52 1 2012 1 JournalofDalianUniversityofTechnology Vol52 No1 Jan2 0 1 2 :1000-8608(2012)01-0132-07 * ( 116024 ) : Dempster CCD 5 ; ; ; Dempster D-S : ; ;D-S ; :TP24 :A 0 UGV [10] (UGV) UGV UGV Dempster [1] Mathies 1 11 [2] [3] CCD 2D 3D UGV [4~7] CCD 2D 1 [5] 2 UGV 3D 2 Θ [8] Brenneke [9] Buluswar :2010-04-10; :2011-12-03 : (WY-YY/M-200818JY001) : * (1979-) E-mail:zhaoyibing005@163com
1 : 133 1 Fig1 Fusionsystemusedforcapturingimage (1) f1 ( ) 3 : [182310001]; [015] [009]; 0 2 Fig2 Registeringinterfaceoffusionsystem Θ D-S Θ (a) A m(a) 2 Θ [01] : m( )=0 (1) 烆 m(a)= 1 A 2 Θ m(a) A A ; D-S Θ N (N ) N 4 (c) 3 f1 Θ = {stonetrunkshrubwater} Fig3 Curvegraphoff1 3 (b) (2) f2 5 Θ 4 trunk 90%
134 52 (3) f3 [00235 00328]; [0049 6 00762]; [0016800319]; [0009000228] (4) f4 (b) 4 [0452 0757]; [05210786]; [03460435]; [04970945] (5) f5 HSV S [040] H [3060]; S [1050] H [3050]; S [ 10 60]H [30140]; S [010]H [60230] S H 4 m ij [11] (a) (c) (d) 4 f4 stone Fig4 Curvegraphoff4 trunk shrub water 5 fi(i=12 5) U = {u 1 u 2 u n } m ; C fi (i=12 m)c fi 5 E ic fi u j C fi μ ( C u )= 1 C fi E i = {u j P i (j)j (j = 1234); μ ( C u ) 1} m i (j) 1; μ C fi )=1 X Ei )= (2) fi 烆 0; E i C fi e i [01] e i = (e i1 e ij e in ) e ij
1 : 135 = X Ei )m E e 1 烌 E = e 2 (3) 烆 em 烎 m C fi M fc M fc = (m fc1 m fci m fcm ) T m fci C fi m fci = (m i1 m in ) T 烆 m ij =μ ( C u ) m ij (4) (1) μ C fi )=1 m ij =1 m ij = t 1 n S u k ) t= eik e ik = E i C fi =1 k=1 ;S u k ) E i u j u k S u k )= 1-d u k )/d max d u k ) E i u j u k d max E i m ij M fc = (P i (j))= u j U C fi 0250 1 1 1 0250 烌 (2) m ij 0016 0141 0391 1 1 m fci j m ij 1 1 0391 1 1 m ij = mij 2 烆 0391 1 1 0391 0 141 烎 C fi E i P i (j) i j m fci 5 6 U 4 C fi [12] N ;M 5 : U = { f1 =7825f2 =0f3 =00294 αi = max {P i (j) j=12 N} f4 =0638f5 = S =48 烆 H =39} ζ i = Nλi N ;i=12 M U u j P i (j) j=1 Θ 4-1 βi =ζi C fi M fc u j C f1 M -1 ; M 2i=12 M C f2 C f3 C f4 C fi R i = λiαi β i ;i=12 M M u j C fi u j λiαi β i 烆 C fi e ij =μ ( i=1 C u )=1; e ij = μ ( C u )=0 C fi 0 1 1 1 0 烌 m i (j)= 0 0 0 1 1 E = 1 1 0 1 1 烆 0 1 1 0 0 烎 5 Fig5 Colorimageforshrub 6 Fig6 Laserimageforshrub m ij ;λi i ; : (5) i j P i (j) N P i (j)+n(1-r i )(1-λiαi β i) j=1 (6)
136 52 i (8) K 0763 3 Θ i N(1-R i )(1-λiαi m(θ)= β i) N P i (j)+n(1-r i )(1-λiαi β i) 1 j=1 3 (7) 1 5 2 1 m(a)= K -1 m 1 (B i )m 2 (C j );A 5 Dempster K =1- m 1 (B i )m 2 (C j ) (8) ijb i C j = Dempster 5 1 : fi 4 (stone)=0217 (trunk)=0122 Dempster m f1 f 2 f 3 f 4 f 5 (shrub)=0416 [13] (water)=0177 (Θ)=0068 Dempster 4 (1) ; 1 m(obj) : 1 m(obj)= max {m(a)a Θ}= m f1 f 0 2 f 3 f 4 f 5 (shrub)=0416 Obj K -1 i jb i C j =A 1 Tab1 Basicprobabilityassignmentfunction f1 f2 f3 f4 f5 stone m1(1)=0091 m2(1)=0184 m3(1)=0205 m4(1)=0173 m5(1)=0059 trunk m1(2)=0006 m2(2)=0026 m3(2)=0080 m4(2)=0173 m5(2)=0236 shrub m1(3)=0362 m2(3)=0184 m3(3)=0080 m4(3)=0173 m5(3)=0236 water m1(4)=0141 m2(4)=0184 m3(4)=0205 m4(4)=0067 m5(4)=0033 Θ m1(θ)=0400 m2(θ)=0422 m3(θ)=0430 m4(θ)=0415 m5(θ)=0435 2 5 Dempster Tab2 