Share Email Print

Proceedings Paper

Morphological image processing for locating minelike objects from side-scan sonar images
Author(s): Gordon L. Swartzman; William C. Kooiman
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

A morphological image-processing algorithm was developed to facilitate the rapid identification of bottom minelike objects in side scan sonar acoustic backscatter images. Because large numbers of images are being processed, the emphasis is on rapid computation. The algorithm achieves computational efficiency by dividing the image into bins of a pre-chosen size and performing a binary opening operation for each bin with a 2 X 2 structuring element on each bin having sufficient pixels within the threshold range. Thresholding can be either above a high backscatter level to highlight bright proud objects or below a low backscatter level to highlight low backscatter shade-like objects. The morphological operating highlights continuous pixels within the threshold range without distortion and eliminates objects smaller than the structuring element. A connected component algorithm was used to locate all identified contiguous pixels and to tabulate their centroids and sizes. The identified objects were then screened as possible targets by checking the proximity of bright and dark objects within some threshold radius and chosen direction of each other. The chosen targets were either graphed or archived. Algorithm performance was evaluated by comparison with other target identification algorithms and was found to be compatible. An advanced interactive mode allows using different structuring elements and different morphological operations for possible improvement of the batch mode algorithm. The algorithm, while potentially effective for target identification, is primarily useful for false target identification. By operating on standard survey images the algorithm can isolate areas of potential false target proliferation. Spatial statistical methods, based on k- nearest neighbor distributions and hierarchical and k-means clustering were used to delineate regions of high false target density within the survey area.

Paper Details

Date Published: 2 August 1999
PDF: 15 pages
Proc. SPIE 3710, Detection and Remediation Technologies for Mines and Minelike Targets IV, (2 August 1999); doi: 10.1117/12.357026
Show Author Affiliations
Gordon L. Swartzman, Univ. of Washington (United States)
William C. Kooiman, Univ. of Washington (United States)

Published in SPIE Proceedings Vol. 3710:
Detection and Remediation Technologies for Mines and Minelike Targets IV
Abinash C. Dubey; James F. Harvey; J. Thomas Broach; Regina E. Dugan, Editor(s)

© SPIE. Terms of Use
Back to Top