In areas of ecological interest, the detection and control of seaweed such as Posidonia Oceanica is usually performed by divers. Due to the limited capacity of the scuba tanks and the human security protocols, this task involves several short immersions leading to poor temporal and spatial data resolution. Thus, it is desirable to automate this task by means of underwater robots. This paper describes a method to autonomously detect Posidonia Oceanica in the imagery gathered by an underwater robot. The proposed approach uses a set of Gabor filters to characterize an image. This characterization is used to detect the regions containing seaweed by means of a Support Vector Machine. The experiments, conducted with an Autonomous Underwater Robot in several marine areas of Mallorca, show promising results towards the automated seafloor classification from extended video sequences.

This paper reports on a novel technique to visually detect loop closings in feature-poor underwater environments in order to increase the accuracy of vision-based localization systems. The main problem of the classical visual Simultaneous Localization and Mapping (SLAM) for underwater vehicles is the lack of robust, stable and matchable features in certain aquatic environments. The presence of sandbanks, seagrass or other underwater phenomena cause the visual features to concentrate in regions heavily textured, leaving great image areas completely free of visual information. In this situation, the classical loop closing detection algorithms fail, resulting in no corrections for the SLAM system. Our novel method proposes to reinforce the loop closing detection by clustering visual keypoints present in multiple keyframes and to match features of clusters instead of features of keyframes. This new technique is assessed on the particular application of navigating an Autonomous Underwater Vehicle (AUV) in marine environments colonized with seagrass or with the presence of sandbanks. Experiments conducted in several coastal zones on the Balearic Islands show a high degree of success in the visual registration of overlapping areas.

Effectiveness in loop closing detection is crucial to increase accuracy in SLAM (Simultaneous Localization and Mapping) for mobile robots. The most representative approaches to visual loop closing detection are based on feature matching or BOW (Bag of Words), being slow and needing a lot of memory resources or a previously defined vocabulary, which complicates and delays the whole process. This paper present a new visual LSH (Locality Sensitive Hashing)-based approach for loop closure detection, where images are hashed to accelerate considerably the whole comparison process. The algorithm is applied in AUV (Autonomous Underwater Vehicles), in several aquatic scenarios, showing promising results and the validity of this proposal to be applied online.

This paper proposes three techniques to reduce the computational cost, both in terms of memory and CPU usage, of visual underwater trajectory-based SLAM. On the one hand, geometric constraints involving the camera Field of View (FOV) are used to decide when a new node has to be added to the trajectory estimate. On the other hand, the camera FOV geometry is also used to preselect the candidate images that have to be registered. Finally, the trajectory-based structure is exploited to foresee loop closures and concentrate the computational efforts to these situations, reducing the CPU work when possible. As a result of these three techniques, the resolution of the estimated trajectory is adjusted dynamically and the image registration process, which is usually the most expensive, is only executed with images that are likely to provide useful information.

This paper presents a new method to remove the effects of uneven ensonification in Side-Scan Sonar (SSS) acoustic images, facilitating their use in underwater robotics. A Lambertian model is used for the sonar scattering process, relating the received echo with the ensonification intensity, the sea bed reflectivity and the incidence angle. As for the ensonification intensity, a sensitivity pattern model is adapted to SSS sensors. By means of these models, SSS imagery is significantly improved.

A one-shot sensor for underwater 3D reconstruction is presented and tested underwater in a water tank. The system is composed of a RGB CCD camera and a 532 nm green laser with a Diffractive Optical Element attached to it. The laser projects a pattern of parallel lines into the scene. The deformed pattern obtained in the camera frame is then processed to obtain a non-dense 3D point cloud that can be later used for autonomous manipulation and grasping, or for detailed mapping of textureless objects or scenarios.

