Segment Anything Model Combined with Multi-Scale Segmentation for Extracting Complex Cultivated Land Parcels in High-Resolution Remote Sensing Images

Remote Sensing, Год журнала: 2024, Номер 16(18), С. 3489 - 3489

Опубликована: Сен. 20, 2024

Accurate cultivated land parcel data are an essential analytical unit for further agricultural monitoring, yield estimation, and precision agriculture management. However, the high degree of landscape fragmentation irregular shapes parcels, influenced by topography human activities, limit effectiveness extraction. The visual semantic segmentation model based on Segment Anything Model (SAM) provides opportunities extracting multi-form parcels from high-resolution images; however, performance SAM in requires exploration. To address difficulty obtaining extraction that closely matches true boundaries complex large-area this study used patches with boundary information obtained unsupervised as constraints, which were then incorporated into subsequent multi-scale segmentation. A combined method was proposed, it evaluated different scenarios. In plain areas, precision, recall, IoU improved 6.57%, 10.28%, 9.82%, respectively, compared to basic extraction, confirming proposed method. comparison point-prompt conditional segmentation, achieved considerable improvements parcels. This confirms that, under zero-shot conditions, demonstrates strong cross-region cross-data source transferability across large areas.

Язык: Английский

---

Improving crop type mapping by integrating LSTM with temporal random masking and pixel-set spatial information

ISPRS Journal of Photogrammetry and Remote Sensing, Год журнала: 2024, Номер 218, С. 87 - 101

Опубликована: Окт. 19, 2024

Язык: Английский

---

Temporal segmentation method for 30-meter long-term mapping of abandoned and reclaimed croplands in Inner Mongolia, China

International Journal of Applied Earth Observation and Geoinformation, Год журнала: 2025, Номер 136, С. 104399 - 104399

Опубликована: Фев. 1, 2025

Язык: Английский

---

Recent Advances in Deep Learning-Based Spatiotemporal Fusion Methods for Remote Sensing Images

Sensors, Год журнала: 2025, Номер 25(4), С. 1093 - 1093

Опубликована: Фев. 12, 2025

Remote sensing images captured by satellites play a critical role in Earth observation (EO). With the advancement of satellite technology, number and variety remote have increased, which provide abundant data for precise environmental monitoring effective resource management. However, existing imagery often faces trade-off between spatial temporal resolutions. It is challenging single to simultaneously capture with high Consequently, spatiotemporal fusion techniques, integrate from different sensors, garnered significant attention. Over past decade, research on has achieved remarkable progress. Nevertheless, traditional methods encounter difficulties when dealing complicated scenarios. development computer science, deep learning models, such as convolutional neural networks (CNNs), generative adversarial (GANs), Transformers, diffusion recently been introduced into field fusion, resulting efficient accurate algorithms. These algorithms exhibit various strengths limitations, require further analysis comparison. Therefore, this paper reviews literature learning-based methods, analyzes compares algorithms, summarizes current challenges field, proposes possible directions future studies.

Язык: Английский

---

FieldSeg: A scalable agricultural field extraction framework based on the Segment Anything Model and 10-m Sentinel-2 imagery

Computers and Electronics in Agriculture, Год журнала: 2025, Номер 232, С. 110086 - 110086

Опубликована: Фев. 15, 2025

Язык: Английский

---

Cropping and Transformation Features of Non-Grain Cropland in Mainland China and Policy Implications

