an application of dsmt in ontology-based fusion systems

8
An application of DSmT in ontology-based fusion systems Ksawery Krenc RS-SD R&D Marine Technology Centre Gdynia, Poland [email protected] Adam Kawalec The Institute of Radioelectronics WAT Military University of Technology Warsaw, Poland [email protected] Abstract The aim of this paper is to propose an ontology framework for preselected sensors due to the sensor networks’ needs, regarding a specific task, such as the target’s threat recognition. The problem will be solved methodologically, taking into account particularly non-deterministic nature of functions assigning the concept and the relation sets into the concept and relation lexicon sets respectively and vice-versa. This may effectively enhance the efficiency of the information fusion performed in sensor networks. Keywords: Attribute information fusion, DSmT, belief function, ontologies, sensor networks. 1 Introduction Ontologies of the most applied sensors do not take into account needs of sensor networks [1]. Sensors, in particular the more complex ones, like radars or sonars are intended to be utilized autonomously. The foundation of the sensor networks (SN), comprehended as the networks of cooperative monitoring, is understanding information obtained from some elements by another ones. Thus the question of the common language is very important. The ontology of sensor network should be unified and structured. The key problem in this paper is neither a direct application of existing solutions in the field of ontologies for the sensor networks nor a design of a new ontology, ready to implement. The aim is to propose the ontology framework for networks, consisting of preselected sensors, due to the sensory needs, to perform a specific task, such as recognizing the target threat. The selection of the sensors will be taken in four particular steps, namely: 1. Describing, what particular pieces of information are required to define the target threat; 2. Describing, what particular sensors enable to gain the mentioned pieces of information; 3. Identification of all information possible to acquire by preselected sensors; 4. The specific sensor selection; 2 Sensor type selection This section focuses on creating the ontology of a sensor network, processing information related to the target threat attribute. Mentioned information may be classified, according to its origin, as: Observable – originated directly from sensors or visual sightings; Deductable (abductable) – designated by the way of deductive reasoning, based on the other observable attributes, gathered previously; Observable and deductable – designated both: on the basis of observation and by the way of deductive reasoning; Confirmed – verified by other information center or external sensor network; The observable attributes may be defined based on information originated from diverse sensors. For the purpose of this paper the scope of sensors (possible to utilize) will be constrained to the set, which in the authors’ opinion fully reflects the required information about the target in the real world. It is a very important assumption that the selection of sensor types is conditioned ontologically. That means neither any particular sensor model nor communication protocol nor any other element of the SN organizations will be discussed. From the observer’s point of view (whose main duty is to assess the target threat) it is important to define the following features of the target: Key attribute of the target: the threat (based on observations); Additional target attributes (as the basis for deduction reasoning about the threat) i.e. the platform, (frigate, corvette, destroyer) and the activity (attack, reconnaissance, search & rescue); Auxiliary characteristics: target position; 12th International Conference on Information Fusion Seattle, WA, USA, July 6-9, 2009 978-0-9824438-0-4 ©2009 ISIF 1218

Upload: don-hass

Post on 19-Aug-2015

6 views

Category:

Documents


2 download

DESCRIPTION

The aim of this paper is to propose anontology framework for preselected sensors due to the sensor networks’ needs, regarding a specific task, such asthe target’s threat recognition.

