High-Precision Tea Plantation Mapping with Multi-Source Remote Sensing and Deep Learning DOI Creative Commons

Yicheng Zhou,

Lingbo Yang, Lin Yuan

et al.

Agronomy, Journal Year: 2024, Volume and Issue: 14(12), P. 2986 - 2986

Published: Dec. 15, 2024

Accurate mapping of tea plantations is crucial for agricultural management and economic planning, yet it poses a significant challenge due to the complex variable nature cultivation landscapes. This study presents high-precision approach in Anji County, Zhejiang Province, China, utilizing multi-source remote sensing data advanced deep learning models. We employed combination Sentinel-2 optical imagery, Sentinel-1 synthetic aperture radar digital elevation models capture rich spatial, spectral, temporal characteristics plantations. Three models, namely U-Net, SE-UNet, Swin-UNet, were constructed trained semantic segmentation Cross-validation point-based accuracy assessment methods used evaluate performance The results demonstrated that Swin-UNet model, transformer-based capturing long-range dependencies global context superior feature extraction, outperformed others, achieving an overall 0.993 F1-score 0.977 when using multi-temporal data. integration with slightly improved classification accuracy, particularly areas affected by cloud cover, highlighting complementary imagery all-weather monitoring. also analyzed influence terrain factors, such as elevation, slope, aspect, on plantation mapping. It was found at higher altitudes or north-facing slopes exhibited improves increasing likely simpler land cover types tea’s preference shade. findings this research not only provide valuable insights into precision but contribute broader application

Language: Английский

Automated shoreline extraction process for unmanned vehicles via U-net with heuristic algorithm DOI Creative Commons
Katarzyna Prokop, Dawid Połap, Marta Włodarczyk-Sielicka

et al.

Alexandria Engineering Journal, Journal Year: 2024, Volume and Issue: 102, P. 108 - 118

Published: June 5, 2024

Detecting the shoreline is an important task for its potential use. The allows cropping of image into two separate areas that present water area and shore. It particularly interesting because images can be used to analyze pollution, land development, or even waterfront erosion. Unfortunately, automatic detection a complex problem due numerous physical atmospheric issues. In this paper, we solution based on U-net convolutional network, trained dedicated database. database automatically generated by applying processing techniques heuristic algorithm. Using heuristics, optimal values mask generation parameters are determined. Consequently, automation generating set masks analyzing boundary line efficiency segmentation network. proposed analysis coastline, where obstacles occurring waves quickly detected. To evaluate solution, tests were carried out in real conditions, which showed effectiveness model. addition, publicly available database, allowed obtaining higher results than existing methods.

Language: Английский

Citations

1
Evaluation of Deep Learning Transfer Techniques for Mangrove Segmentation with Images of the Sentinel-2A DOI
André Luís Pereira de Albuquerque,

Helyane Bronoski Borges

Published: Sept. 30, 2024

Fine-tuning techniques allow the use of weights from pre-trained networks in other models across different contexts, potentially improving training performance as it generally requires fewer computational resources and less data. Finetuning has become more widespread natural domain (RGB) with availability model ImageNet database. However, same are not readily available for remote sensing domain, such mangrove identification. Both nationally state Paraná, there few studies employing deep learning segmentation. Developing using transfer can help establish automated monitoring systems. Thus, this study evaluated fine-tuning segmentation Paraná U-Net encoders sensing, domain. The dataset was generated bands Sentinel-2A satellite annotations MapBiomas project maps. fine-tuned discussed accurately identified mangroves all achieving accuracies above 95.1% F-scores greater than 92.6%.

Language: Английский

Citations

0
Rapid Expansion of Coastal Mangrove Forest in Guangxi Beibu Gulf: Patterns, Drivers, and Impacts DOI Creative Commons
Ziyu Sun, Weiguo Jiang, Ziyan Ling

et al.

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Journal Year: 2024, Volume and Issue: 18, P. 510 - 522

Published: Sept. 5, 2024

Language: Английский

Citations

0
High-Precision Tea Plantation Mapping with Multi-Source Remote Sensing and Deep Learning DOI Creative Commons

Yicheng Zhou,

Lingbo Yang, Lin Yuan

et al.

Agronomy, Journal Year: 2024, Volume and Issue: 14(12), P. 2986 - 2986

Published: Dec. 15, 2024

Accurate mapping of tea plantations is crucial for agricultural management and economic planning, yet it poses a significant challenge due to the complex variable nature cultivation landscapes. This study presents high-precision approach in Anji County, Zhejiang Province, China, utilizing multi-source remote sensing data advanced deep learning models. We employed combination Sentinel-2 optical imagery, Sentinel-1 synthetic aperture radar digital elevation models capture rich spatial, spectral, temporal characteristics plantations. Three models, namely U-Net, SE-UNet, Swin-UNet, were constructed trained semantic segmentation Cross-validation point-based accuracy assessment methods used evaluate performance The results demonstrated that Swin-UNet model, transformer-based capturing long-range dependencies global context superior feature extraction, outperformed others, achieving an overall 0.993 F1-score 0.977 when using multi-temporal data. integration with slightly improved classification accuracy, particularly areas affected by cloud cover, highlighting complementary imagery all-weather monitoring. also analyzed influence terrain factors, such as elevation, slope, aspect, on plantation mapping. It was found at higher altitudes or north-facing slopes exhibited improves increasing likely simpler land cover types tea’s preference shade. findings this research not only provide valuable insights into precision but contribute broader application

Language: Английский

Citations

0