Vis enkel innførsel

dc.contributor.advisorSorensen, Asgeir
dc.contributor.advisorHugel, Vincent
dc.contributor.advisorWeingertner, Philippe
dc.contributor.authorHirtzig, Noémie
dc.date.accessioned2023-10-10T17:19:28Z
dc.date.available2023-10-10T17:19:28Z
dc.date.issued2023
dc.identifierno.ntnu:inspera:140295966:127411795
dc.identifier.urihttps://hdl.handle.net/11250/3095579
dc.description.abstract
dc.description.abstractOur research has focused on exploring the integration of FLS and SAS data for target relocation, contributing to the advancement of underwater exploration and survey missions for both industrial and research applications. Through our background research, we identified a gap in the literature concerning the combination of two different sonar modalities for object relocation. Motivated by this observation, we embarked on investigating the possibilities of combining data from distinct sonar sources to enhance object relocation capabilities. Our approach successfully extracted, synchronised, and combined various crucial data parameters, such as attitude, latitude, longitude, altitude, position in the Cartesian world frame, FLS range, aperture, gain, and frequency. This process provided valuable insights into the floor coverage area of the FLS during the mission and created functions that can be applied in future similar data processing tasks. Our efforts to explore and test various feature detectors for object matching in SAS and FLS images encountered challenges in achieving robust and consistent detections. As a result, we utilised the detection framework provided by the company to extract objects of interest from the data. Nevertheless, obtaining similar results proved to be challenging, leaving the association of data from the detections an open problem. To address this association issue in the future, obtaining specifically floor-oriented FLS images may yield more promising results, enabling effective data association and fusion across different sonar modalities. Our proposed approaches, involving the use of the Hungarian Algorithm and exploring the potential of Deep Learning methods, offer potential solutions for addressing the object relocation challenge. Ultimately, our objective is to assist in the relocation of ROVs, which may lack the high-performing navigation equipment of AUVs. By achieving reliable detections and establishing markers as "anchors", akin to SLAM, we can enhance the localisation capabilities of ROVs and correct potential drifts in the navigation systems of AUVs, leading to improved localisation schemes, especially through real-time processing of FLS data.
dc.languageeng
dc.publisherNTNU
dc.titleMulti-Sensors Relocation
dc.typeMaster thesis


Tilhørende fil(er)

Thumbnail

Denne innførselen finnes i følgende samling(er)

Vis enkel innførsel