Improvement in Land Cover and Crop Classification based on Temporal Features Learning from Sentinel-2 Data Using Recurrent-Convolutional Neural Network (R-CNN)

Abstract

Understanding the use of current land cover, along with monitoring change over time, is vital for agronomists and agricultural agencies responsible for land management. The increasing spatial and temporal resolution of globally available satellite images, such as provided by Sentinel-2, creates new possibilities for researchers to use freely available multi-spectral optical images, with decametric spatial resolution and more frequent revisits for remote sensing applications such as land cover and crop classification (LC&CC), agricultural monitoring and management, environment monitoring. Existing solutions dedicated to cropland mapping can be categorized based on per-pixel based and object-based. However, it is still challenging when more classes of agricultural crops are considered at a massive scale. In this paper, a novel and optimal deep learning model for pixel-based LC&CC is developed and implemented based on Recurrent Neural Networks (RNN) in combination with Convolutional Neural Networks (CNN) using multi-temporal sentinel-2 imagery of central north part of Italy, which has diverse agricultural system dominated by economic crop types. The proposed methodology is capable of automated feature extraction by learning time correlation of multiple images, which reduces manual feature engineering and modeling crop phenological stages. Fifteen classes, including major agricultural crops, were considered in this study. We also tested other widely used traditional machine learning algorithms for comparison such as support vector machine SVM, random forest (RF), Kernal SVM, and gradient boosting machine, also called XGBoost. The overall accuracy achieved by our proposed Pixel R-CNN was 96.5%, which showed considerable improvements in comparison with existing mainstream methods. This study showed that Pixel R-CNN based model offers a highly accurate way to assess and employ time-series data for multi-temporal classification tasks.

Semantically Enriched Crop Type Classification and Linked Earth Observation Data to Support the Common Agricultural Policy Monitoring

“…Their effectiveness stems from their ability to accurately describe the nonlinear relationships between crops' physical condition and their spectral characteristics, while being particularly insensitive to noise and overfitting. Finally, there are important studies that have used Convolutional Neural Networks (CNN) or Recurrent Neural Networks (RNN) or a combination of both [16], which allow the learning of time and space correlation over the Sentinel time-series, thus reducing manual feature engineering.…”

“…It is considered as one of the best triplestores available in terms of storage, supported functionalities, performance and execution time [38]. Other RDF triple stores that provide geospatial support include 10 https://www.sqlshack.com/understanding-benefits-of-graph-databasesover-relational-databases-through-self-joins-in-sql-server/ 11 http://graphdb.ontotext.com/ RDF4J 12 , Virtuoso 13 [39], OntopSpatial 14 [40], Oracle spatial and Graph 15 , AllegroGraph 16 , Stardog 17 , uSeekM 18 and Parliament 19 . The storage and query capabilities of our framework capitalise on an existing RDF triple store, on top of which SPARQL and GeoSPARQL standards are used to form the queries that support the rules of agriculture policies.…”

Fusion of Remote Sensing and Applied Geophysics for Sinkholes Identification in Tabular Middle Atlas of Morocco (the Causse of El Hajeb): Impact on the Protection of Water Resource

“…Sentinel-2 images are frequently applied in many fields, particularly for water management and land protection purposes [43], as well as the management of agricultural plots [44], the identification…”

“…The data acquired by Sentinel satellites constellation is managed by the European Commission and distributed by the European Space Agency (ESA) via the Copernicus Open Access Hub. The images can be accessed and downloaded for free at https://scihub.copernicus.eu/dhus/#/home for regions located within 84 • N and 56 • S. Sentinel-2 images are frequently applied in many fields, particularly for water management and land protection purposes [43], as well as the management of agricultural plots [44], the identification and monitoring of crops [45] and for the risk management of floods [46], forest fires [47], and coastal water contaminant [48]. Additional applications include geological mapping [49] and the monitoring of geological waste [50].…”

Spatially Enhanced Spectral Unmixing Through Data Fusion of Spectral and Visible Images from Different Sensors

“…Here we use an NN for this purpose due to its little need for prior assumptions. Furthermore, NNs were found in many cases to outperform standard machine learning procedures in solving remote sensing tasks, as shown, for example, in [34]. Unlike traditional methods, we fuse data sources without full spatial, spectral, and temporal overlapping between the two data types.…”