TRANSCRIPT

An application of DSmT in ontology-based fusion systems KsaweryKrenc RS-SD R&D Marine Technology Centre Gdynia, Poland [email protected] Adam Kawalec The Institute of Radioelectronics WAT Military University of Technology Warsaw, Poland [email protected] AbstractTheaimofthispaperistoproposean ontologyframeworkforpreselectedsensorsduetothe sensor networks needs, regarding a specific task, such as thetargetsthreatrecognition.Theproblemwillbe solved methodologically, taking into account particularly non-deterministicnatureoffunctionsassigningthe concept and the relation sets into the concept and relation lexiconsetsrespectivelyandvice-versa.Thismay effectivelyenhancetheefficiencyoftheinformation fusion performed in sensor networks. Keywords:Attributeinformationfusion,DSmT,belief function, ontologies, sensor networks. 1Introduction Ontologiesofthemostappliedsensorsdonottakeinto accountneedsofsensornetworks[1].Sensors,in particular the more complex ones, like radars or sonars are intended to be utilized autonomously. Thefoundationofthesensornetworks(SN), comprehended as the networks of cooperative monitoring, isunderstandinginformationobtainedfromsome elementsbyanotherones.Thusthequestionofthe commonlanguageisveryimportant.Theontologyof sensor network should be unified and structured.Thekeyprobleminthispaperisneitheradirect application of existing solutions in the field of ontologies forthesensornetworksnoradesignofanewontology, readytoimplement.Theaimistoproposetheontology frameworkfornetworks,consistingofpreselected sensors,duetothesensoryneeds,toperformaspecific task, such as recognizing the target threat. Theselectionofthesensorswillbetaken in four particular steps, namely: 1.Describing, what particular pieces of information are required to define the target threat; 2.Describing,whatparticularsensorsenabletogain the mentioned pieces of information; 3.Identificationofallinformationpossibletoacquire by preselected sensors; 4.The specific sensor selection; 2Sensor type selection Thissectionfocusesoncreatingtheontologyofasensor network,processinginformationrelatedtothetarget threat attribute. Mentioned information may be classified, according to its origin, as: Observableoriginateddirectlyfrom sensors or visual sightings; Deductable (abductable) designated by the way ofdeductivereasoning,basedontheother observable attributes, gathered previously; Observable and deductable designated both: on thebasisofobservationandbythewayof deductive reasoning; Confirmed verified by other information center or external sensor network;Theobservableattributesmaybedefinedbasedon informationoriginatedfromdiversesensors.Forthe purposeofthispaperthescopeofsensors(possibleto utilize)willbeconstrainedtotheset,whichinthe authorsopinionfullyreflectstherequiredinformation about the target in the real world.It is a very important assumption that the selection of sensortypesisconditionedontologically.Thatmeans neitheranyparticularsensormodelnorcommunication protocolnoranyotherelementoftheSNorganizations will be discussed. From the observers point of view (whose main duty istoassessthetargetthreat)itisimportanttodefinethe following features of the target: Keyattributeofthetarget:thethreat(basedon observations); Additionaltargetattributes(asthebasisfor deductionreasoningaboutthethreat)i.e.the platform,(frigate,corvette,destroyer)andthe activity(attack,reconnaissance,search& rescue); Auxiliary characteristics: target position; 12th International Conference on Information FusionSeattle, WA, USA, July 6-9, 2009978-0-9824438-0-4 2009 ISIF 12182.1Types of sensors Preselectedtargetfeaturesmayberegisteredbyvarious means of observation, namely: Position:Radar(allspatialdimensions),sonar, IRsensor(mostlytodefinetargetazimuthand elevation); Threat:IFF,visualsightings(human),video camera (daylight or noctovision); Platform:visualsightings,videocamera, thermo-vision camera; Activity: visual