Multi-year vector dynamic time warping-based crop mapping

Recent automated crop mapping via supervised learning-based methods have demonstrated unprecedented improvement over classical techniques. Classification accuracies of these degrade considerably in cross-year mapping. Cross-year is more useful as it allows the prediction following years’ maps using previously labeled data. We propose vector dynamic time warping (VDTW), an innovative multi-year classification approach based on angular distances between phenological vectors. The results prove that proposed VDTW method robust to temporal and spectral variations compensating for different farming practices, climate atmospheric effects, measurement errors years. also describe a determining most discriminative window high with limited carried out tests Landsat 8 time-series imagery from years 2013 2015 corn cotton Harran Plain southeastern Turkey. In addition, we tested soybean Kansas, US 2017 2018 Harmonized Sentinel improved overall by 3% fewer training samples compared other state-of-the-art approaches.