Deep Learning-Based Robust Analysis of Laser Bio-Speckle Data for Detection of Fungal-Infected Soybean Seeds

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F62690094%3A18450%2F23%3A50020635" target="_blank" >RIV/62690094:18450/23:50020635 - isvavai.cz</a>

Výsledek na webu

<a href="https://ieeexplore.ieee.org/document/10216973" target="_blank" >https://ieeexplore.ieee.org/document/10216973</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/ACCESS.2023.3305273" target="_blank" >10.1109/ACCESS.2023.3305273</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Deep Learning-Based Robust Analysis of Laser Bio-Speckle Data for Detection of Fungal-Infected Soybean Seeds

Popis výsledku v původním jazyce

Seed-borne diseases play a crucial role in affecting the overall quality of seeds, efficient disease management, and crop productivity in agriculture. Detection of seed-borne diseases using machine learning (ML) and deep learning (DL) can automate the process at large-scale industrial applications for providing healthy and high-quality seeds. ML-based methods are accurate for detecting and classifying fungal infection in seeds; however, their performance degrades in the presence of noise. In this work, we propose a laser bio-speckle based DL framework for detection and classification of disease in seeds under varying experimental parameters and noises. We develop a DL-based spatio-temporal analysis technique for bio-speckle data using DL networks, including neural networks (NN), convolutional neural networks (CNN) with long-short-term memory (LSTM), three-dimensional convolutional neural networks (3D CNN), and convolutional LSTM (ConvLSTM). The robustness of the DL models to noise is a key aspect of this spatio-temporal analysis. In this study, we find that the ConvLSTM model has an accuracy of 97.72% on the test data and is robust to different types of noises with an accuracy of 97.72%, 94.31%, 98.86%, and 96.59%. Furthermore, the robust model (ConvLSTM) is evaluated for variations in experimental data parameters such as frame rate, frame size, and number of frames used. This model is also sensitive towards detecting bio-speckle activity of different order, and it shows average test accuracy of 99% for detecting four different classes.

Název v anglickém jazyce

Deep Learning-Based Robust Analysis of Laser Bio-Speckle Data for Detection of Fungal-Infected Soybean Seeds

Popis výsledku anglicky

Seed-borne diseases play a crucial role in affecting the overall quality of seeds, efficient disease management, and crop productivity in agriculture. Detection of seed-borne diseases using machine learning (ML) and deep learning (DL) can automate the process at large-scale industrial applications for providing healthy and high-quality seeds. ML-based methods are accurate for detecting and classifying fungal infection in seeds; however, their performance degrades in the presence of noise. In this work, we propose a laser bio-speckle based DL framework for detection and classification of disease in seeds under varying experimental parameters and noises. We develop a DL-based spatio-temporal analysis technique for bio-speckle data using DL networks, including neural networks (NN), convolutional neural networks (CNN) with long-short-term memory (LSTM), three-dimensional convolutional neural networks (3D CNN), and convolutional LSTM (ConvLSTM). The robustness of the DL models to noise is a key aspect of this spatio-temporal analysis. In this study, we find that the ConvLSTM model has an accuracy of 97.72% on the test data and is robust to different types of noises with an accuracy of 97.72%, 94.31%, 98.86%, and 96.59%. Furthermore, the robust model (ConvLSTM) is evaluated for variations in experimental data parameters such as frame rate, frame size, and number of frames used. This model is also sensitive towards detecting bio-speckle activity of different order, and it shows average test accuracy of 99% for detecting four different classes.

Druh

J_imp - Článek v periodiku v databázi Web of Science

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2023

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

IEEE Access

ISSN

2169-3536

e-ISSN

2169-3536

Svazek periodika

11

Číslo periodika v rámci svazku

August

Stát vydavatele periodika

US - Spojené státy americké

Počet stran výsledku

18

Strana od-do

89331-89348

Kód UT WoS článku

001058758000001

EID výsledku v databázi Scopus

2-s2.0-85168296417