Finding Ocean Structures Using Mathematical Morphology
Abstract
IR satellite images are used extensively to study ocean dynamics; their synoptic coverage complements traditional oceanographic measurements. Data interpretation is complicated by cloud cover and the features' spatial and temporal variation. The opening and closing operation mathematical morphology smoothes edges and finds objects simultaneously. Adaptions needed to find Gulf Stream rings in NOAA AVHRR images and comparisons of the results with those from both human analysts and other automated techniques are shown. Processing consists of inverting the image, applying opening and closing to separate very hot objects (generally land, the Gulf Stream, and some clouds) from the rest of the image and to remove them, and applying opening and closing to the result to separate remaining hot objects from the background (usually cold water), which is also removed. In opening and closing, objects are defined by minimum size and intensity threshold criteria. The latter criterion varies locally in the image. The structure element, a pattern of pixels whose properties are those sought in the image, is a uniform intensity 3 x 3 pixel array. Opening and closing involves alternately considering whether the entire structure element fits under the image's intensity surface when the structure element is centered on a particular pixel, and whether and part of the structure element fits under the intensity surface. (mm)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1990
- Accession Number
- ADA229818
Entities
People
- Matthew Lybanon
- Suzanne M. Lea
Organizations
- United States Naval Research Laboratory