An integrated multi-modal data fusion framework incorporating mechanical sonar for 3D underwater boundary reconstruction

Accurate and reliable reconstruction of three-dimensional underwater environments remains a persistent challenge, particularly when employing low-cost acoustic sensors onboard compact robotic platforms. In this work, an integrated multi-modal data-fusion framework is presented, which combines a mechanical scanning sonar, an inertial measurement unit (IMU), and a depth sensor to generate high-fidelity 3D boundary representations of underwater scenes. The proposed framework incorporates motion-compensation and depth-anchoring mechanisms that mitigate the effects of slow update rates and motion distortions inherent to mechanical scanning sonars; hence, two-dimensional acoustic profiles are consistently extended into boundary-layer point clouds. The controlled experiments were conducted in a saltwater pool environment by using ground-truth anchors for quantitative validation. The fusion system achieved a root-mean-square error of 6.2 cm, a mean absolute deviation of 4.8 cm, and a standard deviation of 2.1 cm, thereby demonstrating reliable sub-decimeter accuracy. Compared with sonar-only reconstructions, which yielded a root-mean-square error of 9.4 cm, the proposed fusion approach improved accuracy by approximately 35 percent. Vertical consistency was also maintained, with average deviations across depth layers of 5.5 cm. The reconstructed maps accurately represent both linear and curved geometries, confirming the robustness and accuracy of the proposed framework. Experimental results demonstrate that this low-cost multi-sensor fusion approach is a useful and efficient method for various marine applications, such as aquaculture monitoring, underwater infrastructure inspection, and confined space surveys.