While commercially available autonomous underwater vehicles (AUVs) are routinely used in survey missions, a new set of applications exist which clearly demand intervention capabilities: the maintenance of permanent underwater structures as well as the recovery of benthic stations or black-boxes are a few of them. These tasks are addressed nowadays using manned submersibles or work-class remotely operated vehicles (ROVs), equipped with teleoperated arms under human supervision. In the context of the TRITON Spanish funded project, a subsea panel docking and an intervention procedure are proposed. The light-weight intervention AUV (I-AUV) Girona 500 is used to autonomously dock into a subsea panel using a funnel-based docking method for passive accommodation. Once docked, an autonomous fixed-based manipulation system, which uses feedback from a digital camera, is used to turn a valve and plug/unplug a connector. The paper presents the techniques used for the autonomous docking and manipulation as well as how the adapted subsea panel has been designed to facilitate such operations.

This paper proposes a novel approach to perform underwater Simultaneous Localization and Mapping (SLAM) using a Mechanically Scanned Imaging Sonar (MSIS). This approach starts by processing the MSIS data in order to obtain range scans while taking into account the robot motion. Then, the relative motions between consecutively gathered scans are stored in the state vector. Thus, the whole sequence of robot motions between gathered scans is used to perform SLAM using an Extended Kalman Filter (EKF). One of the novelties is that this sequence is not represented with respect to a world-fixed coordinate frame, but with respect to a coordinate frame locked to the robot. Thanks to this, EKF linearization errors are reduced. The experimental results in underwater environments validate the proposal comparing the new robocentric approach to the world-centric trajectory method.

This paper proposes a mobile robot localization approach based on a monocular vision system. The proposal operates in three main steps. First, image features are extracted from frames gathered at different positions. Second, the obtained features are classified as obstacles or ground points. Third, under the assumption of a flat floor, the ground point coordinates are computed and used to perform localization. The experimental results, performed both in simulation and real environments, validate the proposal.

Likelihood fields (LF) have been used in the past to perform localization. These approaches infer the LF from range data. However, an underwater Mechanically Scanned Imaging Sonar (MSIS) does not provide distances to the closest obstacles but echo intensity profiles. In this case, obtaining ranges involves processing the acoustic data. The proposal in this paper avoids the range extraction to build the LF. Instead of processing the acoustic images to obtain ranges and then using these ranges to infer a LF, this paper proposes the use of the acoustic image itself as a good approximation of the LF. The experimental results show the potential benefits of using this idea to define a measurement model to perform mobile robot localization.

Nearly all advanced mobile robotic tasks require some knowledge of the robot location in the environment. Autonomous Underwater Vehicles are usually endowed with acoustic devices to perform localization, because underwater scenarios pose important limitations to light based sensors. One of these acoustic devices is the Mechanically Scanned Imaging Sonar (MSIS). This sensor scans the environment by emitting ultrasonic pulses and it provides echo intensity profiles (beams) of the scanned area. Our goal is to obtain range scans instead of the beams as they are provided by the MSIS. Accordingly, the proposal of this paper is to process the acoustic images in order to compute accurate distances from the sensor to the relevant obstacles in the beam. These range scans are suitable to be used in scan matching, SLAM or other approaches to estimate the robot pose.

This paper proposes a framework to perform Simultaneous Localization and Mapping (SLAM) using the scans gathered by a Mechanically Scanned Imaging Sonar (MSIS). To this end, the acoustic profiles provided by the MSIS are processed to obtain range data. Also, dead reckoning is used to compensate the robot motion during the sonar mechanical scanning and build range scans. When a new scan is constructed, its estimated position with respect to the previously gathered one is used to augment the SLAM state vector. Also, each new scan is matched against the previously detected ones by means of scan matching techniques. As the state vector contains relative positions between consecutively gathered scans, the measurement update explicitly takes into account the robot trajectory involved in each loop closure.