sightings. Theabovestatementmayberegardedasapre-selectionofsensorset,usedinthefollowing considerations of this paper. It is important to notice, that someofthementionedsensorsmayacquireinformation relatedtomorethanoneattribute.Therefore,areversed assignment(sensorstoattributes)seemstobemore adequate.2.2Sensor-originated information Figure1presentsthepreselectedtarget features and their inclusionrelations.Additionally,itwaspointedoutthe examplesensors,whichenabletoacquirethementioned information. PositionRadarIFFVideocameraThreatPlatformActivityVisualsightings Figure 1 Information scope originated from diverse types of sensors. Itshouldbenotedthatalthoughsomeofthese sources allow for obtaining information on more than one attribute, it is possible to identify a hierarchy of relevance ofthisinformation.Thatmeansthatsomeofthe attributes,however,possibletorevealfrommultiple sources,forsomesourcesperformtheprimary information while for others the secondary information: Radar: position1; IFF: position, threat; Video camera: position, platform, threat; Visualsightings:position,threat,platform, activity; For visual sightings, where the human plays the role ofthesensor,itisdifficulttoidentifytheprimary information.Amongtheabovesourcesthevisual recognition is the most reliable way of defining the target activity.Therefore,takingintoaccountthefactthatit allows to identify the target threat and platform, the visual recognitionmaybeconsideredasaspecificsourceof information.Theseobservationsarehighlyimportantforfuture considerations,whichwillbeeffectivelyusedincreation ofthehierarchyoftheconceptlexiconsaswellasin defining the relations among concepts of SN ontology. Someofthesesensorsperformverycomplex devicesandrequiretheintroductionofcertaininterfaces, allowingtheautomaticacquisitionofusefulinformation (intermsofsensornetworks).Anexampleofsucha sensor is a video camera. In order to make effective use of animagefromthevideocameraaspecificmoduleis necessarytointerpretthetakenpicture,identifyingthe significantfeaturesoftheobjectofinterest.Inthatcase, theontology,thevideocameraisdefinedinthatvery moduleanditismodifiableaslongasthereisaccessto theconfigurationofthatmodule.Thisleadstoanother possible classification of sensors: Constant (invariant) ontology sensors, e.g. IFF; Variantontologysensors,e.g.videocamera equippedwithinterpretationmoduleorvisual sightings; Guidedbytheprincipleofmaximuminformation growth,innextstagesofcreatingtheSNontologythe followingsourcesofattributeinformationwillbetaken into account: IFF, video camera (VC) and visual sightings (VS). 3Defining sets of SN ontologies Referringtoataxonomyofthetermofontology[1]the authorswouldliketonoticethattheproblemofSN ontology concerns, in particular, the so-called method and task ontologies. Therehavebeeneffectivelyutilizedconcept lexiconsofJointC3InformationExchangeDataModel 1 Underline means the prime information. 1219[2],constrainingtheconsiderationstothreeoftheJC3 model attributes: threat: object-item-hostility-status-code; platform: surface-vessel-type-category-code; activity: action-task-activity-code; While defining the attribute relation functions, the Dezert-SmarandacheTheory(DSmT)ofplausibleand paradoxical reasoninghas been utilized [3]. 3.1Rules for sensor network ontologies selection In section 2.2 there was proposed a sensor distinction for variantandinvariantontologysensors.Consideringthis divisionisfundamentalwhilecreatingSNontology, which takes place in four stages: 1.Creatingthefundamentalconceptlexiconfora sensornetwork,basedoninvariantconcept lexicons of particular sensors; 2.Creating the auxiliary concept lexicon for sensor network,basedonvariantconceptlexiconsof particular sensors; 3.Extendingthefundamentalconceptlexiconwith the auxiliary lexicon; 4.Definingrelationsamongtheconceptsinsensor network;According to the definition of ontology, given in [4], [5], SN ontology may be formulated as follows:R C G F G F L O , , , , , , = (1) where: L is either concept or relation lexicon; Flexiconelementstoconceptsassigning function; Glexiconelementstorelations assigning function;F afunctionreversedtoF, assigningconceptstoelementsoftheconcept lexicon;G afunction reversed to G, assigning relations to elements of the relation lexicon;Casetof the whole concepts used in SN;Raset of the whole relations used in SN. AccordingtothelexiconsofJC3model,theabove mentionedconceptsandfunctionswillbedefinedinthe following subsections. 