Dempsterfusionresultsbasedonfivekindsoffeatures m f5(stone)=0188f5(trunk)=0188f5(shrub)=0188f5(water)=0012 f5(θ)=0425 f1 f2 f3 f4(stone)=0257 m(stone)=0025 m( )=0 m( )=0 m( )=0 m(stone)=0183 f1 f2 f3 f4(trunk)=0078 m( )=0 m(trunk)=0030 m( )=0 m( )=0 m(trunk)=0055 f1 f2 f3 f4(shrub)=0345 m( )=0 m( )=0 m(shrub)=0133 m( )=0 m(shrub)=0246 f1 f2 f3 f4(water)=0224 m( )=0 m( )=0 m( )=0 m(water)=0013 m(water)=0159 f1 f2 f3 f4(θ)=0096 m(stone)=0009 m(trunk)=0037 m(shrub)=0037 m(water)=0005 m(θ)=0068
1 : 137 : (2) λ1 =025 (shrub)- (trunk)=0294 (3) m i (Θ) λ2 =015 (Θ)=0068<015 (4) m i (Θ) : m(obj)=0416 (Θ)=0068 4 7 shrub Fig7 Shrubidentificationresults UGV UGV 040 m UGV 040 mugv ; UGV UGV 8 ; UGV Fig8 Trunkidentificationresults ; UGV UGV UGV ( ) Dempster UGV 7 8 5 6 5 [1] : [J] 2004(7):6-10 [2]MATTHIES L KELLY A LITWIN Tetal Obstacledetectionforunmannedgroundvehicles:A
138 52 progressreport [C]// Proceedings of the 1995 Inteligent Vehicles SymposiumPiscataway:IEEE [8]HANCOCK J A Laser intensity based obstacle detectionand tracking [D]Pitsburgh:Carnegie 1995:66-71 [3] [9]BRENNEKE C WAGNER B A scan based [D] : 2007 [4] robots in man-made environments [C ] // [D] : 2002 International Conference of Systems Engineering [5] [J] (ICSE)Coventry:[sn]2003:88-93 ( )200535(s2):71-74 MelonUniversity1999 navigationsystemforautonomousoperationofmobile [10]BULUSWARS DDRAPER B AColor machine [6]MACEDO J MATTHIES L MANDUCHI R Ladar-baseddiscriminationofgrassfromobstaclesfor autonomous navigation [M ]// Lecture Notesin ControlandInformationSciencesLondon:Springer- Verlag2000 [7]HUANGJLEE A BMUMFORD DStatisticsof rangeimages[j]proceedingsoftheieeecomputer Society Conferenceon Computer Vision and Patern Recognition20001:324-331 visionfor autonomous vehicles [J]Engineering ApplicationsofArtificialInteligence199811(2): 245-256 [11] [M]1 : 2006 [12] D-S [J] 2006(4):22-23 [13] [M]1 : 2004 Obstacleidentificationincross-countryenvironment forunmannedgroundvehicle ZHAO Yi-bing * GUO Lie ZHANG Ming-heng LI Lin-hui (StateKeyLaboratoryofStructuralAnalysisforIndustrialEquipment DalianUniversityofTechnologyDalian116024China) Abstract: Aimingattheproblem ofcross-countryenvironmentperception ofunmanned ground vehicledempsterfusionrulesareappliedtoidentifyingobstaclefirstlyfivekindsofrepresentative featuresareselectedbasedonccdandlasersensorsecondlysensordataistransformedtoevidence spaceandtheobstacleidentification membershipiscomputedbyusingfuzzyinterpolative method thencorrelationcoeficientisobtainedthirdlyaccordingtoobstacleidentityandweightcorrelation experimentalformulaisselectedtocomputebasicprobabilityassignmentfunctionfinalybasedon Dempster fusion rulesthe ultimate basic probability assignment function is acquired the identificationanddecision-makingrulesaresettodetermineobstacleclassificationtestresultsshow thegoodrobustnessandreal-timepropertybyusingd-stheorytoidentifyobstacle Key words:unmanned ground vehicles;environment perception;d-s theory ofevidence;basic probabilityassignmentfunction