,GSS i RP RPs I, I=1,2,,N, N ;A C I ;C I sporadic,gss) RP 8 AP;O C I AP GSS i O i,o i={sn E,sn EN,sn N,sn NW,sn W,sn WS, RSSI 1.2 NN, 0 1(

Size: px

Start display at page:

Download ",GSS i RP RPs I, I=1,2,,N, N ;A C I ;C I sporadic,gss) RP 8 AP;O C I AP GSS i O i,o i={sn E,sn EN,sn N,sn NW,sn W,sn WS, RSSI 1.2 NN, 0 1("

包陈
5 years ago
Views:

1 GeomaticsandInformationScienceofWuhanUniversity Vol.39No.11 Nov.2014 DOI /j.whugis (2014) NN 1 1 1,, 针对室内环境基于 RSSI 定位不稳定问题, 提出了以几何信息改进基于指纹库的 NN 定位算法根据室内几何布局建立了聚类指纹库, 提出了表征点位几何特性的点散发性强度 (geometricstrengthofsporad- ic,gss) 概念利用最邻近样本点的 GSS 判别移动终端所在参考点 RP 控制网结构以动态选择 NN 关键参数, 构建最佳多边形为约束准则自适应选取后 -1 个邻近点, 建立了基于几何聚类指纹库的约束加权 NN 室内定位模型结果表明, 改进后定位模型可以更好地估计终端位置信息, 其中几何聚类指纹库是改善定位准确性的关键, 约束 NN 能够有效地提高室内定位精度室内定位 ; 聚类指纹库 ; 几何特性 ;RP 控制网结构 ; 约束 NN TN A, (referencepoint,rp) RSSI IEEE (ac-, RSSI cessedpoints,aps) RSSI, (X RP,Y RP ) [5], i [1-4] P(A 1O 1 Pt i ) P(A ko 1 Pt i ) 熿燄 R i =, 燀 P(A 1O v Pt i ) P(A ko v Pt i ) 燅 [5-7] (1) [8-9],,,A AP;O RSSI ;Pt,, NN,,,,, RSSI,, 1 (2) RP i = P(A 1O 1 (θi C I )) P(A ko 1 (θi C I )) 熿燄 1.1 燀 P(A 1O v (θi C I )) P(A ko v (θi C I )) 燅, (2),,θi=(Pt i,gss i ) RP (2013AA12A201); ; (2013RC16); (NCET ),, lcy sia@163.com

2 ,GSS i RP RPs I, I=1,2,,N, N ;A C I ;C I sporadic,gss) RP 8 AP;O C I AP GSS i O i,o i={sn E,sn EN,sn N,sn NW,sn W,sn WS, RSSI 1.2 NN, 0 1(a) 4 MT(mobileterminal) RP MT GSS , 1(b) 3 [10] MT GSS 7 8 5,, T( ) GSS RP, RP, GSS i 2,, S,T S N M r= (T nm S nm ) (3) n=1 m=1, r= N M [(T nm -μ ) (S T nm -μ )] S n=1 m=1 N M n=1 m=1 槡 (T nm -μ T ) N 2 槡 n=1 M m=1 (S nm -μ ) S 2 (4), μ T μ S, NN, 2 NN Honkavirta [7], NN NN, RP ;Beomju [11] 3 4, RP, 2, GSS, RP -1, 2, -1 ; 3 (geometricstrengthof, sn S,sn SE } RP i 8,sn, 2.1 2, MT1, 1 1 1(a), (RP ) 2(RP ),, MT2 (RP 6-7 ) ; 2 1(b), 1 RP Fig.1 LayoutofWLANSignalReferencePoint 2.2 NN,, RP,,