3.1.1Concepts Concepts are representations of a certain group of objects ofthesamecharacteristics,whichmaybedirectly identifiedbyselectedsubsetofelementsoftheconcept lexicon[5].Thatmeans,thatassigningforexamplean attribute hostile target to a target uses the concept of the hostiletarget,whichistheelementoftheset(C)ofall possible concepts for a given sensor network. Anotherquestionisarepresentationoftheconcept hostiletargetinthelanguageoftheparticularsource. For instance: for IFF device it will be the value of FOE, andforavideocamerathevalue,definedinthe interpreting module as HOSTILE.Mathematically,theFassignmentisnotabijection in general, moreover: it is not a function. In case multiple sourcesareutilized,theFisnotaninjection,whereasif the concept set is rich, comparably to the poor lexicon theFisnotinjective.ThismayoccuriftheSN,prepared fordefiningfullytargetthreat,isusedfordeciding whether the target is either friend or hostile. Then, the F willinterpretconceptsoftraininghostile,training suspectandassumedfriendasfriendassigningthe lexical value ofFRIEND [6]. InordertoillustrateFand Fassignmentitis suggested to consider the following example.Example 1:Letthesetofconceptsbedefinedas follows:C = {friend, assumed friend ,assumed hostile, hostile} (2) and the concept lexicon is defined as follows: Lc = {FRIEND, HOSTILE, ASSUMED} (3) Thus,itispossibletodefinesubsetsoftheconcept lexiconelementsinsuchawaythattheFassignment would be a bijection (Figure 2).FRIENDASSUMED FRIENDASSUMED HOSTILEHOSTILEfriendly targetassumed friendlytargetassumed enemytargetenemy targetFFFFFFFF Figure 2 F-assignment as a bijection. 1220Definingsubsetsoflexicalelementsassingletons leads to non-function F assignment (Figure 3). ASSUMEDfriendly targetassumed friendlytargetassumed enemytargetenemy targetFRIEND HOSTILEF Figure 3 F as a non-function assignment. Incaseofrichconceptlexiconsetsitisimportant to express subsequent target types as conjunctions of their distinctive features. Example 2: Table 1 Example definitions of surface platforms Transporter AUX AIR D TRAN Command AUXS&MCALAIR C2 where: AUX auxiliary vessel; S&MCALequippedwithartilleryofsmalland medium caliber; AIR against the air targets; D performs landing operations; C2 command & control; TRAN transport of landing forces; 3.1.2Relations Relationsdefinetherelationshipsamongconcepts. Relationmaybehierarchicalorstructural.Moreover,for the purpose of sensor networks, they may be classified as: Relations I, among the observable attributes ofa diverse type; RelationsII,amongattributesofmiscellaneous origin; RelationsIII,amongtheidenticalattributes, originated from diverse sources; Relationsamongtheobservableattributesofa diversetypeenableadeductionofsomeattributes values based on observable values of another ones. For instance: therelationsbetweenthethreatandtheplatformofthe targetenablethedeductionoftargetactivity.Linkingthesubsequentobservableattributesisperformedaccording to mentioned in previous section distinctive features of the target.Thismeansthatforexample:defining(basedon observations)thetargetplatformisequaltoassigningto thetargetsomeofdistinctivefeatures,whichthetarget, performing the particular activity, has to possess. Relationsamongattributesofmiscellaneousorigin: observableanddeductableresultinso-calledobservable-deductableattribute.Theeffectiveinformationfusion from multiple sources is performed according to the rules ofcombinationandconditioning,obtainedfromDSmT [7], [8]. This process is going