This paper presents an approach to perform Simultaneous Localization and Mapping (SLAM) in underwater environments using a Mechanically Scanned Imaging Sonar (MSIS) not relying on the existence of features in the environment. The proposal has to deal with the particularities of the MSIS in order to obtain range scans while correcting the motion induced distortions. The SLAM algorithm manages the relative poses between the gathered scans, thus making it possible to correct the whole Autonomous Underwater Vehicle (AUV) trajectories involved in the loop closures. Additionally, the loop closures can be delayed if needed. The experiments are based on real data obtained by an AUV endowed with an MSIS, a Doppler Velocity Log (DVL) and a Motion Reference Unit (MRU). Also, GPS data is available as a ground truth. The results show the quality of our approach by comparing it to GPS and to other previously existing algorithms.

In this paper, a new method to perform scan matching using sparse sets of noisy readings is presented and tested with sonar sensors. This method, the LF/SoG, does not rely on the establishment of correspondences. The underlying idea is to model one of the scans by a function of x-y-coordinates, the Likelihood Field, depicting the likelihood of obstacle detection. Then, the scan matching problem turns into an optimization problem involving the Likelihood Field and the other scan. Because no correspondences are established, the main problems of ICP-based algorithms are not present in the LF/SoG. The presented experimental results validate the LF/SoG approach and compare it with other well known scan matching algorithms in terms of robustness, accuracy and convergence rate.

This paper presents a probabilistic framework to perform scan matching localization using standard time-of-flight ultrasonic sensors. Probabilistic models of the sensors as well as techniques to propagate the errors through the models are also presented and discussed. A method to estimate the most probable trajectory followed by the robot according to the scan matching and odometry estimations is also presented. Thanks to that, accurate robot localization can be performed without the need of geometric constraints. The experiments demonstrate the robustness of our method even in the presence of large amounts of noisy readings and odometric errors.

This paper describes a probabilistic sonar Altering technique dealing with sonar cross-talking and multiple reflections. The filtered sonar data is then used to feed the probabilistic scan matching algorithm spIC, which is also described. This scan matching algorithm deals with sonar range and angular errors. Experimental results demonstrate the utility of the combined use of probabilistic filtering and scan matching by evaluating the robustness, accuracy and computational cost. Also, the results obtained with spIC are compared with those provided by well known scan matching algorithms such as ICP and IDC.

Frequently, autonomous underwater vehicles (AUVs) must operate in turbulent ocean environments with complex spatio-temporal variability. Energetic flows, instabilities and currents can strongly perturb safety conditions and development of AUVs operations. A priori knowledge of this ocean variability would allow to adequately plan AUV missions, minimizing the possible negative effects of the environment on its operation. Unfortunately, present oceanographic technology is only capable to forecast large and slow components of the ocean variability, being fast and small scales of variability still unpredictable. This partial knowledge about future environmental conditions degrades AUV robustness, reducing its autonomy and safety. In this article, we propose a class of real-time algorithm to support AUVs navigation in strong turbulent ocean environments characterized by unpredictable small scale variability. The algorithm infers the two-dimensional structure of the current field in a limited region in front of the AUV, using the one-dimensional information obtained from an horizontal acoustic doppler current profiler (H-ADCP). Then, a planner locally optimizes the performance of the AUV in the inferred field. In this work reducing travelling time was of interest. The proposed procedure has been tested on simulated turbulent environments. Results indicate that, although sub-optimals, solutions provided by the algorithm required significantly less travelling time than straight line paths. An example where the proposed methodology increases AUV safety and autonomy is also provided

Autonomous Underwater Vehicles (AUVs) operate in ocean environments characterized by complex spatial variability which can jeopardize their missions. To avoid this, planning safety routes with minimum energy cost is of primary importance. This work explores the benefits, in terms of energy cost, of path planning in marine environments showing certain spatial variability. Extensive computations have been carried out to calculate, by means of an A* search procedure, optimal paths on ocean environments with different length scale of eddies and different current intensities. To get statistical confidence, different realizations of the eddy field and starting-ending points of the path have been considered for each environment. Unlike previous works, the more realistic and applied case of constant thrust power navigation is considered. Results indicate that substantial energy savings of planned paths compared to straight line trajectories are obtained when the current intensity of the eddy structures and the vehicle speed are comparable. Conversely, the straight line path between starting and ending points can be considered an optimum path when the current speed does not exceed half of the vehicle velocity. In both situations, benefits of path planning seem dependent of the spatial structure of the eddy field.

