Samples and data were collected to the project on the cruise for the Arctic Mid-Ocean Ridge (AMOR) in 2016. Using ROV both geological and biological material were provided for analysis – and three sites were mapped using the AUV Hugin. The scientific focus for the work package is methods for data interpretation and efficient data collection for Mid-Ocean ridge exploration with emphasis on features connected with mineral deposits. This challenge has been addressed on multiple spatial ranges. Starting with large area mapping, a method has been developed to enable AUVs to do data driven and adaptive seabed mapping by online data processing and automatic mission re-planning. As the acoustic signal strength differs between substrates, and the signature can be used to classify different seabed types using Hidden Markov Random Fields (HMRF) to perform unsupervised segmentation. The outcome of
this analysis was then directly used to plan and conduct an autonomous near-seabed camera survey to verify the classification results. Methods are also developed to prove supervised seabed classification of synthetic aperture sonar (SAS) data. Seabed classification is proposed by a convolutional neural network (CNN) and tested on seabed partly covered with stony corals. For seabed survey of higher detail and more limited extent – hyperspectral imaging has been explored and methods developed based on AUVs and ROVs. This is the first time a full scale hyperspectral imager has been mounted on AUVs and shows that it is feasible to deploy underwater hyperspectral imagers (UHI) on autonomous underwater vehicles with sufficient lighting. For the cruise an ROV-based core drill system was developed and tested. The system is a single stroke drill and is simple, robust and possible to use from various larger ROVs. The system provides a probing tool for marine mineral exploration, with low cost, low operational complexity and low risk. To increase the value of hyperspectral imaging a method to provide reference spectra of typical seabed material from hydrothermal vent areas were developed by analyses from a lab experiment. The identification of possible end-members from field data requires prior information in the form of representative signatures for distinct materials. Hyperspectral imaging was applied to a selection of materials from the Loki Castle active hydrothermal vent site in a laboratory setting. A method for compensating for systematic effects and recovering the reflectance spectra were presented and applied to recover the spectral signatures from the samples.
The Ecotone UHI mounted on Hugin. The AUV has a roll angle of 90 degrees in the image (underside visible). The lamps are the four smaller squares inside the white rectangle. The camera is mounted in the first circular opening to the right of the lamps, indicated by the white circle. (Sture et al 2017)