Perpustakaan Digital ITB

Underwater topography mapping, known as bathymetry survey, is important to provide initial information on seafloor topography. Current technology requires shipborne systems to carry out such surveys. However, another alternative system must be considered in some areas where shipborne systems can not enter, such as shallow water areas, caves, archaeological sites, etc. In accessing difficult underwater technology, robots have been introduced as a teleoperation capability that extends the ability to access the ocean environment. Vision-based mapping (VbM) is one of the fundamental origins of automation in remote and autonomous spatial data acquisitions. Complexity in obtaining accurate data arises when such a method is applied in the underwater environment. This research proposes an integrated backscatter removal and refraction calibration model within the VbM and navigation pipeline. It is argued that the proposed VbMdedicated models can significantly improve the conformity of objects’ underwater positions around the camera's motion during underwater deployment. The research uses automatic refraction adjustment with a correction map to undistort images in real-time, addressing issues including refraction distortion and non-uniform light exposure that arise in underwater environments. Furthermore, haze removal, contrast adjustment, and color correction are combined to achieve backscatter removal for image enhancement algorithms. The research is structured into three phases to evaluate the synthesized pipeline: simulation, fieldwork, and accuracy assessment. The methodology commences by employing a simulated dataset from underwater simulation tools and introducing varying turbidity levels from 0% to 90% turbidity to assess the image enhancement algorithm. As the simulation does not directly model refraction due to the presence of water and camera lens medium, the simulated underwater camera operates with predefined parameters for both in-air and underwater simulations. The subsequent fieldwork is tailored with GoPro 10 hardware, which features a 109-degree wideangle lens recording images at a high resolution of 1980 x 1020, in Pramuka Island Waters, Indonesia. This setup offers real-world context for the research's relevance to distinct underwater circumstances and aims to examine the calibration impact on the lens and refraction distortion. The accuracy assessment involves comparing discrepancies among the VbM algorithms running offline and online (real-time process). The visual-inertial dataset collected from in-air and controlled environments, named the EASI dataset, is also utilized. The EASI dataset shares similar hardware specifications to analyze the performance and robustness of the derived camera parameters. Hence, it is helpful for testing the visual-inertial VbM for both in-air and underwater environments. The research shows efficient backscatter removal improves feature detection robustness, especially in murky water conditions. Refraction correction also eliminates the bowing effect from missing ground control points in underwater environments. The research is significant because it emphasizes how vital image enhancement and refraction calibration are to obtaining centimeter-level map accuracy of underwater VbM. The results highlight the need for a comprehensive strategy to advance underwater mapping and navigation technology to deliver accurate and dependable outcomes in various underwater situations.