The work presented in this paper deals with scan matching localization using ultrasonic range sensors. Our contribution resides in the extension of ICP based algorithms to be used with ultrasonic sensor data. This extension consists of a pre-process step, where ultrasonic sensor readings are grouped to overcome their sparseness, and a post-process step, where the whole robot trajectory involved in the grouping process is corrected. Thanks to that a great improvement with respect to odometry is obtained. Experimental results show that even huge odometric errors are corrected with the presented method.

For a mobile robot to execute useful missions it has to be endowed with navigation ability. To navigate successfully, information about the environment and the robot position is needed. This article presents a robust system for detecting, identifying, analyzing, and comparing environmental landmarks obtained from standard range sensors and using them to generate and incrementally improve a hybrid topological-metric map of the environment. This map is used in turn to correct the estimation of the robot position. The control architecture that integrates this mapping and localization system is described, as well as the experimental results obtained on both a simulated robot and a real one. © 2005 Wiley Periodicals, Inc. Int J Int Syst 20: 499–521, 2005.

The irradiance measurement performed by vision cameras is not noise-free due to both processing errors during CCD fabrication and the behaviour of the electronic device itself. A proper characterization of sensor performance, however, allows either removing the resulting noise from the image or accounting for it within image processing algorithms. This paper proposes new methods for estimating the distribution parameters of two of such sources of noise: dark current and the so-called fixed pattern noise. Since both methods require knowledge about the scene illumination, an estimation method using a calibrating sphere is also presented. This method models illumination as the combination of directional and ambient lighting. Experimental results can be found at the end of the paper.

Most shape from shading (SFS) algorithms assume that objects in the scene are Lambertian reflectors, i.e., they do not absorb any radiation for any wavelength. However, real objects reflect only part of the incoming radiation, so that these algorithms usually perform poorly on images with non-uniform reflectivity. In this paper an adaptation for such algorithms is proposed. The strategy mainly consists of guiding the SFS algorithm through the image in order not to violate the typical albedo uniformity assumption. To do this, the pixels in the image are iteratively grouped into regions coming from the same scene until there are no more adjacent pixels satisfying this condition. The decision about the aggregation of a new pixel to a region is based on estimating its albedo and comparing it with the albedo of the region. During this process, SFS is applied only to the considered region and its results are used in the albedo estimation procedure. Some experimental results with synthetic and real images are discussed.

Nowadays, the surveillance and inspection of underwater installations, such as power and telecommunication cables and pipelines, is carried out by operators that, being on the surface, drive a Remotely Operated Vehicle (ROV) with cameras mounted over it. This is a tedious and high time-consuming task, easily prone to errors mainly because of loss of attention or fatigue of the human operator. Besides, the complexity of the task is increased by the lack of quality of typical seabed images. In this study, the development of a vision system guiding an Autonomous Underwater Vehicle (AUV) able to detect and track automatically an underwater power cable laid on the seabed has been the main concern. The vision system that is proposed tracks the cable with an average success rate above 90%. The system has been tested using sequences coming from a video tape obtained in several tracking sessions of various real cables with a ROV driven from the surface.

The interest in fuzzy algorithms is increasing in a wide range of pattern recognition applications. This paper describes a recognition system using fuzzy algorithms and parameters, with particular enhancement in the system architecture. This system is able to recognize flat polygonal objects in real time. The main aim of the system is to perform an object recognition based on contour description and being independent of noise and indeterminations in the digitization process. This system is made up of three modules. The first one is a break points detector. The second block performs a contour description by means of fuzzy parameters that allows us to recognize the picture in the third block.>