Fast Unsupervised Seafloor Characterization in Sonar Imagery Using Lacunarity

Abstract

A new unsupervised approach for characterizing seafloor in side-looking sonar imagery is proposed. The approach is based on lacunarity, which measures the pixel-intensity variation, of through-the-sensor data. No training data are required, no assumptions regarding the statistical distributions of the pixels are made, and the universe of (discrete) seafloor types need not be enumerated or known. It is shown how lacunarity can be computed very quickly using integral-image representations, thereby making real-time seafloor assessments on-board an autonomous underwater vehicle feasible. The promise of the approach is demonstrated on high-resolution synthetic-aperture-sonar imagery of diverse seafloor conditions measured at various geographical sites. Specifically, it is shown how lacunarity can effectively distinguish different seafloor conditions and how this fact can be exploited for target-detection performance prediction in mine-countermeasure operations.

Comparison of possibilistic fuzzy local information C-means and possibilistic K-nearest neighbors for synthetic aperture sonar image segmentation

“…Before applying PFLICM and PKNN to each SAS image, feature extraction and superpixel segmentation were implemented. For each image, a 34 dimensional feature vector was computed using Sobel features 19 (eight edge orientations and square mask sizes five, nine, eleven and fifteen) and lacunarity 24 (window sizes of [31,21] and [21,11]) to aid in capturing texture information. After the features are extracted, superpixels are computed on the grayscale image.…”

Resolution dependence of rough surface scattering using a power law roughness spectrum

“…The scintillation index (also called structure (Wang and Bovik, 2002), lacunarity (Williams, 2015), or contrast (Marston and Plotnick, 2015)) was shown to be is highly dependent on all the parameters studied: resolution, grazing angle, spectral strength, and spectral exponent. For moderate to low grazing angles, SI monotonically increases as grazing angle decreases, resolution cell decreases, and spectral strength increases.…”

“…Median filters are often used to "remove" the speckle or intensity fluctuations from acoustic or electromagnetic images before use in remote sensing or target detection algorithms (e.g. (Galusha et al, 2018;Kwon and Nasrabadi, 2005;Williams, 2015Williams, , 2018 and references therein). Although the broadband scattering cross section, which uses the arithmetic mean of the intensity, is insensitive to resolution, the median is not, since the median is highly dependent on the probability density function.…”

“…The resolution dependence of the scintillation index has implications for target detection performance, synthetic aperture autofocus algorithms (e.g. (Marston and Plotnick, 2015)), and preprocessing algorithms for automated target recognition (Galusha et al, 2018;Kwon and Nasrabadi, 2005;Williams, 2015).…”

Possibilistic fuzzy local information C-means with automated feature selection for seafloor segmentation

“…Figure 1 illustrates the forward feature selection process. The suite of features used for the selection process are Sobel, 22 histogram of oriented gradients, 23 local binary patterns, 24 mean, variance, shape, 25,26 Haralick texture features (contrast,correlation,energy, and homogeneity), 27 a Gabor filter, 28 a Gaussian filter, 29 lacunarity, 30 and a Laplacian of Gaussian filter. 6,29 An…”