[1] Here we demonstrate with a study of the Lucky Strike hydrothermal field that image mosaicing over large seafloor areas is feasible with new image processing techniques, and that repeated surveys allow temporal studies of active processes. Lucky Strike mosaics, generated from >56,000 images acquired in 1996, 2006, 2008 and 2009, reveal the distribution and types of diffuse outflow throughout the field, and their association with high-temperature vents. In detail, the zones of outflow are largely controlled by faults, and we suggest that the spatial clustering of active zones likely reflects the geometry of the underlying plumbing system. Imagery also provides constraints on temporal variability at two time-scales. First, based upon changes in individual outflow features identified in mosaics acquired in different years, we document a general decline of diffuse outflow throughout the vent field over time-scales up to 13 years. Second, the image mosaics reveal broad patches of seafloor that we interpret as fossil outflow zones, owing to their association with extinct chimneys and hydrothermal deposits. These areas encompass the entire region of present-day hydrothermal activity, suggesting that the plumbing system has persisted over long periods of time, loosely constrained to hundreds to thousands of years. The coupling of mosaic interpretation and available field measurements allow us to independently estimate the heat flux of the Lucky Strike system at $200 to 1000 MW, with 75% to >90% of this flux taken up by diffuse hydrothermal outflow. Based on these heat flux estimates, we propose that the temporal decline of the system at short and long time scales may be explained by the progressive cooling of the AMC, without replenishment. The results at Lucky Strike demonstrate that repeated image surveys can be routinely performed to characterize and study the temporal variability of a broad range of vent sites hosting active processes (e.g., cold seeps, hydrothermal fields, gas outflows, etc.), allowing a better understanding of fluid flow dynamics from the sub-seafloor, and a quantification of fluxes.

One of the leading causes of overfishing is the catch of unwanted fish and marine life in commercial fishing gears. Echosounders are nowadays routinely used to detect fish schools and make qualitative estimates of the amount of fish and species present. However, the problem of estimating sizes using acoustic systems is still largely unsolved, with only a few attempts at real-time operation and only at demonstration level. This paper proposes a novel image-based method for individual fish detection, targeted at drastically reducing catches of undersized fish in commercial trawling. The proposal is based on the processing of stereo images acquired by the Deep Vision imaging system, directly placed in the trawl. The images are pre-processed to correct for nonlinearities of the camera response. Then, a Mask R-CNN architecture is used to localize and segment each individual fish in the images. This segmentation is subsequently refined using local gradients to obtain an accurate estimate of the boundary of every fish. Testing was conducted with two representative datasets, containing in excess of 2600 manually annotated individual fish, and acquired using distinct artificial illumination setups. A distinctive advantage of this proposal is the ability to successfully deal with cluttered images containing overlapping fish.

No abstract

Abstract-A common problem in video surveys in very shallow waters is the presence of strong light fluctuations, due to sun light refraction. Refracted sunlight casts fast moving patterns, which can significantly degrade the quality of the acquired data.Motivated by the growing need to improve the quality of shallow water imagery, we propose a method to remove sunlight patterns in video sequences. The method exploits the fact that video sequences allow several observations of the same area of the sea floor, over time. It is based on computing the image difference between a given reference frame and the temporal median of a registered set of neighboring images. A key observation is that this difference will have two components with separable spectral content. One is related to the illumination field (lower spatial frequencies) and the other to the registration error (higher frequencies). The illumination field, recovered by lowpass filtering, is used to correct the reference image. In addition to removing the sunflickering patterns, an important advantage of the approach is the ability to preserve the sharpness in corrected image, even in the presence of registration inaccuracies.The effectiveness of the method is illustrated in image sets acquired under strong camera motion containing non-rigid benthic structures. The results testify the good performance and generality of the approach.

Abstract-The fusion of several images of the same scene into a single and larger composite is known as photo-mosaicing. Unfortunately, the seams along image boundaries are often noticeable, due to photometrical and geometrical registration inaccuracies. Image blending is the merging step in which those artifacts are minimized.Processing bottlenecks and the lack of medium-specific processing tools have restricted underwater photo-mosaics to small areas despite the hundreds of thousands of square meters that modern surveys can cover. Large underwater photo-mosaics are increasingly in demand for the characterization of the seafloor for scientific purposes. Producing these mosaics is difficult due to the challenging nature of the underwater environment and the image acquisition conditions, including extreme depth, scattering and light attenuation phenomena and difficulties in vehicle navigation and positioning. This paper proposes strategies and solutions to tackle the problems of very large underwater optical surveys (Giga-mosaics), presenting contributions in the image preprocessing, enhancing and blending steps, resulting in an improved visual quality in the final photo-mosaic. A comprehensive review of the existing methods is also presented and discussed. Our approach is validated by a large optical survey of a deep-sea hydrothermal field, leading to a high-quality composite in excess of 5 Gigapixel.

Abstract-Detecting and selecting proper landmarks is a key issue to solve Simultaneous Localization and Mapping (SLAM). In this work, we present a novel approach to perform this landmark detection. Our approach is based on using three sources of information: 1) three-dimensional topological information from SLAM; 2) context information to characterize regions of interest (RoI); and 3) features extracted from these RoIs. Topological information is taken from the SLAM algorithm, i.e. the three-dimensional approximate position of the landmark with a certain level of uncertainty. Contextual information is obtained by segmenting the image into background and RoIs. Features extracted from points of interest are then computed by using common feature extractors such as SIFT and SURF. This information is used to associate new observations with known landmarks obtained from previous observations. The proposed approach is tested under a real unstructured underwater environment using the SPARUS AUV. Results demonstrate the validity of our approach, improving map consistency.

No abstract

Projective homography sits at the heart of many problems in image registration. In addition to many methods for estimating the homography parameters [5], analytical expressions to assess the accuracy of the transformation parameters have been proposed [4]. We show that these expressions provide less accurate bounds than those based on the earlier results of Weng et al. [7]. The discrepancy becomes more critical in applications involving the integration of frame-to-frame homographies and their uncertainties, as in the reconstruction of terrain mosaics and the camera trajectory from flyover imagery. We demonstrate these issues through selected examples.