Spatially explicit active learning for crop-type mapping from satellite image time series

Beatrice Kaijage, Mariana Belgiu*, Wietske Bijker

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

1 Citation (Scopus)
29 Downloads (Pure)

Abstract

The availability of a sufficient number of annotated samples is one of the main challenges of the supervised methods used to classify crop types from remote sensing images. Creating these samples is time-consuming and costly. Active Learning (AL) offers a solution by streamlining sample annotation, resulting in more efficient training with less effort. Unfortunately, most of the developed AL methods overlook spatial information inherent in remote sensing images. We propose a novel spatially explicit AL that uses the semi-variogram to identify and discard redundant, spatially adjacent samples. It was evaluated using Random Forest (RF) and Sentinel-2 Satellite Image Time Series in two study areas from the Netherlands and Belgium. In the Netherlands, the spatially explicit AL selected 97 samples achieving an overall accuracy of 80%, compared to traditional AL selecting 169 samples with 82% overall accuracy. In Belgium, spatially explicit AL selected 223 samples and obtained 60% overall accuracy, while traditional AL selected 327 samples and obtained an overall accuracy of 63%. We concluded that the developed AL method helped RF achieve a good performance mostly for the classes consisting of individual crops with a relatively distinctive growth pattern such as sugar beets or cereals. Aggregated classes such as ‘fruits and nuts’ posed, however, a challenge.

Original languageEnglish
Article number2108
JournalSensors
Volume24
Issue number7
DOIs
Publication statusPublished - Apr 2024

Keywords

  • agriculture
  • remote sensing
  • scarce label environments
  • spatial autocorrelation
  • supervised classification
  • ITC-GOLD
  • ITC-ISI-JOURNAL-ARTICLE

Fingerprint

Dive into the research topics of 'Spatially explicit active learning for crop-type mapping from satellite image time series'. Together they form a unique fingerprint.

Cite this