Underwater Sonar Fusion for Enhanced 3D Reconstruction using Background Data and Voxel Initializer Mapping for AUV
2
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This study proposes an underwater 3D object reconstruction method using forward looking sonar (FLS) and profilingsonar (PS) fusion through voxel-based space-carving. Underwater environments commonly rely on sonar sensors because of opticalequipment limitations. However, the inherent dimensional ambiguity of conventional sonar systems prevents accurate object localization,necessitating sensor fusion approaches to achieve precise 3D reconstruction. The proposed methodology employs a complementaryfusion strategy: FLS assigns probabilistic weights to all potential object regions considering vertical ambiguity, whereasPS complements this by executing conservative carving operations to remove regions where objects are confirmed absent. Thisapproach enables comprehensive background and object mapping followed by object segmentation, unlike conventional region-ofinterest-based reconstruction. Validation using ray-tracing simulation with objects (slope box) demonstrated that PS effectively carvesaway ambiguous vertical regions initially predicted by FLS, resulting in significantly improved 3D reconstruction accuracy. Theresults confirm that the proposed sensor fusion approach successfully resolves dimensional ambiguity inherent to single-sensor methods,achieving precise object localization through systematic probabilistic space-carving in underwater environments.
Abstract
This study proposes an underwater 3D object reconstruction method using forward looking sonar (FLS) and profilingsonar (PS) fusion through voxel-based space-carving. Underwater environments commonly rely on sonar sensors because of opticalequipment limitations. However, the inherent dimensional ambiguity of conventional sonar systems prevents accurate object localization,necessitating sensor fusion approaches to achieve precise 3D reconstruction. The proposed methodology employs a complementaryfusion strategy: FLS assigns probabilistic weights to all potential object regions considering vertical ambiguity, whereasPS complements this by executing conservative carving operations to remove regions where objects are confirmed absent. Thisapproach enables comprehensive background and object mapping followed by object segmentation, unlike conventional region-ofinterest-based reconstruction. Validation using ray-tracing simulation with objects (slope box) demonstrated that PS effectively carvesaway ambiguous vertical regions initially predicted by FLS, resulting in significantly improved 3D reconstruction accuracy. Theresults confirm that the proposed sensor fusion approach successfully resolves dimensional ambiguity inherent to single-sensor methods,achieving precise object localization through systematic probabilistic space-carving in underwater environments.