Simple Methodology for Deriving Continuous Shorelines from Imagery: Application to Rivers

Abstract

A methodology is developed to extract and process shoreline data, the interface between land and water, identified from imagery. Initially, image pixels containing water (water points) and pixel locations of the land/water interface (edge points) are extracted from an image using either a supervised, threshold approach or a newly developed, automated texture-based analysis. Both are described and demonstrated. Subsequently applied is a procedure for processing these edge and water point locations to obtain oriented and ordered shoreline coordinates. The described methodology has several advantages: (1) shoreline processing is independent of imagery source and resolution, that is, specification of search directions based on image resolution or desired shoreline resolution is unnecessary and (2) a need for additional postprocessing of remote-sensed data or extracted-edge data are obviated, that is, edge data need not be of high quality or vectorized. Details of the entire methodology, including algorithms for water and edge point extraction from imagery and specifics of the processing applied to obtain an ordered, oriented shoreline, are presented. Execution of the complete shoreline extraction algorithm is demonstrated independently by application to sections of the East Pearl River, Mississippi, and the Kootenai River, Idaho. A quantitative performance measure of the ability of the algorithm to produce realistic-ordered, oriented shoreline coordinates from a set of edge and water point data extracted from imagery results in RMS errors of 1 m or less or two times the ground sample distance of the imagery.