You are here

Shoreline Extraction Algorithm

The US Navy seeks a partner to license and commercialize a simple and automated algorithm to extract shorelines from remotely sensed imagery

Image Right
 The US Navy has developed an innovative combination of water pixels (top) and water edge pixels (middle) to create accurate water boundaries (bottom) from a wide range of imagery sources
The US Navy has developed an innovative combination of water pixels (top) and water edge pixels (middle) to create accurate water boundaries (bottom) from a wide range of imagery sources
The Technology: 

The US Naval Research Laboratory detachment at Stennis Space Center has developed an algorithm and software to quickly and easily extract shoreline data from remotely sensed imagery. The Navy’s innovative approach extracts water and non-water point pixels, and employs a unique interrelation between the two in order to generate highly accurate boundaries that are properly ordered and oriented for GIS applications. The system determines water/non-water points by calculating the entropy or roughness in the image. This approach requires only a single-channel image of sufficiently high resolution and positional accuracy for the desired result—there are no a priori requirements of image format, size, color space, or sensor used.

Edge points identified at the water/land interface are then properly ordered and oriented by creating three-point segments with normal vectors pointing into the center of the river. Proper order is then calculated, resulting in open segments and closed loops for islands within the channel and lakes outside that are quite insensitive to image noise. User-defined settings fine-tune the threshold for noise tolerance and the inclusion/exclusion of smaller lakes, puddles, and islands.

Benefits: 
  • Accurate: The innovative combination of edge and water pixels determined by entropy and using those points together to properly orient each edge point typically results in a root-mean-square deviation less than twice as large as the ground sample distance of the imagery
  • Image Agnostic: Requires only a single-channel image in any georeferenced format and from any sensor source with sufficient resolution for the desired output
  • Simple and Tunable: The technique does not require pre- or post-processing but the user can specify up to three parameters to refine the sensitivity and resolution of the output
The Opportunity: 
  • Available for license:
  • Potential for collaboration with NRL Stennis Space Center researchers
Contacts: