An expectation-maximization framework for the estimation of bathymetry from side-scan sonar images

Abstract

In this paper a new procedure for the computation of seabed altitude information from side-scan sonar data is presented. Although side-scan sensors do not provide direct measures of seabed elevation, their images are directly related to seabed topography. Using a mathematical model for the sonar ensonification process, approximations to the seabed characteristics can be inferred from the sonar image. The problem is however severely under-constrained, in the sense that not all the parameters involved in the image formation process can be directly determined from the side-scan image. To overcome this difficulty, we propose the utilization of a multi-resolution expectation-maximization framework to select the most probable parameters from the solution space. At every iteration the estimated solution is used to simulate a side-scan image of the observed scene, which is then be compared to the side-scan image actually observed; solution parameters are then refined using gradient-descent optimization. The process is repeated until convergence is achieved.

Supervised target detection and classification by training on augmented reality data

“…The sonar inversion process is the means by which we are able to reconstruct these sets of parameters for a given side-scan image. A more detailed explanation of this procedure can be found in the work of Coiras et al [11].…”

“…The inverse problem, that of obtaining the scene parameters from a given real side-scan image, is much more complex and requires the utilisation of statistical optimisation techniques. Using the approach proposed by Coiras et al [11], we can solve the optimisation problem by selecting an adequate initialisation for R, Z The intensity scattered back to the sensor will depend on the incidence angle u( p) formed by the direction of illumination r and the normal N at point p…”

“…where l is a constant that controls the rate of change, and is typically set to a value of 0.25. Application of this approach using a multi-resolution implementation [11] yields an estimation of the scene parameter maps R, Z and F. An example of the results is presented in Fig. 3, which shows an original side-scan image and the three parameter images R, F and Z obtained from it.…”

“…A compromise between using real and synthetic data can be found in augmented reality (AR) simulation [10], where synthetic target models are embedded on a real image of the seafloor. A computer model of the seafloor is constructed from the side-scan image by an inversion process [11], which determines the parameters that characterise the observed scene. Then the computer model for the seafloor and that of the target are combined and rendered to obtain a new AR image that realistically integrates the synthetic target within the observed scene.…”

Target recognition in synthetic aperture and high resolution sidescan sonar

“…A compromise between using real and synthetic data can be found in the Augmented Reality (AR) simulation [13], where synthetic target models are embedded on a real image of the seafloor. A computer model of the seafloor is constructed from the sidescan image by an inversion process [14], which determines the parameters that characterize the observed scene. Then the computer model for the seafloor and that of the target are combined and rendered to obtain a new AR image that realistically integrates the synthetic target within the observed scene.…”