to be described in details in section 3.2. Relationsamongtheidenticalattributes,originated fromdiversesourcesarethetypeofrelations,wherethe keyquestionisalexicalvarietyofconceptsusedby particularsources.Forinstance:thethreatattributevalue acquiredfromIFFmaybeeitherFRIENDorFOE, whereas the same attribute obtained from visual sightings maybeof{FRIEND,HOSTILE,UNKNOWN,JOKER, FAKER,}.InsuchacaseavalueofFRIEND,gained fromIFF,correspondstotheexactvalueofthevisual sightings.ThevalueofFOEisequaltoHOSTILE, whereastherelationsamongvaluesofFRIEND,gained fromIFFandFAKER(orJOKER),gainedfromthe visualsightingsarenotsoobviousandtheymustbe defined,accordingtothedefinitionsofthesetraining types (JOKER, FAKER).3.2Proposition of sensor network ontology Thissectionpresentsapropositionofanontology framework for a sensor network, dedicated to monitor the targetthreat.Inthesolutiontherewereutilizedconcepts andconceptlexiconsofJC3model.Theauthors intention was to show the way relations of three attributes (threat,platformandactivity)shouldbedefined,rather than to present the complete SN ontology. Table2presentsabijectiveassignmentofconcepts toelementsofaconceptlexicon.Asitwasmentioned before, this assignment need not be a bijection, however it isdesirableespeciallyifsetsofvaluesforattributesof platform and activity are numerous. Table 2 SN ontology: concepts and concept lexicon. ConceptsConcept lexicon AnOBJECT-ITEMthat isassumedtobea friend becauseofits characteristics,behavior or origin. ASSUMED FRIEND Threat AnOBJECT-ITEMthat object-item-hostility-status-codeHOSTILE 1221ispositivelyidentifiedas enemy. according to JC3 according to JC3 Generaldesignatorfor aircraft/multi-role aircraft carrier; AIRCRAFT CARRIER, GENERAL Craft40metersorless employedtotransport sick/woundedand/or medical personnel. AMBULANCE BOAT Platform according to JC3 surface-vessel-type-category-code according to JC3 Toflyoveranarea, monitorand,where necessary, destroy hostile aircraft,aswellas protectfriendlyshipping inthevicinityofthe objective area. PATROL, MARITIME Emplacementor deploymentofoneor more mines. MINE-LAYING Activity according to JC3 action-task-activity-code according to JC3 Theassignmentofrelationsamongattributesto relationlexicons(Table3)isasurjection.Inorderto definetherelationsamongattributesDSmTcombining andconditioningruleshavebeenapplied.Thepreferred ruleforconditioningistheruleno.12.Whencombining evidence,thereisapossibilitytousemanycombination rules,dependingtheparticularrelation.However,for simplicity,itissuggestedtoapplytheclassicruleof combination(DSmC),whichhaspropertiesof commutativity and associativity. Table 3 SN ontology: relations and relation lexicon. RelationsRemarksRelation lexicon cond(.)Based on DSmTConditioningRel. I: Accordingto distinctive features Implication cond(.)Based on DSmTConditioningRel. II: Based on DSmTCombination cond(.)Based on DSmTConditioningRel. III: BasedonDSmT (combinationrule neednotbeidentical withoneinRelations II) Combination Below,therehavebeenpresentedexamplesof particular types of relations. In case of the relation of type Iitispossibletoreasonaboutavalueofacertain attribute,basedontheknowledgeabouttheotherones. However,iftheunambiguousdeductionofthethird attributeisnotpossible,duetothemajorityofpossible solutions, an application of abductive reasoning (selection of the optimal variant) seems to be justified.Relations I: (Threat,Platform)Activity:(FAKER,FRIGATE TRAINING) TRAIN OPERATIONS; (Threat,Activity)Platform:(FAKER,TRAIN OPERATIONS) TRAINING CRAFT; (Platform,Activity)Threat:(HOUSEBOAT, PROVIDE CAMPS) NEUTRAL; Relations II: FAKER=cond(obs(FAKER) ded(FAKER) obs(FRIEND)); Relations III: FAKER=cond(obs(FAKER) VS(FAKER) IFF(FRIEND)); Theabductivereasoningprocessmaybesystemizedby applicationofDSmT,wheretheselectionoftheoptimal valuetakesplaceaftercalculatingthebasicbelief assignment. Example 3: (Threat,Activity)Platform:(FRIEND,MINE HUNTING MARITIME) MINEHUNTER COASTAL (MHC)

