Page 62 - Summer 2021

ECHO CLASSIFICATION
trees (radar) that will interfere with the process of detecting the target of interest.
Beyond detection and imaging, the fluctuations of echoes between successive transmissions are commonly exploited in classification applications. The statistics of these fluctuations are based on multiple transmissions as the sensor system scans an area. Through modeling, the statistics can be classified in terms of important prop- erties of the scatterers, such as type of scatterer and its numerical density. For example, a large fluctuation in a sonar echo could be from a large fish feeding on an aggre- gation of smaller marine organisms such as zooplankton. Or a change in the degree of fluctuations in the echoes from a medical ultrasound device could represent the presence of cancerous tissue.
Key elements and applications of echo statistics are briefly summarized. A comprehensive treatment of the subject is given in the tutorial by Stanton et al. (2018). Also, although the material in this article and in the tuto- rial in 2018 was inspired by the authors’ own research in aquatic applications, the diverse examples presented in each publication demonstrate the ubiquity and broad range of usefulness of echo statistics.
The Echo Classification Process and Statistics of Echo Fluctuations
A principal method for classifying echoes involves analyzing the time it takes for an echo to return to the transmitter as well as the amplitude of the echo. The measure of time is related to the distance between the transmitter and the object that causes the echo. For exam- ple, if a ship is using sonar to help navigate in the dark, it is useful to know the distance to an iceberg. To determine the type of object that is causing the echo, the amplitude (a measure of its “strength”) of the echo is measured. The amplitude is related to the size and material properties of the object. Back to the iceberg, if there is a loud echo measured by the sonar that comes back quickly, then it may be time to steer the ship in another direction!
There are many ambiguities in using the echo amplitude to classify the object because echoes may be the same or similar for completely different scenarios. For example, the echo from a single large object may be mistaken as the echo from many small, closely spaced objects of similar
material properties. Or the echo from a large object with material properties similar to those of the surrounding medium may be confused with the echo from a small object with material properties with a strong contrast rela- tive to that of the surrounding medium.
There are a variety of methods to eliminate such ambi- guities. One method involves analyzing how the echo amplitude varies through repeated transmissions as the sensor system scans an area. For example, does the echo remain approximately the same value or does it vary by a large amount across many transmissions? Is the echo generally small most of the time but with occasional large values? Qualitatively, the variability of echoes within an acoustic image or “echogram” of an area is analogous to texture in a photograph of an object. The degree to which the image is grainy or smooth will provide information about the type of object causing the echo. For example, the shiny image of a smooth metallic spoon has a dramati- cally different texture than that of a wooden spoon of the same size. In this case, the texture of the image provides information about the material properties and smoothness of the object. The challenge for all cases lies in extracting information from the data for quantitative classification.
Physics of Scattering
To quantitatively understand and exploit these variations or texture for classification, the physics of the scattering must first be understood. The scattering of the objects can be modeled in terms of their size, shape, orientation, and material properties as well as their spatial distribu- tion. The echoes will also depend on properties of the sensor system, including the frequency, beamwidth, and duration of the transmitted pulse. The key scattering and sensor properties can be accounted for in a physics-based model of an echo. Specifically, the echo magnitude ã from an aggregation of N scatterers is given in Stanton et al. (2018) by
(1)
where the magnitude of the echo voltage from the ith scatterer as received through the sensor system is
(2)
62 Acoustics Today • Summer 2021