Image segmentation by unsupervised adaptive clustering in the distribution space for AUV guidance along sea-bed boundaries using vision

Tenas, A.; Rendas, Maria João; Folcher, Jean-Pierre

doi:10.1109/oceans.2001.968778

Cited by 10 publications

(7 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We assume that the two regions induce distinctive signatures on both the video images and on the profiles returned by the sonar. In previous work, we have already demonstrated the feasibility of discriminating different types of sea bottom for these two sensing modalities [2], [3]. Assuming that both sensors are able to acquire data incoming from both regions 'The ROV Phantom SOOXP.…”

Section: Introduction and Problem Formulationmentioning

confidence: 99%

Exploiting natural contours for automatic sonar-to-video calibration

Barat¹,

Rendas²

2005

Europe Oceans 2005

Self Cite

View full text Add to dashboard Cite

A&stract-The paper addresses the problem of co-registration of data acquired by two distinct ego-centric exteroceptive sensors mounted in an autonomous underwater vehicle: a video camera and a mechanically scanning profiler sonar. As it is the case for the http://www.laas.fr/reforme-CNRShtmI experimental piatform used in this study', these two perception sensors can often be installed in flexible geometric configurations on the robot's crash frame, and thus their relative orientation and displacement may change from mission to mission, even if during each mission they are kept fixed. The algorithm exploits the presence in the sea bottom of a boundary between two regions that can be discriminated both in the video frames and on the sonar profiles, and requires no specially tuned artificial settings. The registration is solved in a completely autonomous manner, using the ability of the robot to track the natural contour based on the video data. Our algorithm h d s the mapping from the sonar coordinates to the image coordinates that minimizes the average entropy of a probabilitic model that is fit to the inverse mapping of the segmented images. It does not require reconstruction of the robot's trajectory during data acquisition, and is thus insensitive to positioning problems.

show abstract

Section: Introduction and Problem Formulationmentioning

confidence: 99%

Exploiting natural contours for automatic sonar-to-video calibration

Barat¹,

Rendas²

2005

Europe Oceans 2005

Self Cite

View full text Add to dashboard Cite

show abstract

“…The solution proposed here improves upon the work presented in [3] where contour estimation involves a complete segmentation of each video frame. The major goal when designing the previous algorithm was to keep the analysis free of (I priori assumptions about the visual appearance of the observed sea bed regions.…”

Section: O D U C I 1 O Nmentioning

confidence: 99%

“…In [3] tracking is enforced by closing a classical controller acting on the vehicle's beading rate (of PID type) around an error signal equal to the distance of the outgoing point with respect to the the middle of the upper edge of the image. To be able to iteratively locate in a stable manner the wtgoing point in the current frame we must estimate the entire contour that is observed inside each frame.…”

Section: Problemformirationmentioning

confidence: 99%

“…(3)) involve an integration of the gradient of the potential over the contour. We approximate this continuous integral by a summation over a set of discrete sample points along the contour.…”

Section: A Regisnationmentioning

confidence: 99%

“…However, as [he image shows, there are significant variations of the background along the sequence.Figure 3(right) illushates the initialisation of the snake based on a coarse segmentation of the image[3].…”

mentioning

confidence: 99%

See 2 more Smart Citations

Statistical snakes: robust tracking of benthic contours under varying background

Rolfes

Rendas

2004 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (IEEE Cat. No.04CH37566)

Self Cite

View full text Add to dashboard Cite

Absfmcl-Thir paper presemta the lormdhm 01 rhtisHrnl snakes lor contour estimation, and applles It to the problem of trseklng a nsturnl benthic cootom using am A W equipped with a video mmem To close the controi Imp that malahins the vehicle on the contour, we need to estimate Its lorstion on each lmme of the video sequence. We propose a new criterion for cootour estimation which Is approprhte to the complexity and variability of natural environments. The criterion is bared c m a non-parametric 6t1tisticd modeliog of the regions adjacent to the tracked boundary, s mixtures of the probabllity distributions corresponding to the areas on each side of the contour. It Is shown on the paper that minimizing the proposed criterion leads to an estimated contour such th.t the regions on each side have mixture cormelrats close to zero and one (meanlug that they are "pure" regions). Examples 01 appllmtion 01 the proposed algorithm to P r e d underwater image sequence are gkm. Keywords-Conlour ~ocking, Viion, Staisflcal s w k s 1. m O D U c I 1 O NThe ability to back benthic contours -that delineate regions of the sea bed occupied by a given type of material -is not only an efficient observation behaviour in the context of biological studies (analysis of the patchy structure induced in the sea-bed), but also a necessary capability for mapping large scale geometric feahlres of the sea-bed, which provides the most useful and reliable information for robot localisation purposes. This paper addresses backing of benrhic contours using visual infomation.Autonomous contour tracking using vision involves hvo problems: (a) to estimate the relative position of the robot with respect to the contour, (8) to generate appropriate control actions that steer the vehicle such that the contour is kept inside the field of vision of the onboard video camera. This paper focus on problem (a), proposing a fast robust algorithm for contour tracking accross video jiames.The solution proposed here improves upon the work presented in [3] where contour estimation involves a complete segmentation of each video frame. The major goal when designing the previous algorithm was to keep the analysis free of (I priori assumptions about the visual appearance of the observed sea bed regions. The algorithm is grounded on fundamental results from probabilistic Information Theory, making no assumptions about the characteristics of the probability dishibutious associated to the distinct regions that may be present in the images. The solution presented here solves the hvo main drawback associated to our previous work (1) its large computational cost, by focusing image processing in the vicinity of the hacked contour; (2) the sensitivity with respect to background variations, by explicitly adapting to the local characteristics of the sea bed. These gains are obtained by resorting the formalism of defomable contours (snakes) [I], and using the basic measures of Information Theory already involved in the previous solution to defme the energy functional that these deforma...

show abstract

Statistical snakes: Application to tracking of benthic contours

Rolfes¹,

Rendas²

2004

IFAC Proceedings Volumes

View full text Add to dashboard Cite

Image segmentation by unsupervised adaptive clustering in the distribution space for AUV guidance along sea-bed boundaries using vision

Cited by 10 publications

References 2 publications

Exploiting natural contours for automatic sonar-to-video calibration

Exploiting natural contours for automatic sonar-to-video calibration

Statistical snakes: robust tracking of benthic contours under varying background

Statistical snakes: Application to tracking of benthic contours

Contact Info

Product

Resources

About