Automated detection of dust clouds, sources, and direction from NOAA-AVHRR satellite imagery
This research is mainly focused on automatically locating dust sources and estimating their transport direction in NOAA-AVHRR images. Applications of AVHRR images may include agricultural assessment, producing maps of large area, and to retrieve various geophysical parameters. The AVHRR simultaneously records five bands of data, in our research, we focused on bands four and five since these wavelengths highlight the absorption and subsequent emission of thermal radiation by the silicate particles in the dust storms. In this research, we propose a new method to detect dust clouds on satellite images. The approach we propose involves the use of an image segmentation technique known as region growing. This method starts with a set of seed points obtained from a band math image, which is known to be a good indicator of dust clouds due to physical reasons, the result will be a binary image that represents the improved dust cloud region estimate. We also propose a method to locate dust sources automatically in satellite images. Previous methods were mainly based on the user manually selecting the source points in the satellite image. The method developed here uses corner detection on the boundary of the dust cloud region. Also, we propose a new technique for finding the direction of the dust transport in the dust cloud region using individual bands from the NOAA-AVHRR imagery. Multi-resolution filters and state-of-the-art directional filters, based on the Contourlet transform, are used to help us determine the direction with more precision and consistency among the relevant subimages than in previous approach. Before applying the directional filtering to the candidate region of the multispectral image, a preprocessing step involves passing the image through a nonsubsampled pyramidal transform to apply selective amplification of high frequency information. This preprocessing step enhances the directional streaks before the directional filtering used. For AVHRR images, our methodology involves applying directional filtering on bands 4 or 5. Directional filtering is applied to blocks of the image targeting a decision among 8 or 16 different angles. From the filter outputs, energy measurements are computed to find the prominent direction of the dust storm in each block. The presence of consistent prominent directions in the texture of the blocks that constitute region of the dust storm can be used as a verification of its presence.
Alkhatib, Mohammed Qassim, "Automated detection of dust clouds, sources, and direction from NOAA-AVHRR satellite imagery" (2011). ETD Collection for University of Texas, El Paso. AAI1503700.