Use of the Dempster-Shafer algorithm for the detection of SAR ship wakes

Rey, María Teresa de Tena; Tunaley, J. K. E.; Sibbald, Timothy

doi:10.1109/36.263786

Cited by 23 publications

(7 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The power of evidential reasoning stems both from the generality of its formulation, and also Downloaded by [Temple University Libraries] at 18:15 20 November 2014 from the fact that it was designed speci cally to provide a rich set of calculus for combining disparate, multisource data with di V erent properties and levels of information. It has been used in a variety of applications such as medicine (Gordon and Shortli V e 1985), forestry (Goldberg et al 1985), expert systems and arti cial intelligence (Shafer and Logan 1987), route selection (Garvey 1987), geology (Moon 1990(Moon , 1993, classi cation (Srinivasan andRichards 1990, Wilkinson andMégier 1990), water resources (Caselton and Luo 1992) and target detection and reconnaissance studies (Rey et al 1993). In the classi cation context, evidential reasoning regards information from di V erent data sources as individual pieces of evidence over a set of prede ned or hierarchical classes.…”

Section: T He Evidential Reasoning Approachmentioning

confidence: 98%

Evidential reasoning with Landsat TM, DEM and GIS data for landcover classification in support of grizzly bear habitat mapping

Franklin¹,

Peddle²,

Dechka³

et al. 2002

International Journal of Remote Sensing

View full text Add to dashboard Cite

Multisource data consisting of satellite imagery, topographic descriptors derived from DEMs, and GIS inventory information have been used with a detailed, eld-based landcover classi cation scheme to support a quantitative analysis of the spatial distribution and con guration of grizzly bear (Ursus arctos horribilis) habitat within the Alberta Yellowhead Ecosystem study area. The map is needed to determine if bear movement and habitat use patterns are a V ected by changing landscape conditions and human activities. We compared a multisource Evidential Reasoning (ER) classi cation algorithm, capable of handling this large and diverse data set, to a more conventional maximum likelihood decision rule which could only use a subset of the available data. The ER classi er provided an acceptable level of accuracy (ranging to 85% over 21 habitat classes) for a level 3 product, compared to 71% using a maximum likelihood classi er.

show abstract

Section: T He Evidential Reasoning Approachmentioning

confidence: 98%

Evidential reasoning with Landsat TM, DEM and GIS data for landcover classification in support of grizzly bear habitat mapping

Franklin¹,

Peddle²,

Dechka³

et al. 2002

International Journal of Remote Sensing

View full text Add to dashboard Cite

show abstract

“…All of them are uses characterized by very low radar cross section targets embedded in strong clutter and not easily differentiable using Doppler processing. Conventional Track Before Detection (TBD) techniques [8], [9] are no more well suited, and new processing schemes have to be included in order to detect targets rivaling the sea clutter [10], [11]. Its behavior, or statistical characterization, is no more "classic," due to HRR features [12], and requires new studies in order to improve the detection algorithms.…”

Section: A Clutter Power Reductionmentioning

confidence: 98%

Instrumental CWLFM High-Range Resolution Radar in Millimeter Waveband for ISAR Imaging

et al. 2011

View full text Add to dashboard Cite

This paper shows a High-Range Resolution (HRR) Radar prototype development, transmitting in millimeter-wave band. The waveform selected to be used is the continuous-wave linear frequency modulation (CWLFM), which allows, at the same time, good range resolution, moderate unambiguous doppler range and reasonable operational range taking into account that high-frequency band is used. The radar prototype looks for High-Range Resolution (HRR) characteristics as well as flexibility, in order to obtain the best Inverse Synthetic Aperture Radar (ISAR) images possible depending on the applications and targets involved. That is, depending on the speed and target's trajectories, some radar features, such as transmitted bandwidth, modulator frequency or sampling frequency, play a key role in obtaining the best possible results while using lowest electrical requirements.In Section IV, several different ISAR images are shown, pointing out not only the radar features achieved and their variety, but also their quality thanks to the hardware and electronics improvements developed.Index Terms-Continuous-wave linear frequency modulation (CWLFM), doppler processing, high-range resolution (HRR), inverse synthetic aperture radar (ISAR) images, millimeter waveband, RF zoom scheme.

