The impact of band selection on gas detection algorithms Academic Article uri icon


  • ABSTRACT A new method is developed for evaluating band selection for detecting gases in hyperspectral images. We use a no-gas background to estimate the sample correlation matrix; we detect anomalies in the gas-present image. After separating the gas and background pixels, we then calculate the SNR. We find that increasing the number of bands tends to lower our overall performance.

publication date

  • August 13, 2010