DM_CorrMatch: A Semi-Supervised Semantic Segmentation Framework for Rapeseed Flower Coverage Estimation Using UAV Imagery DOI Creative Commons
Jie Li, Chengyong Zhu, Chenbo Yang

et al.

Research Square (Research Square), Journal Year: 2025, Volume and Issue: unknown

Published: March 28, 2025

Abstract Background Rapeseed(Brassica napus L.) inflorescence coverage is a crucial phenotypic parameter for assessing crop growth and estimating yield. Accurate cover assessment typically performed using Unmanned Aerial Vehicles (UAVs) in combination with semantic segmentation methods. However, the irregular variable morphology of rapeseed inflorescences presents significant challenges segmentation. To address these challenges, advanced methods that can improve accuracy, particularly under limited data conditions, are needed. Results In this study, we propose cost-effective high-throughput approach semi-supervised learning framework, DM_CorrMatch. This method enhances input images through strong weak augmentation techniques, while leveraging Denoising Diffusion Probabilistic Model (DDPM) to generate additional samples data-scarce scenarios.We an automatic update strategy labeled dilute proportion erroneous labels manual Furthermore, novel network architecture, Mamba-Deeplabv3+, proposed, combining strengths Mamba Convolutional Neural Networks (CNNs) both global local feature extraction. architecture effectively captures key features, even varying poses, reducing influence complex backgrounds. The proposed validated on Rapeseed Flower Segmentation Dataset (RFSD), which consists 720 UAV from Yangluo experimental station Oil Crops Research Institute Chinese Academy Agricultural Sciences (CAAS). results showed our outperforms four traditional eleven deep methods, achieving Intersection over Union (IoU) 0.886, Precision 0.942, Recall 0.940. Conclusions The learning-based method, combined Mamba-Deeplabv3+ demonstrates superior performance accurately segmenting challenging conditions. Our handles backgrounds various poses inflorescences, providing reliable tool flower estimation. aid development high-yield cultivars monitoring UAV-based technologies.

Language: Английский

A review of regional and Global scale Land Use/Land Cover (LULC) mapping products generated from satellite remote sensing DOI
Yanzhao Wang, Yonghua Sun, X. L. Cao

et al.

ISPRS Journal of Photogrammetry and Remote Sensing, Journal Year: 2023, Volume and Issue: 206, P. 311 - 334

Published: Nov. 28, 2023

Language: Английский

Citations

61
A research review on deep learning combined with hyperspectral Imaging in multiscale agricultural sensing DOI

Luyu Shuai,

Zhiyong Li, Ziao Chen

et al.

Computers and Electronics in Agriculture, Journal Year: 2024, Volume and Issue: 217, P. 108577 - 108577

Published: Jan. 5, 2024

Language: Английский

Citations

35
Applications of hyperspectral imaging technology in the food industry DOI
Da‐Wen Sun, Hongbin Pu, Jingxiao Yu

et al.

Nature Reviews Electrical Engineering, Journal Year: 2024, Volume and Issue: 1(4), P. 251 - 263

Published: March 26, 2024

Language: Английский

Citations

25
Hyperspectral Image Analysis and Machine Learning Techniques for Crop Disease Detection and Identification: A Review DOI Open Access
Yimy Edisson García Vera, Mauricio Andrés Polochè Arango, Camilo A. Mendivelso-Fajardo

et al.

Sustainability, Journal Year: 2024, Volume and Issue: 16(14), P. 6064 - 6064

Published: July 16, 2024

Originally, the use of hyperspectral images was for military applications, but their has been extended to precision agriculture. In particular, they are used activities related crop classification or disease detection, combining these with machine learning techniques and algorithms. The study a wide range wavelengths observation. These allow monitoring agricultural crops such as cereals, oilseeds, vegetables, fruits, other applications. ranges wavelengths, conditions maturity index nutrient status, early detection some diseases that cause losses in crops, can be studied diagnosed. Therefore, this article proposes technical review main applications perspectives challenges combine artificial intelligence algorithms deep vegetables. A systematic scientific literature carried out using 10-year observation window determine evolution integration technological tools support sustainable agriculture; among findings, information on most documented is highlighted, which cereals citrus fruits due high demand large cultivation areas, well vegetables integrating technologies. Also, being worked summarized classified, wavelength prediction, analysis tasks physiological characteristics production. This useful reference future research, based mainly classification, decision making, implement appropriate

Language: Английский

Citations

20
Hyperspectral image classification using graph convolutional network: A comprehensive review DOI

Guoyong Wu,

Mohammed A. A. Al‐qaness, Dalal AL-Alimi

et al.