Land, Год журнала: 2025, Номер 14(3), С. 561 - 561

Опубликована: Март 7, 2025

The decrease in grain plantation areas poses a growing concern for global food security. China, with its large population, increasingly diversified demands, and relatively small cultivated lands, has suffered deeply from this phenomenon (non-grain production, NGP) recent years. Since 2020, the central government of China claimed to deal problem by attracting agriculturalists organizations involved plantation. In context, understanding NGP national situation is vital policy making. Remote sensing regarded as most effective accurate method purpose, but existing studies have mainly focused on algorithms operating at local scale or exploring grain-producing capability perspective agricultural space. As such, characterization remains deficient. study, we tried bridge gap through spatio-analysis newly published nationwide crop pattern land use geo-datasets; quantitative, spatial, structural features, well utilization cropland year 2019, were observed. results showed that about 60% was used non-grain About 15% parcels grains least three times past 4 years, these 40% double- single-season plantation, respectively, which could result 16–22% increase grain-sown area compared 2019. Forest grassland dominant non-cropping categories transferred into, indicating more time economic cost regaining grains. also presented spatio-heterogeneity regarding cropping intensity transformation. Parcels double-season mostly emerged northern, central, southern provinces, while those always located northeastern western provinces. into forest appeared Inner Mongolia, northern continued cropping. According results, propose remediation policies focusing raising provinces due their advantages water, heat, terrain, change features. Future work warranted based study’s deficiencies uncertainties. forerunner, study provides holistic observation mainland scale, findings can inform improvements concerning production security China.

Язык: Английский

---

A novel architecture for automated delineation of the agricultural fields using partial training data in remote sensing images

Computers and Electronics in Agriculture, Год журнала: 2025, Номер 234, С. 110265 - 110265

Опубликована: Март 18, 2025

Язык: Английский

---

Identification of Interannual Variation Frequency of Cropland Cropping Intensity Based on Remote Sensing Spatiotemporal Fusion and Crop Phenological Rhythm: A Case Study of Zhenjiang, Jiangsu

Agriculture, Год журнала: 2025, Номер 15(9), С. 1004 - 1004

Опубликована: Май 6, 2025

The scientific evaluation of cropland resource utilization efficiency is crucial for ensuring food security and promoting sustainable agricultural development. At present, the research on resources primarily focuses multiple cropping index intensity, but these data are insufficient to reveal long-term trends potential future changes in crop production. To fill this knowledge gap, study took Zhenjiang City, Jiangsu Province, as a case proposed method determine distribution spatiotemporal change frequency single- double-season patterns using fusion phenological rhythm. By combining Sentinel-2 NDVI MOD13Q1 satellite data, dataset with 10 m resolution was developed show interannual three area. accuracy revealed that intensity exhibited good verification accuracy, an average overall Kappa coefficient 81.53% 0.68, respectively. This provides essential support government agencies assess production develop policies improving use efficiency.

Язык: Английский

---

Extraction of Cropland Based on Multi-Source Remote Sensing and an Improved Version of the Deep Learning-Based Segment Anything Model (SAM)

Agronomy, Год журнала: 2025, Номер 15(5), С. 1139 - 1139

Опубликована: Май 6, 2025

Fine extraction of cropland parcels is an essential prerequisite for achieving precision agriculture. Remote sensing technology, due to its large-scale and multi-dimensional characteristics, can effectively enhance the efficiency collecting information on agricultural land parcels. Currently, semantic segmentation models based high-resolution remote imagery utilize limited spectral rely heavily a large amount fine data annotation, while pixel classification medium-to-low-resolution multi-temporal are by mixed problem. To address this, study utilizes GF-2 Sentinel-2 data, in conjunction with basic image model SAM, additionally introducing prompt generation module (Box Auto module) achieve automatic The research results indicate following: (1) mIoU SAM Box 0.711, OA 0.831, showing better performance, 0.679, 0.81, yielding higher-quality masks; (2) combination various prompts (box, point, mask), along hierarchical strategy, improve performance SAM; (3) Employing more accurate source significantly boost performance. superior-performing increased 0.920, raised 0.958. Overall, improved reducing demand mask annotation training, high-precision

Язык: Английский

---

Uncovering the Location of Photovoltaic Power Plants Using Heterogeneous Remote Sensing Imagery

Energy and AI, Год журнала: 2025, Номер unknown, С. 100527 - 100527

Опубликована: Май 1, 2025

Язык: Английский

---