled-haploid is considered as a significant method in maize breeding,
genetic programs, and modern-crop enhancement to maximize the breeding efficiency. Nevertheless, the
breeding process is more difficult and expensive when choosing haploid seeds. This study focuses on this issue
and treats it as a computer vision task that provides a quick, low-cost model for choosing maize seeds. The CNN
methods are utilized to automatically identify diploid and haploid maize seeds via a transfer learning technique
with ResNet, GoogLeNet, WGNet, and AlexNet.

Veeramani, Raymond and Chanda (2017) the current research focused on identifying the corn seeds shape,
embryo pose, and phenotypic expression that captured in the camera. Moreover, the ML-classifiers utilizes
features based on texture, morphology, and color. DeepSort-based CNN used in the realistic settings in maize
production.

Altuntaş, Kocamaz, CÖMERT,Cengiz,& Esmeray,(2018) the main focus of this research to find out the maize
breeding in homozygosity and time. The main process in this technique was the haploid seeds selection.
Therefore, this automation process will enhance the success rate and minimize the time and labor. DH - Doubled
Haploid technique has been utilized in maize breeding.

Javanmardi,Hassan, Martynenko (2020) the high-level (CNN-extracted) and low-level (texture, morphology,
and color) corn seeds features utilized for classification. Moreover, CNN was implemented as a feature extractor
especially in the classification process with the 9 varieties of corn seeds. Deep-CNN has been utilized as a
generic feature extractor were classified with the ANN - Artificial Neural Network, quadratic SVM, SVM -

www.tojdel.net Copyright © The Online Journal of Distance Education and e-Learning 422
 The Online Journal of Distance Education and e-Learning, January 2023 Volume 11, Issue 1

Support Vector Machine, bagged tree, boosted tree, LDA - linear discriminant analysis, and weighted k-nearest-
neighbor.

Ren , Liu, Zhang , Liu 1 , Xi 1 , Kang 3 , Li 1,2 and Zhang 1,2(2020) the effectiveness of the various strategies
has been assessed in the current study's crop classification process. Using sentinel-2 images, it was easy to
distinguish the seed maize from other crops. The 4-classification algorithms includes KNN, SVC, LSTM, and
RF has been utilized in the study and among these LSTM was considered as an effective method.
Chipindu, Mupangwa, Mtsilizah, Nyagumbo, & Zaman-Allah,(2020)In the current research, different
techniques are used to classify and predict the maize kernal abortion among the total images of 66 and
remaining images for performance testing. The techniques such as Binary ML techniques, including k-NN,
CART - classification tree, AdaBoost, SVM, LR - Logistic Regression, and deep CNN, have been utilized in the
detection and classification of maize kernel abortion.

García-Santillán, Montalvo & Pajares, (2017) identifying straight and curved crop rows in gathered images that
are captured at the crop’s initial stage in maize fields. This sequence categorizes into three phases: image
segmentation, Initial stage identification, and crop row prediction. The DAGP - Detection by Accumulation of
Green pixels has been utilized for predicting straight, curved crop rows. The performance has been studied
through quantitative analysis and the techniques utilized. Further, Standard Hough Transform has been used to
predict straight crop rows.

Dönmez,(2020) the technique used for maize Seeds classification as diploid and haploid utilizing pre-trained
CNN. Therefore, 3000 diploid & haploid seeds dataset has been used in the experimentation. Further, the maize
seed classification was processed computer-aided with the deep features acquired using the pre-trained CNN
method. Eventually, these can be applied to different agricultural products, The categorization of maize seeds as
diploid and haploid uses pre-trained CNN and classifier-based techniques including SVM, k-nearest neighbour,
and DT - Decision Tree methods.

Zhao,C.,Quan, Liu, R., Wang, Feng & Sin2020) This study focuses on electromagnetic vibration application
exploration and DL-technology in the prediction of maize kernel. Using color images as input data and
emphasizing on the quality concerns while identifying the maize kernels, the Faster R-CNN model has been
utilized to choose maize kernels.

Research Gap
In this literature review study identified the research gap is lack of comparative analysis & Most of the reviewed
studies utilize different machine learning techniques to classify and recognize maize seeds. However, there is a
lack of comparative analysis among the various methods, which can help identify the most effective approach
for high yield maize production. Limited focused on specific seed characteristics Some of the studies focus on
specific seed characteristics, such as haploid and diploid seeds or maize kernel abortion, but do not address other
essential features that can affect maize production.

Inadequate consideration of practical implementation While the reviewed studies show promising results in
maize seed classification and recognition, there is a lack of consideration for practical implementation of these
techniques in real-world settings, such as on farms and in seed production facilities. Limited exploration of
novel machine learning techniques Although the reviewed studies utilize various machine learning techniques,
there is a lack of exploration of novel techniques, such as deep reinforcement learning or transfer learning,
which can enhance the accuracy and efficiency of maize seed classification and recognition systems.

