Crop mapping using supervised machine learning and deep learning: a systematic literature review

International Journal of Remote Sensing, Journal Year: 2023, Volume and Issue: 44(8), P. 2717 - 2753

Published: April 18, 2023

The ever-increasing global population presents a looming threat to food production. To meet growing demands while minimizing negative impacts on water and soil, agricultural practices must be altered. make informed decisions, decision-makers require timely, accurate, efficient crop maps. Remote sensing-based mapping faces numerous challenges. However, recent years have seen substantial advances in through the use of big data, multi-sensor imagery, democratization remote sensing success deep learning algorithms. This systematic literature review provides an overview history evolution using techniques. It also discusses latest scientific field mapping, which involve machine models. protocol involved analysis 386 peer-reviewed publications. results show that areas such as rotation double cropping, early further exploration. LiDAR tool for needs more attention, hierarchical is recommended. comprehensive framework future researchers interested accurate large-scale from multi-source image data

Language: Английский

---

Crop Mapping without Labels: Investigating Temporal and Spatial Transferability of Crop Classification Models Using a 5-Year Sentinel-2 Series and Machine Learning

Remote Sensing, Journal Year: 2023, Volume and Issue: 15(13), P. 3414 - 3414

Published: July 5, 2023

Multitemporal crop classification approaches have demonstrated high performance within a given season. However, cross-season and cross-region presents unique transferability challenge. This study addresses this challenge by adopting domain generalization approach, e.g., training models on multiple seasons to improve new, unseen target years. We utilize comprehensive five-year Sentinel-2 dataset over different agricultural regions in Slovakia diverse scheme (eight classes). evaluate the of machine learning algorithms, including random forests, support vector machines, quadratic discriminant analysis, neural networks. Our main findings reveal that across years differs between regions, with Danubian lowlands demonstrating better (overall accuracies ranging from 91.5% 2022 94.3% 2020) compared eastern 85% 91.9% 2020). Quadratic networks consistently scenarios. The forest algorithm was less reliable generalizing scenarios, particularly when there significant deviation distribution domains. finding underscores importance employing multi-classifier analysis. Rapeseed, grasslands, sugar beet show stable seasons. observe all periods play crucial role process, July being most important August least important. Acceptable can be achieved as early June, only slight improvements towards end Finally, approach allows for parcel-level confidence determination, enhancing reliability maps assuming higher classifiers yield similar results. To enhance spatiotemporal generalization, our proposes two-step approach: (1) determine optimal spatial accurately represent type distribution; (2) apply interannual capture variability helps account various factors, such rotation practices, observational quality, local climate-driven patterns, leading more accurate nationwide monitoring.

Language: Английский

---

Monitoring the Degree of Mosaic Disease in Apple Leaves Using Hyperspectral Images

Remote Sensing, Journal Year: 2023, Volume and Issue: 15(10), P. 2504 - 2504

Published: May 10, 2023

Mosaic of apple leaves is a major disease that reduces the yield and quality apples, monitoring for allows its timely control. However, few studies have investigated status pests diseases, especially mosaic using hyperspectral imaging technology. Here, images healthy infected were obtained near-ground high spectrometer anthocyanin content was measured simultaneously. The spectral differences between analyzed. in estimated by optimal model to determine degree disease. exhibited stronger reflectance at range 500–560 nm as increased. correlation processed Gaussian1 wavelet transform significantly improved compared corresponding results with original spectrum. VPs-XGBoost estimation performed best, which sufficient monitor findings provide theoretical support quantitative leaf remote sensing disease; they lay foundation large-scale sensing.

Language: Английский

---

Eucalyptus Plantation Area Extraction Based on SLPSO-RFE Feature Selection and Multi-Temporal Sentinel-1/2 Data

Forests, Journal Year: 2023, Volume and Issue: 14(9), P. 1864 - 1864

Published: Sept. 13, 2023

An accurate and efficient estimation of eucalyptus plantation areas is paramount significance for forestry resource management ecological environment monitoring. Currently, combining multidimensional optical SAR images with machine learning has become an important method classification, but there are still some challenges in feature selection. This study proposes a selection that combines multi-temporal Sentinel-1 Sentinel-2 data SLPSO (social particle swarm optimization) RFE (Recursive Feature Elimination), which reduces the impact information redundancy improves classification accuracy. Specifically, this paper first fuses data, then carries out by to mitigate effects redundancy. Next, based on features such as spectrum, red-edge indices, texture characteristics, vegetation backscatter coefficients, employs Simple Non-Iterative Clustering (SNIC) object-oriented three different types machine-learning models: Random Forest (RF), Classification Regression Trees (CART), Support Vector Machines (SVM) extraction areas. Each model uses supervised-learning method, labeled training guiding regions. Lastly, validate efficacy selecting performance SLPSO–RFE comparative analysis undertaken against results derived from single-temporal ReliefF–RFE scheme. The findings reveal employing significantly elevates precision plantations across all classifiers. overall accuracy rates were noted at 95.48% SVM, 96% CART, 97.97% RF. When contrasted outcomes ReliefF–RFE, trio models saw increase 10%, 8%, 8.54%, respectively. enhancement was even more pronounced when juxtaposed ReliefF-RFE, increments 15.25%, 13.58%, 14.54% insights research carry profound theoretical implications practical applications, particularly identifying extracting leveraging

