Localized Radon transform tracks ship-wake patterns

Because a ship moving through a body of water normally creates a characteristic v-shaped pattern of waves, its wake components usually appear as linear features in synthetic-aperture-radar tracking images that may appear brighter or darker than the background. Often, the Radon transform is used for finding such features in images because it is robust in the presence of noise. In operation, the transform accentuates the linear features by integrating image intensity along all possible lines in an

Localized Radon transform tracks ship-wake patterns

Because a ship moving through a body of water normally creates a characteristic v-shaped pattern of waves, its wake components usually appear as linear features in synthetic-aperture-radar tracking images that may appear brighter or darker than the background. Often, the Radon transform is used for finding such features in images because it is robust in the presence of noise. In operation, the transform accentuates the linear features by integrating image intensity along all possible lines in an image. But, because it is a global transform, it cannot discriminate between long and short lines.

Because of this, Anthony Copeland of the Computer Vision and Robotics Research Laboratory at the University of California, San Diego, has developed a localized version of the transform that can automatically monitor nautical activity by tracking ship-wake patterns. "Because the analysis of features takes place within the localized Radon-transform domain," says Copeland, "the algorithm is a feature space-line detector."

If the wake components are dark and linear, they appear in the transform domain as a narrow dark region bordered above and below by brighter regions. "By performing variance calculations along short segments in the vertical direction, sharp transitions are highlighted," says Copeland. This image is then thresholded to isolate the strongest responses. A morphology process is then performed on the feature space to eliminate false responses and to isolate the linear wake features. After dilation and erosion, the features are mapped back to image space to highlight the linear wake features of the synthetic-aperture-radar image.

To automate shipping surveillance systems, further processing is required to classify the detected linear features as either ocean waves or the effects of a ship`s wake. Linear wake components also need to be analyzed to approximate the position and heading of the ship that caused them. "To accomplish this," says Copeland, "at least two linear wake features must be identified for each ship."

More in Non-Factory