Spectral distortion-based flood detection in multi-temporal images fusion techniques

In changes detection process, the choice of information extraction method plays an important role in the quality of final changes detecting. In this study, Landsat 8 multi-temporal data fusion method based on spectral distortion was used to detect changes and to determine the range of floods. For this reason, both pre and post flood images were fused using the Gram Schmitt algorithm to increase spatial resolution of images. In the following, three algorithms, Gram Schmitt, IHS, PCA, were used to detect changes and determine the extent of flood. In this study, input of each algorithm was pre-flooded as a multicolor image and post-flood infrared image as a panchromatic image selected to determine the extent of flood using the spectral distortion generated in each algorithm. The results showed that the capability of data fusion method based on spectral distortion is very high in detecting of changes. The spectral distortion generated in IHS is the most accurate distortion and the output of this algorithm is highly consistent with the reference data. Also, the output of the Gram Schmitt algorithm has spectral distortions in the unchanged regions. The PCA algorithm, which is highly sensitive to inputs, distorts most image regions, which is not recommended for detecting changes based on spectral distortion.