show abstract

“…The Dempster-Shafer approach provides a heuristic basis for combining evidence from diverse and multi-source data sets and has been applied to problems of remote sensing image classification (Lee et al 1987;Srinivasan and Richards, 1990), geology (Moon, 1990(Moon, , 1993, water resources (Caselton and Luo, 1992), detecting ship wakes on SAR imagery (Rey et al,1993),route selection (Garvey, 1987),and as an interface to expert systems (Goldberg et al, 1985;Wilkinson and Megier, 1990) to measure the level of uncertainties in hypothesis testing. Although considerably less attention has been directed towards the use of DempsterShafer theory for remote sensing image classification compared to well-established neural network approaches, there is, nonetheless, an increasing recognition of the power of evidential reasoning for classifying higher-dimensional multi-source data sets with diverse statistical properties and for handling the inherent information uncertainty and conflicting knowledge typical of divergent data sources.…”

Section: Evidential Reasoning Classificationmentioning

confidence: 99%

Multi-Source Image Classification II: An Empirical Comparison of Evidential Reasoning and Neural Network Approaches

Peddle

Foody

Airong

et al. 1994

Canadian Journal of Remote Sensing

View full text Add to dashboard Cite

RESUMELes methodes classiques reconnues pour la classification d'images, comme la classification par maximum de vraisemblance, ne sont pas toujours eppropriees pour traiter, de nos jours, les donnees provenant de sources multiples en raison du volume eleve et de la diversite de ces donnees. Dans Ie present article, deux methodes sont examinees comme solution de rechange ala methode de classification par maximum de vraisemblance (MV) pour la classification d'images provenant de sources multiples; il s'agit d'un algorithme de classification par raissonement veridique (RV) selon la theorie de Dempster-Shafer et de deux algorithmes de classification par reseeux neuronaux artificiels (RNA). Chaque methode a ete mise al'epreuve dans des classifications distinctes de la couverture d'un sol de type alpin et de la profondeur des couches actives du pergeliso! dans une region montagneuse du sud-ouest du Yukon, aI'aide d'un ensemble de donnees SPOT, de donnees sur la texture des images et de variables geomorphometriques tirees d'un modele numerique d'elevetion. La comparaison du niveau d'exactitude global moyen a demontre que la methode de classification par RV offrait un niveau d'exactitude de 10 % superieur acelui de la methode de classification par MV et que la methode de classification par RNA presenteit un niveau d'exactitude de 5 % superieur acelui de la methode par RV. Les coefficients d'egrement de Kapp les plus eleves obtenus pour la couverture du sol furent 0,79,0,90 et 0,96 avec la methode par MV, la methode par RV et la methode par RNA respectivement. Des differences importantes du niveau d'exactitude des deux algorithmes de classification par RNA ont ete observees dans plusieurs cas. Ces differences sont attribuables aux problemes poses par la determination des parametres optimaux d'entrai'nement des reseeux neuronaux aI'aide de donnees d'entree subjectives. Pour ce qui est de la duree de calcul, I'algorithme de classification par RV s'est revele plusieurs ordres de grandeur plus rapide que les algorithmes de classification par RNA. Les conclusions que tirent les auteurs de ces experiences sont les suivantes : I'algorithme de classification par RV et les algorithmes de classification par RNA presentent des cepecites superieutes ala methode de classification par MV pour traiter des donnees provenant de sources multiples; I'algorithme de classification par raisonnement veridique est plus rapide et moins subjectif que la methode de classification par teseeux neuronaux, cependant que cette demiere, en raison de son niveau d'exactitude plus eieve, est celie que les auteurs recommandent pour I'instant. La methode, en effet, a donne les resultets escomptes compte tenu de I'avancement de la recherche dans Ie domaine des reseeux neuronaux, par rapport ala methode relativement nouvelle du raisonnement veridique dont les applications en teledetection en sont aux premiers stades de developpement; il s'evere iustiti« de mener des recherches plus pousses pour developper les algorithmes de classification par raisonnement verid...

show abstract

Use of the Dempster-Shafer algorithm for the detection of SAR ship wakes

Cited by 23 publications

References 4 publications

Evidential reasoning with Landsat TM, DEM and GIS data for landcover classification in support of grizzly bear habitat mapping

Evidential reasoning with Landsat TM, DEM and GIS data for landcover classification in support of grizzly bear habitat mapping

Instrumental CWLFM High-Range Resolution Radar in Millimeter Waveband for ISAR Imaging

Multi-Source Image Classification II: An Empirical Comparison of Evidential Reasoning and Neural Network Approaches

Contact Info

Product

Resources

About