Ndoi, A Novel Oil Spectral Index: Comparisons and Results

Abstract

Spectral indices provide valuable information about oil spill characteristics, making them an essential tool to achieve effective action plans. In this work, the authors present new features that make NDOI a robust and suitable index for detecting coastal spills. Its main characteristics are that it obviates suspended sand in coastal regions and makes use of simple operations using VNIR spectral bands available in commercial sensors. The index is compared with spectral indices from the state-of-the-art and a radar image for the Deepwater Horizon accident occurred on 2010 in Mexico. An assessment is made based on the confusion matrix resulting from the k-nearest neighbours (kNN) classification, as well as the accuracy and the Fl-score metrics. The results consolidate NDOI in spill monitoring due to its balanced trade-off among its performance and its convenient implementation.