3 39 11 NN ; Fig.2 OptimalPolygonConstraintCriteria, GSS 烄 X MT = 烅 Y MT = 烆 (r j (X RPj C i )) r j (r j (Y RPj C i )) GSS 2, =2 ;GSS>2, =3 4 r j (5),{(X RP,Y RP )C I} I 试验 1 ;r s ;(X MT, Y MT ) 2.3 NN,,, ; AP RP,, -1, NN 3 WiFi, 4, 4, 4(a) 70 m C 1.2 m 1.2 m, NN, 4(b) 3 RSSI NN Fig.3 BasedonRSSIGeometricClustering-constraintWeightedNN Algorithm Model 试验 m,, 2208m 2, 7,, AP 4(c),, 5 ~ 10 3m, dbm [12], NN, 5(c), RP 4 AP, 30, 120 5(a) 5(b) 4,,,

1290 2014 11 4 Fig.4 TestEnvironment 5 7 NR (NCRS,CS) Fig.5 AnalysisofaSamplingPointSignalStrength R (NCS,CRS),, NCS 30 NCRS,,,, NN RP NN,, ; CRS -1 ; NN 3 CS, 试验 1 4, NN 1.3m 1.8m NN 6(a), NN,,, 1.

4 Fig.4 TestEnvironment 5 7 NR (NCRS,CS) Fig.5 AnalysisofaSamplingPointSignalStrength R (NCS,CRS),, NCS 30 NCRS,,,, NN RP NN,, ; CRS -1 ; NN 3 CS, 试验 1 4, NN 1.3m 1.8m NN 6(a), NN,,, 1.36m 3.31m 1 NC C NC - (NCRS) - (NCS)NN,C - (CS) - (CRS)NN;2 NR R NR - NN,R - NN Tab.1 试验 2 17, 1 NN WeightedNN Positioning Models NCRS NCS CS CRS NN 6(b), ; NN, 70%, (a) NC (NCRS,NCS) C (CS,CRS),,, ; 7(b) NC C, NC 5 13, NN -1,

39 11 NN 1291 6 Fig.6 TestsResults 7 2 Fig.7 ErrorAnalysisofDiferentAlgorithms,,2009,34(11)1372-1376) NN 53.1%; [4] LiuJunning,ZhaiChuanrun,SongDier,etal.

5 39 11 NN Fig.6 TestsResults 7 2 Fig.7 ErrorAnalysisofDiferentAlgorithms,,2009,34(11) ) NN 53.1%; [4] LiuJunning,ZhaiChuanrun,SongDier,etal.In-, doorpseudolitesprecisepointpositioningbasedon ImprovedTime-SatelitesDiference[J].Geomatics 3 ;, and Information Science of Wuhan University, 29.8%; 2009,34(1) (,,,. NN 72.6% [J].,2009,34(1), ) [5] LingPei,ChenRuizhi,LiuJingbin.UsingInquiry- 4 [6] LingPei,ChenRuizhi,LiuJingbin.Inquiry-based RSSI BluetoothIndoor Positioning via RSSIProbability,, Distributions[C].2010 SecondInternationalCon-, ferenceonadvancesinsateliteandspacecommu- ; nications,tbd,athens,greece,2010, [7] HonkavirtaV,PeralaT,Ali-LoytyS.A Compara-, tive Survey of WLAN Location Fingerprinting -1 Methods[C].Positioning,NavigationandCommu- [1] Pahlavan,LiXinrong.IndoorGeo-locationSci- enceand Technology[J].IEEE Communications Magazine,2002,40(2) [2] JaegeolYim.IntroducingaDecisionTree-basedIn- door Positioning Technique[J].ExpertSystems with Applications,2008,34(2) [3] Tian Hui,XiaLinyuan,MoZhiming,etal.Signals ofopportunityassistedubiquitouspositioningand ItseyElementsforOutdoor/IndoorEnvironment basedbluetooth RSSIProbability Distributionsfor IndoorPositioning[J].Journalof GlobalPositio- ning Systems,2010,9(2) munications,forth Worth,Texas,USA,2003 NN, nication,hannover,germany,2009 [8] NN YousseF,Agrawala M A,Shankar A,etal.A ProbabilisticClustering-BasedIndoorLocation De- terminationsystem[r].technicalreportsofthe ComputerScience Department, Washington D C, [9] YousseF,Agrawala M A,UdayaShankarA,et al.locationdeterminationviaclusteringandprob- abilitydistributions[c].thefirstieeeinterna- tionalconferenceonpervasivecomputingandcom- [10]YangDongyong,GuDongyuan,FuXiaojie.AnIn- doorlocation Algorithm Base on RSSI-Similarity Degree[J].ChineseJournalof Sensorsand Actua- tors,2009,22(2) (,, [J].GeomaticsandInformation Scienceof Wuhan. RSSI [J]. University,2009,34(11) (,,2009,22(2) ),,. [11]ShinB,LeeJ H,LeeT.Enhanced Weighted - [J]. NearestNeighborAlgorithmforIndoor WiFiPosi-

