The dataset of main grain land changes in China over 1985–2020 DOI Creative Commons
Shidong Liu, Li Wang, Jie Zhang

et al.

Scientific Data, Journal Year: 2024, Volume and Issue: 11(1)

Published: Dec. 24, 2024

Continuous, Accurate, and detailed information on main grain land (MGL) areas is crucial for provisioning food security making policies affecting sustainable agricultural production. It still lacks a long-term MGL distribution dataset with fine spatial resolution. This study aimed to produce long-term, high-resolution map China. Here, we developed the change of resolution 30 m in China period 1985–2020 using Landsat image-based random forest algorithm GEE platform. Finally, planting intensity, gain time loss was calculated. Results indicate that our mapping results are highly consistent annual area various crops according national statistics. A validation based 3113 field survey samples 30-m showed overall accuracy were 93.57%. The full freely available at https://doi.org/10.6084/m9.figshare.26212643.v2 .

Language: Английский

Adaptive Month Matching: A Phenological Alignment Method for Transfer Learning in Cropland Segmentation DOI Creative Commons
Reza Maleki, Falin Wu,

G. Qu

et al.

Remote Sensing, Journal Year: 2025, Volume and Issue: 17(2), P. 283 - 283

Published: Jan. 15, 2025

The increasing demand for food and rapid population growth have made advanced crop monitoring essential sustainable agriculture. Deep learning models leveraging multispectral satellite imagery, like Sentinel-2, provide valuable solutions. However, transferring these to diverse regions is challenging due phenological differences in stages between training target areas. This study proposes the Adaptive Month Matching (AMM) method align of crops areas enhanced transfer cropland segmentation. In AMM method, an optimal Sentinel-2 monthly time series identified area based on deep model performance major common both A month-matching process then selects by aligning this study, covered part Mississippi River Delta, while included across US Canada. evaluation focused crops, including corn, soybeans, rice, double-cropped winter wheat/soybeans. trained was transferred areas, accuracy metrics were compared different chosen various alignment methods. consistently demonstrated strong performance, particularly rice-growing regions, achieving overall 98%. It often matched or exceeded other matching techniques corn segmentation, with average all exceeding 79%

Language: Английский

Citations

0
DeepCropClustering: A deep unsupervised clustering approach by adopting nearest and farthest neighbors for crop mapping DOI
Hengbin Wang, Yuanyuan Zhao, Shaoming Li

et al.

ISPRS Journal of Photogrammetry and Remote Sensing, Journal Year: 2025, Volume and Issue: 224, P. 187 - 201

Published: April 12, 2025

Language: Английский

Citations

0
Automatic crop type mapping based on crop-wise indicative features DOI
Jieqing Yu, Longcai Zhao, Yanfu Liu

et al.

International Journal of Applied Earth Observation and Geoinformation, Journal Year: 2025, Volume and Issue: 139, P. 104554 - 104554

Published: April 27, 2025

Language: Английский

Citations

0
CropSight: Towards a large-scale operational framework for object-based crop type ground truth retrieval using street view and PlanetScope satellite imagery DOI Creative Commons
Yin Liu, Chunyuan Diao,

Weiye Mei

et al.

ISPRS Journal of Photogrammetry and Remote Sensing, Journal Year: 2024, Volume and Issue: 216, P. 66 - 89

Published: Aug. 1, 2024

Crop type maps are essential in informing agricultural policy decisions by providing crucial data on the specific crops cultivated given regions. The generation of crop usually involves collection ground truth various species, which can be challenging at large scales. As an alternative to conventional field observations, street view images offer a valuable and extensive resource for gathering large-scale through imaging roadside fields. Yet our ability systematically retrieve labels scales from operational fashion is still limited. retrieval pixel level with uncertainty seldom considered. In study, we develop novel deep learning-based CropSight modeling framework object-based synthesizing Google Street View (GSV) PlanetScope satellite images. comprises three key components: (1) A cropland field-view imagery method devised acquire representative geotagged types across regions manner; (2) UncertainFusionNet, Bayesian convolutional neural network, developed high-quality collected quantified; (3) Segmentation Anything Model (SAM) fine-tuned employed delineate boundary tailored each image its coordinate as point prompt using imagery. With four dominated US study areas, consistently shows high accuracy retrieving multiple species (overall around 97 %) delineating corresponding boundaries (F1 score 92 %). UncertainFusionNet outperforms benchmark models (i.e., ResNet-50 Vision Transformer) classification, showing improvement overall 2–8 %. SAM surpasses performance Mask-RCNN base delineation, achieving 4–12 % increase F1 score. further comparison product layer (CDL)) indicates that promising mapping products high-quality, diverse holds considerable promise extrapolate over space time operationalizing near-real-time manner.

Language: Английский

Citations

2
The dataset of main grain land changes in China over 1985–2020 DOI Creative Commons
Shidong Liu, Li Wang, Jie Zhang

et al.

Scientific Data, Journal Year: 2024, Volume and Issue: 11(1)

Published: Dec. 24, 2024

Continuous, Accurate, and detailed information on main grain land (MGL) areas is crucial for provisioning food security making policies affecting sustainable agricultural production. It still lacks a long-term MGL distribution dataset with fine spatial resolution. This study aimed to produce long-term, high-resolution map China. Here, we developed the change of resolution 30 m in China period 1985–2020 using Landsat image-based random forest algorithm GEE platform. Finally, planting intensity, gain time loss was calculated. Results indicate that our mapping results are highly consistent annual area various crops according national statistics. A validation based 3113 field survey samples 30-m showed overall accuracy were 93.57%. The full freely available at https://doi.org/10.6084/m9.figshare.26212643.v2 .

Language: Английский

Citations

2