Classification of Citrus Leaf Diseases Using Hyperspectral Reflectance and Fluorescence Imaging and Machine Learning Techniques DOI Creative Commons
Hyun Jung Min, Jianwei Qin, Pappu Kumar Yadav

et al.

Horticulturae, Journal Year: 2024, Volume and Issue: 10(11), P. 1124 - 1124

Published: Oct. 22, 2024

Citrus diseases are significant threats to citrus groves, causing financial losses through reduced fruit size, blemishes, premature drop, and tree death. The detection of via leaf inspection can improve grove management mitigation efforts. This study explores the potential a portable reflectance fluorescence hyperspectral imaging (HSI) system for detecting classifying control group diseases, including canker, Huanglongbing (HLB), greasy spot, melanose, scab, zinc deficiency. HSI was used simultaneously collect images from front back sides leaves. Nine machine learning classifiers were trained using full spectra spectral bands selected principal component analysis (PCA) with pixel-based leaf-based spectra. A support vector (SVM) classifier achieved highest overall classification accuracy 90.7% when employing combined data side leaves, whereas discriminant yielded best 94.5% analysis. Among control, melanose classified most accurately, each over 90% accuracy. Therefore, integration advanced techniques demonstrated capability accurately detect classify these high precision.

Language: Английский

Cross-comparative review of Machine learning for plant disease detection: apple, cassava, cotton and potato plants DOI Creative Commons

James Daniel Omaye,

Emeka Ogbuju, Grace Ataguba

et al.

Artificial Intelligence in Agriculture, Journal Year: 2024, Volume and Issue: 12, P. 127 - 151

Published: May 13, 2024

Plant disease detection has played a significant role in combating plant diseases that pose threat to global agriculture and food security. Detecting these early can help mitigate their impact ensure healthy crop yields. Machine learning algorithms have emerged as powerful tools for accurately identifying classifying wide range of from trained image datasets affected crops. These algorithms, including deep shown remarkable success recognizing patterns signs diseases. Besides detection, there are other potential benefits machine overall management, such soil climatic condition predictions plants, pest identification, proximity many more. Over the years, research focused on using machine-learning detection. Nevertheless, little is known about extent which community explored cover areas management. In view this, we present cross-comparative review applications designed with specific focus four (4) economically important plants: apple, cassava, cotton, potato. We conducted systematic articles published between 2013 2023 explore trends over years. After filtering number based our inclusion criteria, individual prediction accuracy classes associated selected 113 were considered relevant. From articles, analyzed state-of-the-art techniques, challenges, future prospects identification plants. Results show performed significantly well detecting addition, found few references management covering prevention, diagnosis, control, monitoring. or no work recovery Hence, propose opportunities developing learning-based technologies monitoring, recovery.

Language: Английский

Citations

7
Detection of apple mosaic based on hyperspectral imaging and three-dimensional Gabor DOI
Yanfu Liu, Xiaonan Zhao,

Zhenghua Song

et al.

Computers and Electronics in Agriculture, Journal Year: 2024, Volume and Issue: 222, P. 109051 - 109051

Published: May 28, 2024

Language: Английский

Citations

5
Hyperspectral Remote Sensing for Early Detection of Wheat Leaf Rust Caused by Puccinia triticina DOI Creative Commons
Anton Terentev, Vladimir Badenko,

Е. L. Shaydayuk

et al.

Agriculture, Journal Year: 2023, Volume and Issue: 13(6), P. 1186 - 1186

Published: June 2, 2023

Early crop disease detection is one of the most important tasks in plant protection. The purpose this work was to evaluate early wheat leaf rust possibility using hyperspectral remote sensing. first task study choose tools for processing and analyze sensing data. second biochemical profile by chromatographic spectrophotometric methods. third discuss a possible relationship between data results from leaves, analysis. used an interdisciplinary approach, including methods, as well As result, (1) VIS-NIR spectrometry analysis showed high correlation with data; (2) wavebands identification were revealed (502, 466, 598, 718, 534, 766, 694, 650, 866, 602, 858 nm). An accuracy 97–100% achieved fourth dai (day/s after inoculation) SVM.

Language: Английский

Citations

11
Identification of tomato bacterial wilt severity based on hyperspectral imaging technology and spectrum Transformer network DOI
Xin Wang, Wei Yang, Yang Yu

et al.

Ecological Informatics, Journal Year: 2023, Volume and Issue: 78, P. 102353 - 102353

Published: Oct. 24, 2023

Language: Английский

Citations

11
Study on the influence of specular reflection on vegetation index and its elimination method DOI

Guofeng Zhang,

Siyuan Li, Yongcun Zhao

et al.

Computers and Electronics in Agriculture, Journal Year: 2025, Volume and Issue: 231, P. 110051 - 110051

Published: Feb. 4, 2025

Language: Английский

Citations

0
Enhancing field-scale soil moisture content monitoring using UAV hyperspectral-derived multi-dimensional spectral response indices of crop comprehensive phenotypic traits DOI
Hao Liu, Junying Chen, Jiang Bian

et al.

Computers and Electronics in Agriculture, Journal Year: 2025, Volume and Issue: 235, P. 110399 - 110399

Published: April 15, 2025

Language: Английский

Citations

0
Utilizing Hyperspectral Reflectance and Machine Learning Algorithms for Non-Destructive Estimation of Chlorophyll Content in Citrus Leaves DOI Creative Commons

Dasui Li,

Qingqing Hu,

Siqi Ruan

et al.

