A new adaptive band weighting technique for hydrocarbon detection

In this study, a new approach is presented for hydrocarbon detection using hyperspectral data. The algorithm is developed based on an adaptive band weighting (ABW) technique which utilizes information in different spectral bands of the hyperspectral data to enhance the detection of desired oil signatures while suppressed the unwanted background. A constant false alarm rate (CFAR) detector is then used to obtain detected hydrocarbon under a constant false alarm rate. The algorithm has been tested using an AVIRIS hyperspectral data. A comparison study is also carried out between ABW algorithm and the Mixture Tuned Matched Filtering (MTMF) algorithm in a small sub-scene of the AVIRIS data based on the Receiver Operating Characteristic (ROC) curves from the 10 regions of interest. The presented algorithm has a higher probability of hydrocarbon detection and lower false alarm than that of MTMF results.