Synergistic Use of Radar Sentinel-1 and Optical Sentinel-2 Imagery for Crop Mapping: A Case Study for Belgium

Abstract

A timely inventory of agricultural areas and crop types is an essential requirement for ensuring global food security and allowing early crop monitoring practices. Satellite remote sensing has proven to be an increasingly more reliable tool to identify crop types. With the Copernicus program and its Sentinel satellites, a growing source of satellite remote sensing data is publicly available at no charge. Here, we used joint Sentinel-1 radar and Sentinel-2 optical imagery to create a crop map for Belgium. To ensure homogenous radar and optical inputs across the country, Sentinel-1 12-day backscatter mosaics were created after incidence angle normalization, and Sentinel-2 normalized difference vegetation index (NDVI) images were smoothed to yield 10-daily cloud-free mosaics. An optimized random forest classifier predicted the eight crop types with a maximum accuracy of 82% and a kappa coefficient of 0.77. We found that a combination of radar and optical imagery always outperformed a classification based on single-sensor inputs, and that classification performance increased throughout the season until July, when differences between crop types were largest. Furthermore, we showed that the concept of classification confidence derived from the random forest classifier provided insight into the reliability of the predicted class for each pixel, clearly showing that parcel borders have a lower classification confidence. We concluded that the synergistic use of radar and optical data for crop classification led to richer information increasing classification accuracies compared to optical-only classification. Further work should focus on object-level classification and crop monitoring to exploit the rich potential of combined radar and optical observations.

Optimal Timing Assessment for Crop Separation Using Multispectral Unmanned Aerial Vehicle (UAV) Data and Textural Features

“…When considering all of the crops in the study area, the accuracy of crop separation is highest in the mid-season temporal window between 1362 and 2016 AGDD but accuracy reaches a maximum at 1556 AGDD. This is in line with other studies that report best accuracies for crop separation with data sets acquired in July [2,40,41]. At this time of the growing season, the winter crops are in a senescence stage and the summer crops in their most productive stage.…”

“…In the early temporal window, at the beginning of the growing season, maize and sugar beet are gradually sown in. Since their seedlings are difficult to distinguish [40] and can thus be mixed up with pixels from bare soil fields, the AA values are lower than in the subsequent mid-season temporal window. The slight decrease in AA values in the early temporal window is not of statistical significance.…”

“…Sugar beet is the first sown summer crop in the study area and can be accurately separated at 515 AGDD, as the spectral signal is dominated by bare soil and therefore is distinguishable from the other present crops (i.e., winter crops or grassland). At the following observation date (i.e., 635 AGDD), differentiation between maize, sugar beet and bare soil fields becomes more difficult, since the spectral signal of all three classes at this phenological stage is dominated by the signal of bare soil and indistinguishable seedlings [40].…”

Vehicle and Vessel Detection on Satellite Imagery: A Comparative Study on Single-Shot Detectors

“…Imagery intelligence (IMINT) is a discipline which collects information through aerial and satellite means, allowing the monitoring of agricultural crop growth [2], performance of border and maritime surveillance [3,4] and inference of land changes [5] for other applications. Recent advances in computer vision, using deep learning techniques, already allow successful automation of IMINT cases on aerial images [6][7][8].…”

Sub-Pixel Crop Type Classification Using PROBA-V 100 m NDVI Time Series and Reference Data from Sentinel-2 Classifications

“…Compared with other studies the accuracy achieved here is reasonable. For example, Van Tricht et al [40] applied random forest classification algorithm on monthly NDVI time series from Sentinel-2 to map 12 crops and land cover types in Belgium. An overall accuracy of 78% was obtained using an input dataset encompassing the period March-August and 72% using another dataset covering the period March-June.…”