An unsupervised mixture-tuned matched filtering-based method for the remote sensing of opium poppy fields using EO-1 Hyperion data: an example from Helmand, Afghanistan

Abstract

ABSTRACTRemote sensing has special advantages to monitor drug production that causes serious problems to global society. The widely used high spatial resolution images are too costly to make the full coverage of the opium poppy fields in a large area. Although the hyperspectral imagery acquired by Earth Observing-1 (EO-1) Hyperion that is free with medium spatial resolution has been employed, the used unsupervised multiple endmember spectral mixture analysis (MESMA)-based method is time-consuming for a large area. The present study used an unsupervised mixture-tuned matched filtering (MTMF)-based method to detect poppy fields from a Hyperion image covering a study area in Helmand, Afghanistan, and it achieved the producer’s, user’s and overall accuracies of 61%, 73% and 76%, as well as the kappa coefficient of 0.48. This method worked over 10 times faster than the MESMA-based method with similar detection accuracies. This MTMF-based method provides a potential alternative for the United Nations and the Afghanistan government to monitor opium poppy cultivation in Afghanistan.