Cloud Tracking from Satellite Pictures.
Abstract
Satellite pictures taken in short time intervals are employed to estimate the lateral (x-y) and rotational displacement of clouds. Several methods of motion analysis are introduced based on different signal characterizations and various degrees of image abstraction. Using the discrete luminance information (half-tone picture), the maximum of the inner product surface of two consecutive images serves as a lateral displacement estimator. This dissertation demonstrates how the maximum peak detectability can be improved by prefiltering with a non causal planar least squares inverse filter. Although this filter is derived from theoretical considerations, the resulting high-pass frequency characteristics commends its use for another reason: since most of the cloud information, such as edges and texture, is embedded in the high frequency parts of the image, the subsequent inner product operation takes only the homogeneous displacement of these features into account. For clouds that have undergone severe shape changes, a gradual shift to less restrictive frequency characteristics is suggested. If the luminance values are fit to a bivariate normal surface, clouds or cloud fields can be traced by following the ellipses of equal density. This method retrieves lateral displacement, rotation and size changes. Another technique makes use of a closed contour of the cloud, either the boundary or any contour of constant gray-level. This contour is traced, resampled and transformed into a Fourier descriptor (FD) series.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1981
- Accession Number
- ADA111838
Entities
People
- Uwe L. Haass
Organizations
- Colorado State University