Monitoring agriculture areas with satellite images and deep learning

Nguyên, Thành Tâm; Hoang, Thanh Dat; Pham, Minh Quang; Vu, Tuyet Trinh; Nguyen, Thanh Hung; Huynh, Quyet-Thang; Jo, Jun

doi:10.1016/j.asoc.2020.106565

Cited by 97 publications

(35 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…However, fully understanding sense of these valuable images and turning them into practice guidance for agricultural production is a challenging topic within the agricultural community. Some researchers have shown great interest in introducing stateof-the-art deep learning into satellite-based farmland mapping, crop classification, yield prediction, and so on [88][89][90].…”

Section: Land Scale: Land Cover Mappingmentioning

confidence: 99%

A Review of Deep Learning in Multiscale Agricultural Sensing

et al. 2022

View full text Add to dashboard Cite

Population growth, climate change, and the worldwide COVID-19 pandemic are imposing increasing pressure on global agricultural production. The challenge of increasing crop yield while ensuring sustainable development of environmentally friendly agriculture is a common issue throughout the world. Autonomous systems, sensing technologies, and artificial intelligence offer great opportunities to tackle this issue. In precision agriculture (PA), non-destructive and non-invasive remote and proximal sensing methods have been widely used to observe crops in visible and invisible spectra. Nowadays, the integration of high-performance imagery sensors (e.g., RGB, multispectral, hyperspectral, thermal, and SAR) and unmanned mobile platforms (e.g., satellites, UAVs, and terrestrial agricultural robots) are yielding a huge number of high-resolution farmland images, in which rich crop information is compressed. However, this has been accompanied by challenges, i.e., ways to swiftly and efficiently making full use of these images, and then, to perform fine crop management based on information-supported decision making. In the past few years, deep learning (DL) has shown great potential to reshape many industries because of its powerful capabilities of feature learning from massive datasets, and the agriculture industry is no exception. More and more agricultural scientists are paying attention to applications of deep learning in image-based farmland observations, such as land mapping, crop classification, biotic/abiotic stress monitoring, and yield prediction. To provide an update on these studies, we conducted a comprehensive investigation with a special emphasis on deep learning in multiscale agricultural remote and proximal sensing. Specifically, the applications of convolutional neural network-based supervised learning (CNN-SL), transfer learning (TL), and few-shot learning (FSL) in crop sensing at land, field, canopy, and leaf scales are the focus of this review. We hope that this work can act as a reference for the global agricultural community regarding DL in PA and can inspire deeper and broader research to promote the evolution of modern agriculture.

show abstract

Section: Land Scale: Land Cover Mappingmentioning

confidence: 99%

A Review of Deep Learning in Multiscale Agricultural Sensing

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Random walk methods generate random walks starting from the network nodes, and then learn the embeddings for the nodes so that these embeddings can capture the co-occurrences of nodes in the walks [22], [23], [40]. Deep learning methods leverage neural architectures [41], [42] such as graph neural networks [43] and autoencoders to incorporate the node features and an inductive capability in the same model [44]- [46]. To the best of our knowledge, ours is the first study to propose a multi-context graph embedding method for LBSNs in particular and hypergraphs in general.…”

Section: A Graph Embeddingmentioning

confidence: 99%

Learning Holistic Interactions in LBSNs with High-order, Dynamic, and Multi-role Contexts

Huynh

Tong

Nguyên

et al. 2022

IEEE Trans. Knowl. Data Eng.