6 tioningsystems[c].the8thinternationalconfer- ence on Computing Technology and Information Management(ICCM),Seoul,orea,2012 [12]Zhang Minghua,ZhangShensheng,CaoJian.Re- WirelessLANs[J].ComputerScience,2007,34 (6)68-71 (,. [J].,2007,34 (6)68-71) ceived Signal Strength Based Indoor Locationin AConstrainedNNIndoorPositioningModelBasedona GeometricClusteringFingerprintingTechnique LIU Chunyan 1 WANG Jian 1 1 eylaboratoryforlandenvironmentanddisastermonitoring,chinauniversityofminingandtechnology,xuzhou221116,china AbstractThecommonalgorithmsbasedonRSSIpresentlyavailableareunstableintheindoorenvi- ronment.hence,aconstrained NN positioningalgorithm withgeometricalinformationviacluste- ringfingerprintsisproposedtoresolvethisissue.firstly,thegeometricclusteringfingerprintsare builtaccordingtothestructurallayout.then,theconceptofgeometricstrengthofsporadic (GSS) forasamplepoint sgeometrycharacterizationisintroduced.thevalueofgssisusedforidentifying therpcontrolnetworkstructureinwhichthemobileterminalislocatedtodynamicalychoosethekey parameterfornn.whenthenearestpoint(np)isdecided,anoptimalpolygonconstraintcondi- tionisconstructedtochoosethelater(-1)neighbourpoints.itcanbesummarizedasaconstrain- ednnindoorlocalization modelbasedonageometricclusteringfingerprintingtechnique.there- sultsofseriesoftestsindicatethatthenewalgorithmcan moreefectivelyestimatethelocationofa mobileterminal.clusteredfingerprintsplaysakeyroleinimprovingthepositionaccuracy,thusthe impactofthisnew NNalgorithmshouldnotbeoverlooked. eywordsindoorpositioning;clusteringfingerprint;geometrycharacterization;rpcontrolnetwork structure;constrainednn FirstauthorLIU Chunyan,postgraduate,specializesinindoorandtheundergroundwirelesssensorpositioningsystem. lcy sia@ 163.com FoundationsupportTheNationalHighTechnologyResearchandDevelopmentProgramofChina(863Program),No.2013AA12A201; thepriorityacademicprogram DevelopmentofJiangsu HigherEducationInstitutions(PAPD);theFundamentalResearchFundsforthe CentralUniversities,No.2013RC16;theNew CenturyExcelentTalentsinUniversity,No.NCET

40 3 :GPS ~127 GPS,T 16%,,S 1 S 2 S ,3 UNAVCO % [5] 2 GPS, Tab.2 StatisticsforNon-integer-hourDataBlocks ingpsbroadcaste

40 3 :GPS ~127 GPS,T 16%,,S 1 S 2 S ,3 UNAVCO % [5] 2 GPS, Tab.2 StatisticsforNon-integer-hourDataBlocks ingpsbroadcaste 40 3 2015 3 GeomaticsandInformationScienceofWuhanUniversity Vol.40No.3 March2015 DOI:10.13203/j.whugis20130593 :1671-8860(2015)03-0372-07 GPS 1 1,2 1,2,3 1,,430079 2,,430079 3,,541004 :, GPS ;, GPS, ;,