Main Page | Web Tools | Site Map
An Introduction to Fisheries Acoustics
Acoustics | DEIMOS | Ecology and Management |
Current Research Topics | Previous Research Topics
A listing of published acoustics papers and reports | Publications | Reports
The members of the FAR Lab | Current Members | Lab Alumni | Lab Events
Links | Other acoustics related sites | Press articles

Digital Resolution

small logo

When developing the original KRM computer model (written by C. S. Clay in BASIC), data for fish body and swimbladder tracings were digitized using graph paper and a set of human eyes. Subsequent modification of the computer programs included a procedure that read and digitized scanned .gif files (written by J. M. Jech in IDL). We did not know the effect of digital resolution on predicted scattering lengths of the modeled animals. The assumption that higher resolution always improved prediction needed checking.

A series of 9 radiographs from surface adapted Atlantic cod were used to examine the effects of digital resolution on predicted backscatter amplitudes. Fish bodies and swimbladders were digitized at 1 mm resolution and then resampled at coarser resolutions. We found that swimbladder volumes (relative to those calculated at 1 mm) decreased as image resolution decreased. Maximum reduction in swimbladder volume was 40% while most other predicted swimbladder volumes were reduced by 30%. The effect of image resolution on predicted backscatter amplitudes for a single fish, plotted as a function of the ratio of fish length (253 mm in this example) to acoustic wavelength (L/l), was more complex.

In the Rayleigh to resonance region (100 Hz to 5 kHz), backscatter amplitudes depended on swimbladder volume only. Therefore at low frequencies, digital imaging of fish bodies and swimbladders must characterize swimbladder volume. In the geometric scattering region (>5 kHz, L/l=30 kHz), backscatter amplitudes diverge with decreasing image resolution and increasing frequency. RSL values do not diverge at any resolution below L/l=5. As image resolution decreased, scattering curves diverged from the 1 mm resolution curve at lower L/l values (e.g. 3 mm L/l=27; 6 mm L/l=15; 10 mm L/l=9). At L/l=34 (200 kHz) RSL curves separate into two groups labeled A and B. At this time we continue to use 1 mm image resolution for input to our KRM models.

How does digitization affect KRM model predictions?

low resolution ambit image

high resolution ambit image

This image uses low resolution (10 mm body, 10 mm swimbladder) digitization in data files when estimating backscatter amplitudes with KRM models. Notice the large changes in amplitude when moving only a few degrees in the roll plane (calculated in 15 o increments). Think of digital resolution as the number of facets on a gem stone. The greater number of facets the larger and more uniform the reflective surface. Fewer facets reduce the total surface area that reflect light back to the source.

This image uses a high resolution (2 mm body, 2 mm swimbladder) digitization to estimate acoustic backscatter. Changes in predicted echo amplitude (calculated at 2 o increments) across a range of angles are smaller than those seen on the left. A higher digital resolution in the data files provides more 'facets' for the sound to reflect back to the source (i.e. transducer).

 

Relevant Publications
Jech, J.M. and J.K. Horne. 1998. Sensitivity of acoustic scattering models
to fish morphometry. In Proceedings of the 135th Meeting of the Acoustical Society of America, Seattle, 20-26 August 1998. pp. 1819-1820.

©2010 Fisheries Acoustics Research