Remote Sensing, Journal Year: 2023, Volume and Issue: 15(20), P. 4934 - 4934

Published: Oct. 12, 2023

To address the demands of precision agriculture and measurement plant photosynthetic response nitrogen status, it is necessary to employ advanced methods for estimating chlorophyll content quickly non-destructively at a large scale. Therefore, we explored utilization both linear regression machine learning methodology improve prediction leaf (LCC) in citrus trees through analysis hyperspectral reflectance data field experiment. And relationship between phenology LCC estimation was also tested this study. The tree leaves five growth seasons (May, June, August, October, December) were measured alongside measurements reflectance. spectral parameters used evaluating using univariate (ULR), multivariate (MLR), random forest (RFR), K-nearest neighbor (KNNR), support vector (SVR). results revealed following: MLR models (RFR, KNNR, SVR), October December, performed well with coefficient determination (R2) greater than 0.70. In ULR model best, achieving an R2 0.69 root mean square error (RMSE) 8.92. However, RFR demonstrated highest predictive power May, December. Furthermore, accuracy best VOG2 Carte4 0.83 RMSE 6.67. Our findings that just few can efficiently estimate trees, showing substantial promise implementation large-scale orchards.

Language: Английский

Citations

10
Early Detection of Rubber Tree Powdery Mildew by Combining Spectral and Physicochemical Parameter Features DOI Creative Commons

Xiangzhe Cheng,

Mengning Huang,

Anting Guo

et al.

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(9), P. 1634 - 1634

Published: May 3, 2024

Powdery mildew significantly impacts the yield of natural rubber by being one predominant diseases that affect trees. Accurate, non-destructive recognition powdery in early stage is essential for cultivation management The objective this study to establish a technique detection trees combining spectral and physicochemical parameter features. At three field experiment sites laboratory, spectroradiometer hand-held optical leaf-clip meter were utilized, respectively, measure hyperspectral reflectance data (350–2500 nm) both healthy early-stage powdery-mildew-infected leaves. Initially, vegetation indices extracted from data, wavelet energy coefficients obtained through continuous transform (CWT). Subsequently, significant (VIs) selected using ReliefF algorithm, optimal wavelengths (OWs) chosen via competitive adaptive reweighted sampling. Principal component analysis was used dimensionality reduction coefficients, resulting features (WFs). To evaluate capability aforementioned features, above, along with their combinations (PFs) (VIs + PFs, OWs WFs PFs), construct six classes In turn, these input into support vector machine (SVM), random forest (RF), logistic regression (LR), build models results revealed based on perform well, markedly outperforming those constructed VIs as inputs. Moreover, incorporating combined surpass relying single an overall accuracy (OA) improvement over 1.9% increase F1-Score 0.012. model combines PFs shows superior performance all other models, achieving OAs 94.3%, 90.6%, 93.4%, F1-Scores 0.952, 0.917, 0.941 SVM, RF, LR, respectively. Compared alone, improved 1.9%, 2.8%, increased 0.017, 0.016, This showcases viability

Language: Английский

Citations

3
Hyperspectral Analysis and Regression Modeling of SPAD Measurements in Leaves of Three Mangrove Species DOI Open Access
Huazhe Li, Lijuan Cui, Zhiguo Dou

et al.

Forests, Journal Year: 2023, Volume and Issue: 14(8), P. 1566 - 1566

Published: July 31, 2023

Mangroves have important roles in regulating climate change, and reducing the impact of wind waves. Analysis chlorophyll content mangroves is for monitoring their health, conservation management. Thus, this study aimed to apply four regression models, eXtreme Gradient Boosting (XGBoost), Random Forest (RF), Partial Least Squares (PLS) Adaptive (AdaBoost), inversion Soil Plant Development (SPAD) values obtained from near-ground hyperspectral data three dominant species, Bruguiera sexangula (Lour.) Poir. (B. sexangula), Ceriops tagal (Perr.) C. B. Rob. (C. tagal) Rhizophora apiculata Blume (R. apiculata) Qinglan Port Mangrove Nature Reserve. The accuracy model was evaluated using R2, RMSE, MAE. mean SPAD R. (SPADavg = 66.57), with a smaller dispersion (coefficient variation 6.59%), were higher than those 61.56) 58.60). first-order differential transformation spectral improved prediction model; R2 mostly distributed interval 0.4 0.8. XGBoost less affected by species differences best stability, RMSE at approximately 3.5 MAE 2.85. This provides technical reference large-scale detection management mangroves.

Language: Английский

Citations

8
Amplifying Apple Mosaic Illness Detection: Combining CNN and Random Forest Models DOI
Arshleen Kaur, Vinay Kukreja,

Priyanshi Aggarwal

et al.

Published: March 14, 2024

Global apple harvests are seriously threatened by Apple Mosaic Disease (AMD), which calls for accurate identification and scalable control measures. This study uses Deep Learning (DL) Convolutional Neural Networks (CNN) Random Forest (RF) models to investigate AMD categorization across four severity levels. The carefully chooses a broad leaf dataset that includes both healthy diseased samples. To guarantee consistency capture minor differences, the is rigorously preprocessed. large serves as foundation training RF CNN models, allows them identify complex patterns unique AMD. put through extensive get thorough understanding of complexity AMD, rigorous validation procedures used refine parameters improve flexibility. Diverse performance indicators highlight advantages disadvantages model in an examination unpublished simulates real-world situations. performs admirably, with 97.08% diagnosis accuracy demonstrating its superior ability comprehend disease patterns. On other hand, phases levels effectively distinguished model. represents significant advancement treatment agricultural diseases developing accurate, automated methods quick detection. Combining state-of-the-art DL conventional could strengthen crop protection, allow prompt interventions, maximize resource allocation sustainable farming methods.

Language: Английский

Citations

3