Conclusion
This literature review study demonstrates the importance and effectiveness of developing accurate and efficient
classification approaches for maize seeds, which can be useful in various agricultural applications such as seed
sorting and quality control. The use of advanced techniques such as deep learning and computer vision can
significantly improve the accuracy and efficiency of seed classification, and potentially reduce the cost and time
required for manual classification. The literature review indicates that machine learning algorithms such as CNN
and SVM are effective in seed classification and recognition.

The literature review suggests that there is a gap in research on the use of machine learning for seed
classification and recognition in maize production. The proposed system can accurately classify and recognize
maize seeds based on their quality, size, and other characteristics, which can help farmers select the best seeds
for planting and improve the overall yield. Overall, the proposed system has the potential to revolutionize maize

www.tojdel.net Copyright © The Online Journal of Distance Education and e-Learning 423
 The Online Journal of Distance Education and e-Learning, January 2023 Volume 11, Issue 1

production by providing accurate and efficient seed selection, which can lead to higher yields and profits for
farmers.

References
Altuntaş, Y., Cömert, Z., & Kocamaz, A. F. (2019).”Identification of haploid and diploid maize seeds using
convolutional neural networks and a transfer learning approach. Computers and Electronics in
Agriculture”, 163, 104874.
Altuntaş, Y., Kocamaz, A. F., CÖMERT, Z., Cengiz, R., & Esmeray, M. (2018, September). “Identification of
haploid maize seeds using gray level co-occurrence matrix and machine learning techniques”. In 2018
International Conference on Artificial Intelligence and Data Processing (IDAP) (pp. 1-5). Ieee.
An, J., Li, W., Li, M., Cui, S., & Yue, H. (2019). “Identification and classification of maize drought stress using
deep convolutional neural network”. Symmetry, 11(2), 256.
Chipindu, L., Mupangwa, W., Mtsilizah, J., Nyagumbo, I., & Zaman-Allah, M. (2020). “Maize kernel abortion
recognition and classification using binary classification machine learning algorithms and deep
convolutional neural networks.” AI, 1(3), 361.
Dönmez, E. (2020). “Classification of Haploid and Diploid Maize Seeds based on Pre-Trained Convolutional
Neural Networks”. Celal Bayar University Journal of Science,16(3), 323-331.
Fountas, S., Carli, G., Sørensen, C. G., Tsiropoulos, Z., Cavalaris, C., Vatsanidou, A., & Tisserye, B. (2015).
“Farm management information systems: Current situation and future perspectives. Computers and
Electronics in Agriculture”, 115, 40-50.
Giua, C., Materia, V. C., & Camanzi, L. (2020). “Management information system adoption at the farm level:
evidence from the literature”. British Food Journal.
García-Santillán, I., Guerrero, J. M., Montalvo, M., & Pajares, G. (2018). “Curved and straight crop row
detection by accumulation of green pixels from images in maize fields”. Precision Agriculture, 19(1), 18-
41.
Huang, M., He, C., Zhu, Q., & Qin, J. (2016). “Maize seed variety classification using the integration of spectral
and image features combined with feature transformation based on hyperspectral imaging”. Applied
Sciences, 6(6), 183.
Huang, S., Fan, X., Sun, L., Shen, Y., & Suo, X. (2019). “Research on classification method of maize seed
defect based on machine vision”. Journal of Sensors, 2019.
Javanmardi, S., Ashtiani, S. H. M., Verbeek, F. J., & Martynenko, A. (2021).” Computer-vision classification of
corn seed varieties using deep convolutional neural network”. Journal of Stored Products Research, 92,
101800.
Xu, P., Tan, Q., Zhang, Y., Zha, X., Yang, S., & Yang, R. (2022). “Research on Maize Seed Classification and
Recognition Based on Machine Vision and Deep Learning. Agriculture”, 12(2), 232.
Xu, P., Tan, Q., Zhang, Y., Zha, X., Yang, S., & Yang, R. (2022). “Research on Maize Seed Classification and
Recognition Based on Machine Vision and Deep Learning”. Agriculture, 12(2), 232.
Yu, L., Liu, W., Li, W., Qin, H., Xu, J., & Zuo, M.(2018). “Non-destructive identification of maize haploid
seeds using nonlinear analysis method based on their near-infrared spectra”. Biosystems engineering,
172, 144-153.
Zhao, C., Quan, L., Li, H., Liu, R., Wang, J., Feng, H.,& Sin, K. (2020). “Precise selection and visualization of
maize kernels based on electromagnetic vibration and deep learning”. Transactions of the ASABE, 63(3),
629-643.
Zhang, L., Liu, Z., Ren, T., Liu, D., Ma, Z., Tong, L.,& Li, S. (2020). “Identification of seed maize fields with
high spatial resolution and multiple spectral remote sensing using random forest classifier”. Remote
Sensing, 12(3), 362.

www.tojdel.net Copyright © The Online Journal of Distance Education and e-Learning 424