Language: Английский

---

Assessment of multi-date Sentinel-2 data for field-level monitoring of Isabgol (Plantago ovata Forsk) cropping practices in India

Advances in Space Research, Journal Year: 2024, Volume and Issue: 74(11), P. 5305 - 5318

Published: Aug. 6, 2024

Language: Английский

---

Multi-Year Cereal Crop Classification Model in a Semi-Arid Region Using Sentinel-2 and Landsat 7–8 Data

Agriculture, Journal Year: 2023, Volume and Issue: 13(8), P. 1633 - 1633

Published: Aug. 19, 2023

This study developed a multi-year classification model for winter cereal in semi-arid region, the Kairouan area (Tunisia). A random forest was constructed using Sentinel 2 (S2) vegetation indices reference agricultural season, 2020/2021. then applied S2 and Landsat (7 8) data previous seasons from 2011 to 2022 validated field observation data. The achieved an overall accuracy (OA) of 89.3%. Using resulted higher accuracy. Cereal exhibited excellent precision ranging 85.8% 95.1% when utilizing data, while lower (41% 91.8%) obtained only slight confusion between cereals growing with olive trees observed. second objective map as early possible season. An demonstrated accurate results February (four months before harvest), 95.2% OA 87.7%. When entire period, 85.1% 94.2% (42.6% 95.4%) observed general methodology could be adopted other regions similar climates produce very useful information planner, leading reduction fieldwork.

Language: Английский

---

Staple crop mapping potential analysis and application of Sentinel-1 interferometric coherence: A case study in the Songnen Plain, China

Advances in Space Research, Journal Year: 2024, Volume and Issue: 74(7), P. 2993 - 3010

Published: June 20, 2024

Language: Английский

---

Extraction of Hani terraces based on Sentinel-2 and GF-2 images in Honghe prefecture, Yunnan province

Geocarto International, Journal Year: 2024, Volume and Issue: 39(1)

Published: Jan. 1, 2024

Monitoring Hani terraces quickly and accurately using remote sensing technology is crucial for the protecting World Cultural Heritage Sites. However, single image affected by mutual constraints of temporal spatial resolution, making it difficult to concurrently integrate key phenological information accurate extraction. In this study, GF-2 Sentinel-2 images are used extract based on objected-based analysis. Firstly, features objects were obtained multi-resolution segmentation image. Secondly, optimal optimized recursive feature elimination cross-validation separation index, respectively. Finally, all adopted random forest (RF) support vector machine (SVM) classifiers, Comparing deep learning traditional methods, proposed method RF achieved highest accuracy with Kappa coefficient overall 89.45 94.73%,

Language: Английский

---

An Overview of Within-Season Agricultural Monitoring from Remotely Sensed Data

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(24), P. 4706 - 4706

Published: Dec. 17, 2024

Remote sensing data have been widely used to monitor various agricultural activities, such as crop distribution mapping, phenology extraction, farmland soil moisture monitoring, diseases prevention, and ideotype breeding [...]

Language: Английский

---

Crop Diversification Assessment in Tank Ayacut Areas of Lower Palar Sub-Basin of Chengalpattu District, Tamil Nadu, India Using Geo-Spatial Techniques

International Journal of Environment and Climate Change, Journal Year: 2023, Volume and Issue: 13(10), P. 968 - 980

Published: Aug. 23, 2023

For the assessment of crop diversification in major tank Ayacut area Lower Palar sub-basin Chengalpattu district Tamil Nadu, research works were carried out using Sentinel 2 optical data by relating with ground truth data, to identify crops pixel-based classification and further classified Random Forest machine learning algorithms. The total estimated under was 15767.97 28818.17 ha respectively for summer seasons 2018 2021. Since, season experiences high diversification. water spread volume tanks 612.31 1177.89 6,39,248 14,06,056 m3 accuracy points confusion matrix reveals an overall 96.8% (2018) 94.9 % (2021) kappa scores 0.96 0.94 respectively. assessments Simpson Index Diversity values 0.63 0.68 accounted 2021 diversified pattern is significantly correlated availability which increased cropping as confirmed Crop Diversification factor.

Language: Английский

---