MINEHUNTERCOASTALWITHDRONE(MHCD) MINEHUNTER GENERAL (MH) MINEHUNTER INSHORE (MHI) MINEHUNTEROCEAN(MHO) MINEHUNTER/SWEEPER COASTAL (MHSC)

MINEHUNTER/SWEEPER GENERAL (MHS) MINEHUNTER/SWEEPEROCEAN(MHSO) MINEHUNTER/SWEEPER W/DRONE (MHSD) Applying DSmT, for each of possible hypothesis a certain mass of belief is assigned, e.g.: m(MHC) = 0.2, m(MHCD) = 0.3,m(MH)=0.1,m(MHI) = 0.1,m(MHO) = 0.1, m(MHSC)=0.05,m(MHS) = 0.05, m(MHSO) = 0.05,m(MHSD)= 0.05 Based on the obtained basic belief assignment (bba) belief functions,referringtoparticularhypotheses,maybe calculated.Inthesimplestcase,assumingallofthe hypothesesareexclusive,thesubsequentbelieffunctions willbeequaltorespectivemasses,e.g.Bel(MHC)= m(MHC), Bel(MHCD) = m(MHCD), etc. Morecomplexcase,whererelationshipsamong hypothesesaretakenintoaccountwillbeconsideredin the next section. 12224Verificationoftheusefulnessof elaborated ontology sets ThepresentedframeworkoftheSNontology,forthe purposeofthetargetthreatassessment,requiresa verification.Particularly,itisimportanttoverifythe correctnessofreasoningprocessesandacombinationof the reasoning results with observation information.Theproposedsolutionsubstantiallydiffersfromthe existingdeterministicontology-basedmethodsbecauseit introducesexplicitlytheuncertaintyoftherelations amongtargetattributes.Therefore this section was meant tofocusontheverificationoftheserelationreasoning mechanismsratherthanthecompletenessofthetarget representation by the sensor network.4.1Assumptions In order to verify the usefulness of the proposed ontology framework, a specially designed demonstrator application forevaluationofthetargetthreatinformationhasbeen used. This application enables a simulation of acquiring of informationfromdiversesources,like:radar,video camera and visual sightings.It is assumed that the visual sighting is also a source ofinformationaboutatargetplatformandatarget activity.Thebbavaluesforplatformandactivity attributeshavebeenassignedarbitrary.During experimentationtheobservableattributesaswellas deductableattributeshavebeentakenintoaccount. Frames of discernment for observation and deduction may differingeneral.Forthepurposeofverificationof proposedontologysets,anexamplefromthesection3.2 is to be considered. Additionally it is assumed: Application of the hybrid DSmT model: oThe hypotheses are not exclusive; oThehypothesescorrespondtotheJC3 model terminology; InrelationsoftypeIIandIIIthehybridruleof combination (DSmH) has been applied; Theconditioningruleno.12hasbeenusedfor updating evidences; 4.2Numerical experiments Figure4showsarandomlygeneratedtrajectoryofthe targetofwhichthethreatvalueisatstake.Observations aretakenfromthreesources(visualsightings,radar system - IFF and video camera) synchronously.Thegreencolormeanssuccessivelyacquired observations for each of the sources. The red color means theobservationsimpossibletoacquirebecausethetarget wasoutsideofthedetectionzoneforaparticularsource [3].Takingforexamplethelastsample,therespective bba are as Table 4 shows. Figure 4 Randomly generated target trajectory and its threat evaluation based on radar, VS and VC observations. Table4Bbagatheredfromdiversesources:visual sightings, video camera and radar. Threat Visual Sightings Video Camera Radar/IFFHOS 0.00240.00040.0008 UNK 0.00600.0012- NEU 0.00680.0015- JOK 0.0109 - - FRD 0.24000.43680.8773 FAK 0.02920.00490.0119 SUS 0.00320.00050.0011 AFR 0.02150.00460.0088 PEN 0.68000.55000.1000 A relation of type III of combining information from IFF and the visual sightings results in acceptance the target is friendly: FRIEND Threat ThreatIFF VS (4) From the visual sightings it is also acquired that the target activityismine-hunting(MINEHUNTINGMARITIME). Thus,therelationoftypeI,betweenthethreatandthe activity attribute results in selection of the target platform, related to searching for mines. (FRIEND, MINE HUNTING MARITIME) platform(5) 1223Intheconsideredcaseitisassumedtheframeof