No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
In-situ performance assessment for high frequency seabed imaging sonar
Abstract
High frequency sonar is an important and useful tool for acoustic imaging of the seabed in various applications. There are essentially three different types: sectorscanning sonar
which produces a 2D image per transmitted pulse; sidescan sonar (SSS) which operates
in a sidelooking geometry on a moving platform, where multiple pulse-echoes are treated
as independent range lines to form an image of the seabed; and synthetic aperture sonar
(SAS) which operates as a SSS with the additional coherent combination of pings into
imaging pixels. All these three operational modes are subject to loss in image quality;
e.g. due to environmental variations, refractive effects, multipath contamination, vehicle
instabilities, navigation errors, ambient noise, and other noise. When running stand-off type operations from autonomous underwater vehicles (AUV), the sonar operational conditions may vary rapidly. It then becomes important to assess the sensor performance automatically. Recently, the Norwegian Defence Research Establishment
(FFI) has developed MCM Insite, a tool for Mine Countermeasures (MCM) performance
assessment for sidelooking sonar for AUVs, based on in-situ measurements and
a-priori knowledge. The main parameters in the tool are the image quality and the image
complexity, both gathered automatically from the sonar images. In this paper we consider techniques to assess image quality for traditional (non-interferometric) SSS. We suggest that the quality should be divided into parameters that can be assessed directly from the sonar images (through-the-sensor) and accessible meta-data from the platform, without a-priori knowledge other than sensor-centric information. We have developed a coherence based technique, and image texture based techniques. In addition, we have developed a simple model for prediction of quality regions. We evaluate our techniques on data collected by AUVs, in varying conditions, both on 100 kHz and 410 kHz SSS data. We find that the techniques produce reasonable results, and think the approach may be used for automated assessment of image quality for SSS.
Synthetic aperture sonar (SAS) allows for high-resolution imagery of the seafloor with a quality that approaches that of an optical camera. Identifying seabed types within SAS images can prove very useful in a setting where autonomous underwater vehicles (AUVs) are used, since information about the bottom type can be used to determine the behavior of the AUV. In this article, we present a new algorithm for the automatic detection of different seabed types based on SAS image texture. Using a Gaussian process classification (GPC) model, we were able to predict the seabed type, as well as an uncertainty estimate on this prediction. The GPC-based model is able to learn the most informative textures from a training data set, and by construction is robust against outliers or data points that are very different from those in the training set. An additional benefit of the model is that it can be easily configured to hit the desired tradeoff between computational complexity and accuracy.
A technology shift currently takes place within mine hunting, as the search for seafloor mines to an increasing extent is performed by autonomous underwater vehicles (AUV) equipped with high-resolution, side-looking sonar. The search effectiveness for these systems can be quantified by the probability of correctly detecting and classifying mines (Pdc), together with the associated false alarm rate. These performance parameters depend on the full sensor system (including data analysis), mine threat and environment. The influence of the environment is difficult to predict reliably, as the environment may vary rapidly in space and time, and affects the performance through many factors: seafloor, bathymetry, sea state, sea currents, clutter density, etc.
This paper presents a performance model based on a combination of a priori and in-situ parameters as illustrated in the figure. The model is developed for mine search with state-of-the-art synthetic aperture sonar (SAS) systems such as HISAS 1030, but can also be adapted to other side-looking sonar systems (real or synthetic aperture). The two main in-situ parameters are estimated locally from the actual sonar data. Firstly, the image quality estimates how well a potential mine on the seafloor would be represented in the sonar image, taking into account degradations caused by interference noise, multi-path, non-linear sensor track, occlusions, etc. Secondly, the image complexity estimates to what degree Pdc is degraded by intensity variations that resemble or may interfere with mine responses. The prior parameters consist of the mine threat, as target size and shape greatly impact the expected Pdc, and the analysis system, which can be either a human operator or automatic target recognition system. The third prior parameter is the fraction of undetectable mines due to burial within sediments or vegetation. The table contains calibrated performance values for all combinations of image quality, complexity, mine threat and the analysis system, generated either from mine hunting operations with ground truth or simulations.
The performance model has been implemented as a prototype C++ tool, which will be evaluated on varied mine hunting data sets by FFI and others. The paper presents example results from sea trials with HUGIN AUV.
The interpretation of sonar images, and more generally of remote-sensing images, has traditionally been performed visually by trained interpreters. This presents the distinct advantage of using the skill of the interpreter to limits which are often unattainable by computers. But there are also many disadvantages to a purely visual interpretation. First of all, it is subjective: two interpreters with different experience, or different skills, are likely to get different interpretations for some features, depending on their experience of the sonar used or of the environment studied. Visual interpretation is also time-consuming, and a longer amount of time spent on analysis does not ensure higher objectivity. Structural geologists all know that some morphologic trends will be highlighted, unwillingly and unconsciously, when the time spent on interpretation is too long. The other important disadvantage of visual interpretation is that it is qualitative. Objects are outlined, trends and patterns are shown. But their quantitative assessment requires either the interpretation to take place directly on a numeric support (with all the associated problems of screen size and limited range of scales available) or to scan and quantify the interpretation made on physical supports (paper maps, photographs, etc.). The last decade has seen many new useful tools for visualization of data; for example, in 3-D with the ubiquitous Fledermaus software (http://www.ivs3d.com/) and the use of immersive environments (based on virtual reality advances) like the ones used in seismic prospecting (e.g., Lin et al., 2000; Shell et al., 2006). Because of their cost and the necessary investment in hardware and software, these systems are still not within the reach of all sonar users, and they will still not replace the need for a full computer-assisted interpretation.