Self Cite

View full text Add to dashboard Cite

Location-based social networks (LBSNs) have emerged over the past few years. Their exponential network effects depend on the fact that each user can share her daily digital footprints with different communities, in different places, and at different times (for example in the form of check-in activities). Unlike other types of social networks, activities in an LBSN can potentially be performed by several users in a collaborative way. Existing studies of representation learning for LBSNs often consider them as regular graphs and ignore these high-order, dynamic, and multi-role contexts, since their holistic interactions are quite difficult to capture. In this paper, we propose a model in which these holistic interactions can be learned and transferred into node embeddings derived from a hypergraph representation and a persona decomposition process. More specifically, the model learns from friendship edges, check-in hyperedges, and node personas at the same time, and devises multiple presentations for each user that reflects their multiple roles in a social context. The embedding learning process also exploits useful patterns such as user colocation and sequential effects through a carefully designed point-of-interest splitting step. Extensive experiments on real and synthetic datasets show that our model outperforms alternative state-of-the-art embedding methods on friendship and location prediction tasks by an average margin of 45.7% and 29.46%, respectively. We also demonstrate the robustness of our model against adversarial conditions such as structural noise, attribute noise, and hyperparameter sensitivity.

show abstract

“…The problem of EFA detection is a typical land cover classification one that is expected to be faced when using satellite data [43]. The above-mentioned requirements from the ARPEA drove the authors to exclude adoption of machine/deep learning-based (MDL) approaches [44,45] and opt for a simpleR rule-based approach, where defined rules could be immediately agronomically interpreted. Moving towards MDL would make it difficult to interpret eventual classification criticalities and, consequently, to properly adjust the algorithm parameters that, in MDL, cannot be related to tangible factors close to the knowledge domain to which the ideal user of the system would belong.…”

Section: Goalsmentioning

confidence: 99%

Mapping Ecological Focus Areas within the EU CAP Controls Framework by Copernicus Sentinel-2 Data

2022

View full text Add to dashboard Cite

Greening is a Common Agricultural Policy (CAP) subsidy that ensures that all EU farmers receiving income support produce climate and environmental benefits as part of their farming activities. To receive greening support, it is mandatory for the farmer to carry out three agricultural practices that are considered environmentally and climate friendly: (a) crop diversification; (b) maintenance of permanent meadows and pastures; and (c) presence of an Ecological Focus Area (EFA). Contributions are delivered and monitored by paying agencies (PP) that ordinarily perform administrative checks and spot checks. The latter are provided through photo-interpretation of high-resolution satellite or aerial images and, in specific cases, through local ground checks (GC) as well. In this work, stimulated by the Piemonte Regional Agency for Payments in Agriculture (ARPEA), a prototype service to support PPs’ controls within the greening CAP framework was proposed with special concern for EFA detection. The proposed approach is expected to represent a valid alternative or supporting tool for GC. It relies on the analysis of NDVI time series derived from Copernicus Sentinel-2 data. The study was conducted in the provinces of Turin, Asti and Vercelli within the Piedmont Region (NW Italy), and over 12,500 EFA fields were assessed. Since the recent National Report No. 5465 stipulates that mowing and any other soil management operation is prohibited on set-aside land designated as an EFA during the reference period (RP) between 1st March and 30th June, a time series (TS) of NDVI in the same period was generated. Once averaged at plot level, NDVI trends were modelled by a first-order polynomial, and the correspondent statistics (namely, R2, MAE and maximum residual) was computed. These were assumed to play the role of discriminants in EFA detection based on a thresholding approach (Otsu’s method), calibrated with reference to the training dataset. The threshold satisfaction was therefore tested, and, depending on the number of satisfied thresholds out of the possible three, EFA and non-EFA plots were detected with a different degree of reliability. The correspondent EFA map was generated for the area of interest and validated according to GCs as provided by the ARPEA. The results showed an overall accuracy of 84%, indicating that the approach is promising. The authors retain that this procedure represents a valid alternative (or integrating) tool for ground controls by PPs.

show abstract

Monitoring agriculture areas with satellite images and deep learning

Cited by 97 publications

References 42 publications

A Review of Deep Learning in Multiscale Agricultural Sensing

A Review of Deep Learning in Multiscale Agricultural Sensing

Learning Holistic Interactions in LBSNs with High-order, Dynamic, and Multi-role Contexts

Mapping Ecological Focus Areas within the EU CAP Controls Framework by Copernicus Sentinel-2 Data

Contact Info

Product

Resources

About