discernmentoftheplatformattributeoriginatedfromthe video camera is defined as follows: VC = {MHC, MHI, MHO, MSC, MSO, D}(6) where: MHC MINEHUNTER COASTAL;MHI MINEHUNTER INSHORE;MHO MINEHUNTER OCEAN; MSC SWEEPER COASTAL;MSO SWEEPER OCEAN;D DRONE; Additionally,with and operatorsthesecondary hypotheses may be created, which refer to another values oftheplatformattribute(surface-vessel-type-category code) of JC3 model: MHCD=MHCD(MINEHUNTERCOASTAL WITH DRONE);MHIMHOMHCD=MH(MINEHUNTER GENERAL);MHOMSO=MHSO(MINEHUNTER/SWEEPER OCEAN); (MHCMSC)D=MHSD (MINEHUNTER/SWEEPERW/DRONE);(MHOMSO)(MHCMSC)D=MHS (MINEHUNTER/SWEEPER GENERAL); Thebasicbeliefassignmentforthevideocamera observation may be defined as follows: mVC(MHC) = 0.1, mVC (MHCD) = 0.1, mVC (MSC) = 0.2, mVC (MHI) = 0.3,mVC (MHO) = 0.2,mVC (MSO) = 0.1,Due to the implication (5) the above bba may be modified according to BCR12 with a following condition: MHI MHO MHC Truth Cond = : (7) Figure 5 Venn's diagram for the platform attribute. The truth is grey colored. Thus,theresultingbbafortheplatformattributeis updated, as follows: mR(MHC|Cond)=mVC(MHC)+mVC(MHCD)=0.2,mR(MHSC|Cond)=mVC(MSC)=0.2, mR(MHI|Cond)=mVC(MHI)=0.3,mR(MHO|Cond)=mVC(MHO)=0.2, mR(MHSO|Cond)=mVC(MSO)=0.1, which,aftercalculatingtherespectivebeliefand plausibilityfunctions,leadstoacceptationofthe hypothesisofMHC(MINEHUNTERCOASTAL)forthe platform attribute of the whole sensor network. It is worth of notice that the belief function for MHC before updating is of the least value since: BelVC (MHC) = mVC(MHC)= 0.1(8) Afterupdating,duetothefactthatmVC(MHSC)supports thebeliefinMHChypothesis,thishypothesisbecomes the most credible since: BelR (MHC) = mR(MHC)+ mR(MHSC)= 0.4(9) 5Conclusions The results of the numerical experiments, presented in the previoussection,haveproventhattheapplicationof DSmTforthepurposeofdefiningrelationsamongtarget attributes,givesthepossibilityofunificationof informationacquiredfromsensorsaswellasobtained basedonthedeductivereasoning.Thatinfluences effectively the wholeSN ontology, due to the fact the SN conceptlexiconbecomessubstantiallymodified.Itdoes notprovideaunionoflexiconsforeachsensor,which wouldbeexpectableinthedeterministiccase.TheSN conceptlexiconbecomesextendedwithintersectionsand unionsofthehypothesescreateduponthelexiconsof particular sensors. DuringtheexperimentsithasbeenutilizedtheJC3 modelslexiconofsurface-vessel-type-category-code attribute.Itisimportanttonotice,thatdespiteitslarge volume,thelexiconisnotstructured.Thus,anemerging conclusion occurs, that setting JC3 lexicons in a hierarchy wouldbringtangiblebenefitsduetothefactthatthe hierarchyenablescreatingthehypothesesusing and operatorsmoreeffectively,andthisinturnincreases theprecisionofthereasoningprocessesbasedon information acquired from sensors. Acknowledgements The research work described in this document is a part of extensiveworksdevotedtosensornetworksinNEC environment.Itwasfinancedwithsciencemeansfrom 2007 to 2010 as an ordered research project. 1224References [1]K. Krenc: An introductory analysis of the usefulness ofsensornetworkorganizations,MCCConference, Cracow, ISBN 83-920120-5-4, 2008. [2]TheJointC3InformationExchangeDataModel, Edition 3.1b, 2007 [3]K. Krenc, A. Kawalec: An evaluation of the attribute informationforthepurposeofDSmTfusioninC&C systems, Fusion2008, Cologne, ISBN 978-3-00-024883-2, 2008. [4]http://www.aifb.uni-karlsruhe.de/WBS/meh/publications/ehrig03ontology.pdf [5]M.Chmielewski,R.Kasprzyk:Usageand characteristicsofontologymodelsinNetworkEnabled Capabilityoperations,MCCConference,Cracow,ISBN 83-920120-5-4, 2008. [6]NATOStandardizationAgency,TacticalData Exchange Link 16, STANAG No. 5516, Ed. 3. [7]FlorentinSmarandache,JeanDezert,Advancesand ApplicationsofDSmTforInformationFusion,Vol1, American Research Press Rehoboth, 2004. [8]FlorentinSmarandache,JeanDezert,Advancesand ApplicationsofDSmTforInformationFusion,Vol2, American Research Press Rehoboth, 2006. 1225