This paper proposes a new performance metric for interferometric side-looking sonar. With this metric the probability of target detection and classification is reported as a function of coherence instead of across-track range, which is traditionally used. The coherence values are estimated from interferometric sonar processing, and can be converted into equivalent signal to noise ratios. The metric ensures that system performance is estimated based on sensor data quality of the actual mission, and as local values instead of large area averages. The method is demonstrated on synthetic aperture sonar data from HISAS 1030, showing increased consistency between shallow and deep water results, compared to the traditional metric.
In this survey we review the image processing literature on the various approaches and models investigators have used for texture. These include statistical approaches of autocorrelation functions, optical transforms, digital transforms, textural edgeness, structural element, gray tone co-occurrence, run lengths, and autoregressive models. We discuss and generalize some structural approaches to texture based on more complex primitives than gray tone. We conclude with some structural-statistical generalizations which apply the statistical techniques to the structural primitives. -Author
Synthetic aperture sonar (SAS) is a signal processing technique which increases the along-track resolution in sonar images by coherent combination of collected pings. Successful SAS processing is dependent on accurate navigation, knowledge about the ocean environment, and accurate bathymetric maps. In general, the quality of SAS images will vary dependent on these factors. We propose a model for automated prediction and assessment of SAS performance or image quality. The intended use is to give an estimate of the local quality of data in a particular SAS image. The model consists of modules for generalized signal to noise, geometric resolution, radiometric resolution (value accuracy) and image sharpness. In this paper, specifically focus on image sharpness. We show example data sets collected by an interferometric SAS on an autonomous underwater vehicle.
A new adaptive strategy for performing data collection with a sonar-equipped autonomous underwater vehicle (AUV) is proposed. The approach is general in the sense that it is applicable to a wide range of underwater tasks that rely on subsequent processing of side-looking sonar imagery. By intelligently allocating resources and immediately reacting to the data collected in-mission, the proposed approach simultaneously maximizes the information content in the data and decreases overall survey time. These improvements are achieved by adapting the AUV route to prevent portions of the mission area from being either characterized by poor image quality or obscured by shadows caused by sand ripples. The peak correlation of consecutive sonar returns is used as a measure for image quality. To detect the presence of and estimate the orientation of sand ripples, a new innovative algorithm is developed. The components of the overall data-driven path-planning algorithm are purposely constructed to permit fast real-time execution with only minimal AUV onboard processing capabilities. Experimental results based on real sonar data collected at sea are used to demonstrate the promise of the proposed approach.
A classical theorem of statistical optics, the van Cittert-Zernike theorem, is generalized to pulse echo ultrasound. This theorem fully describes the second-order statistics of the spatial fluctuations (the spatial covariance) of the field produced by an incoherent source. As a random scattering medium is insonified, it behaves as an incoherent source. The van Cittert-Zernike theorem can thus predict the spatial covariance of the pressure field backscattered by a random medium. It is shown that this spatial covariance and the incident energy diagram are Fourier pairs. In the case of a focused illumination, the spatial covariance of the backscattered pressure field is proportional to the autocorrelation of the transmitting aperture function. This is independent of frequency and of F/ number. Experimental results obtained with a linear array are in good agreement with theoretical expectations. The implications of this theorem in speckle reduction and in focusing in nonhomogenous media are discussed.
As the exploration of the seafloor is extended ever further,
automated classification and interpretation of sonar images are becoming
increasingly important. However, many of the image processing techniques
employed for these purposes consider the images simply as a collection
of textures and neglect the process by which these images are formed.
The most important aspect thus neglected is the directionality of the
imaging process and its subsequent effect on the image textures
produced. The paper reports the results of a systematic investigation of
the effect of the sonar process on image texture directionality,
illustrating the importance of the relative orientation of the sonar and
the seabed features during the sonar image capture process. A frequency
domain model is developed for examining and quantifying these
orientation effects, and the subsequent effect on a classification
system is demonstrated
Sound reflections: Advanced Applications of Side Scan Sonar
- J P Fish
- H A Carr
J. P. Fish and H. A. Carr, Sound reflections: Advanced Applications of Side Scan
Sonar. LowerCape Publishing, 2001.
APL-UW High-Frequency Ocean Environmental Acoustics Models Handbook
- Apl-Uw
APL-UW, "APL-UW High-Frequency Ocean Environmental Acoustics Models Handbook," Applied Physics Laboratory, University of Washington, Seattle, WA, USA, Tech.
Rep., 1994.
Assessment of shallow water performance using interferometric sonar coherence
- S A Synnes
- R E Hansen
- T O Saebø
S. A. Synnes, R. E. Hansen, and T. O. Saebø, "Assessment of shallow water performance using interferometric sonar coherence," in Proceedings of Underwater Acoustic
Measurements 2009, Nafplion, Greece, June 2009.
HISAS 1030 -High resolution interferometric synthetic aperture sonar
- T G Fossum
- T O Saebø
- B Langli
- H J Callow
- R E Hansen
T. G. Fossum, T. O. Saebø, B. Langli, H. J. Callow, and R. E. Hansen, "HISAS 1030
-High resolution interferometric synthetic aperture sonar," in chc/nsc08, Victoria, BC,
Canada, May 2008.