Expert Systems with Applications, Journal Year: 2024, Volume and Issue: 257, P. 125106 - 125106

Published: Aug. 18, 2024

Language: Английский

Citations

16
Land Use and Land Cover Classification Meets Deep Learning: A Review DOI Creative Commons
Shengyu Zhao,

Kaiwen Tu,

Shutong Ye

et al.

Sensors, Journal Year: 2023, Volume and Issue: 23(21), P. 8966 - 8966

Published: Nov. 3, 2023

As one of the important components Earth observation technology, land use and cover (LULC) image classification plays an essential role. It uses remote sensing techniques to classify specific categories ground as a means analyzing understanding natural attributes Earth’s surface state use. provides information for applications in environmental protection, urban planning, resource management. However, images are usually high-dimensional data have limited available labeled samples, so performing LULC task faces great challenges. In recent years, due emergence deep learning processing methods based on achieved remarkable results, bringing new possibilities research development classification. this paper, we present systematic review deep-learning-based classification, mainly covering following five aspects: (1) introduction main typical networks, how they work, their unique benefits; (2) summary two baseline datasets (pixel-level, patch-level) performance metrics evaluating different models (OA, AA, F1, MIOU); (3) strategies studies, including convolutional neural networks (CNNs), autoencoders (AEs), generative adversarial (GANs), recurrent (RNNs); (4) challenges faced by schemes under training samples; (5) outlooks future

Language: Английский

Citations

30
Appraisal of EnMAP hyperspectral imagery use in LULC mapping when combined with machine learning pixel-based classifiers DOI
Christina Lekka, George P. Petropoulos, Spyridon E. Detsikas

et al.

Environmental Modelling & Software, Journal Year: 2024, Volume and Issue: 173, P. 105956 - 105956

Published: Jan. 10, 2024

Language: Английский

Citations

14
Utilizing a fusion of remote sensing data and machine learning models to forecast flood risks to agriculture in Hanoi City, Vietnam DOI
Anh Ngoc Thi

Letters in Spatial and Resource Sciences, Journal Year: 2024, Volume and Issue: 17(1)

Published: June 24, 2024

Language: Английский

Citations

8
An overview of recent advancements in hyperspectral imaging in the egg and hatchery industry DOI Creative Commons
Md Wadud Ahmed, Alin Khaliduzzaman, J.L. Emmert

et al.

Computers and Electronics in Agriculture, Journal Year: 2024, Volume and Issue: 230, P. 109847 - 109847

Published: Dec. 18, 2024

Language: Английский

Citations

8
A Comparison of Deep Transfer Learning Methods for Land Use and Land Cover Classification DOI Open Access
Hatef Dastour, Quazi K. Hassan

Sustainability, Journal Year: 2023, Volume and Issue: 15(10), P. 7854 - 7854

Published: May 11, 2023

The pace of Land Use/Land Cover (LULC) change has accelerated due to population growth, industrialization, and economic development. To understand analyze this transformation, it is essential examine changes in LULC meticulously. classification a fundamental complex task that plays significant role farming decision making urban planning for long-term development the earth observation system. Recent advances deep learning, transfer remote sensing technology have simplified problem. Deep learning particularly useful addressing issue insufficient training data because reduces need equally distributed data. In study, thirty-nine models were systematically evaluated alongside multiple using consistent set criteria. Our experiments will be conducted under controlled conditions provide valuable insights future research on models. Among our models, ResNet50, EfficientNetV2B0, ResNet152 top performers terms kappa accuracy scores. required three times longer time than EfficientNetV2B0 test computer, while ResNet50 took roughly twice as long. achieved an overall f1-score 0.967 set, with Highway class having lowest score Sea Lake highest.

Language: Английский

Citations

17