Object Modeling from Underwater Forward-Scan Sonar Imagery with Sea-Surface Multipath

Abstract

We propose an optimization technique for 3-D underwater object modeling from 2-D forward-scan sonar images at known poses. A key contribution, for objects imaged in the proximity of the sea surface, is to resolve the multipath artifacts due to the air-water interface. Here, the object image formed by the direct target backscatter is almost always corrupted by the ghost and sometimes by the mirror components (generated by the multipath propagation). Assuming a planar air-water interface, we model, localize, and discard the corrupted object region within each view, thus avoiding the distortion of recovered 3-D shape. Additionally, complementary visual cues from the boundary of the mirror component, distinct at suitable sonar poses, are employed to enhance the 3-D modeling accuracy. The optimization is implemented as iterative shape adjustment by displacing the vertices of triangular patches in the 3-D surface mesh model, in order to minimize the discrepancy between the data and synthesized views of the 3-D object model. To this end, we first determine 2-D motion fields that align the object regions in the data and synthesized views, then calculate the 3-D motion of triangular patch centers, and finally the model vertices. The 3-D model is initialized with the solution of an earlier space carving method applied to the same data. The same parameters are applied in various experiments with 2 real data sets, mixed real-synthetic data set, and computer-generated data guided by general findings from a real experiment, to explore the impact of non-flat air-water interface. The results confirm the generation of a refined 3-D model in about half-dozen iterations.

Figures (11)

• Figure 1: (a) Sonar image of a scene in shallow pool, comprising of object, mirror and ghost components due to both surface and bottom multi path contributions. Black mesh in (b) is the initial 3-D target model to generate the synthetic sonar image (c), comprising of (e) object, (f) mirror and (g) ghost components. Optimizing the 3-D model by our method yields the red mesh in (b). (d) lower parts of object (green) and mirror (red) contours and the non-overlapping region of object and ghost components (blue) used in the optimization method.
• Figure 2: (a) A beam in $\theta$ direction covers azimuthal interval [${\theta}, {\theta}\!+\!{\delta\theta}$]. Image intensity $\!I\!$ of pixel $(x,y)$ depends on cumulative echos from unknown number of surface patches within volume $\!V_{\phi}\!$, arriving at sonar receiver simultaneously. Along this beam, $V_{\phi}$ covers intervals $[-W_{\phi},W_{\phi}]$ in elevation angle and [$\Re$, $\Re\!+{\delta\Re}]$ in range. DIDSON beam-bin data (b) and corresponding polar image (c) of a hemispherical rock on a shallow-pool bottom is corrupted by two near/far highlight bands within shadow region due to bottom and surface multipath. (d) Boundary pixels (green), and a boundary patch (red) with its center (yellow) projecting onto a boundary pixel (magenta).
• Figure 3: (a) Mirror image geometry: transmitted sound waves traveling along ${\bf R}_1$, are scattered at object point $P_s$. Reflected component, propagating along "unique direction" ${\bf R}_2$ to point $P_W$ at water surface with normal ${\bf n}$, is specularly reflected towards the sonar along ${\bf R}_3$. (b) Ghost image geometry: traveling along mirror-image pathway in reverse direction, sound waves along $-{\bf R}_3$ are specularly reflected towards the object along $-{\bf R}_2$. Scattered at $P_s$, component along $-{\bf R}_1$ is captured by the sonar. (c) Relative locations of object (green), mirror (red) and ghost (yellow) components vary as sonar rolls. Object region overlap with ghost (and possibly mirror) component(s) leads to contour distortion and intensity corruption.
• Figure 4: (a) Block diagram of entire algorithm; (b) 3-D motions of vertices and patch centers; sample example shows that each vertex ${\bf P}_k$ appear in two or more patches, and its motion ${\bf V}^*_{P_k}$ have to be consistent with motions ${\bf V}^*_{c_i}$ ($i=1,\ 2,\ \ldots,\ p$) of $p$ neighboring patch centers.
• Figure 5: (a) Beam-bin image of coral rock near water surface with mirror and ghost components; (b) binary image; (c) extracted contours; (d) contours superimposed on image (green: object and ghost regions; red: mirror).