• Optics and Precision Engineering
  • Vol. 32, Issue 4, 578 (2024)
Ning WANG1, Wenxing BAO1,*, Kewen QU1,*, and Wei FENG2
Author Affiliations
  • 1School of Computer Science and Engineering, North Minzu University, Yinchuan75002, China
  • 2School of Electronic Engineering, Xidian University,Xi'an710071, China
  • show less
    DOI: 10.37188/OPE.20243204.0578 Cite this Article
    Ning WANG, Wenxing BAO, Kewen QU, Wei FENG. Hyperspectral unmixing with shared endmember variability in homogeneous region[J]. Optics and Precision Engineering, 2024, 32(4): 578 Copy Citation Text show less

    Abstract

    Due to different lighting conditions, complex atmospheric conditions and other factors, the spectral signatures of the same endmembers show visible differences at different locations in the image, a phenomenon known as spectral variability of endmembers. In fairly large scenarios, the variability can be large, but within moderately localised homogeneous regions, the variability tends to be small. The perturbed linear mixing model (PLMM) can mitigate the adverse effects caused by endmember variability during the unmixing process, but is less capable of handling the variability caused by scaling utility. For this reason, this paper improved the perturbed linear mixing model by introducing scaling factors to deal with the variability caused by the scaling utility, and used a super-pixel segmentation algorithm to delineate locally homogeneous regions, and then designed an algorithm of Shared Endmember Variability in Unmixing (SEVU). Compared with algorithms such as perturbed linear mixing model, extended linear mixing model (ELMM), and other algorithms. The proposed SEVU algorithm was optimal in terms of mean Endmember Spectral Angular Distance (mSAD) and abundance Root Mean Square Error (aRMSE) on the synthetic dataset with 0.085 5 and 0.056 2, respectively. mSAD is optimal on the Jasper Ridge and Cuprite real datasets with 0.060 3 and 0.100 3, respectively. Experimental results on a synthetic dataset and two real datasets verify the effectiveness of the SEVU algorithm.
    yn=Man+rn    s.t. an0K,k=1Kak,n=1(1)

    View in Article

    yn=(M+dMn)an+rn(2)

    View in Article

    yn=MSnan+rn(3)

    View in Article

    yn=(MSn+dMn)an+rn(4)

    View in Article

    s.t. an0K,anT1K=1,Sn0K×K,M0L×K.MSn+dMn0L×K,n=1,,N.(5)

    View in Article

    f(A,M,dM,S)=12n=1Nyn-(MSn+dMn)an22+αΦ(A)+βΨ(M)+γϒ(dM)+ηΓ(S).(6)

    View in Article

    Φ(A)=12n=1Nan1(7)

    View in Article

    Ψ(M)=12l=1Lk=1Km˜l(-IK+ek1KT)22(8)

    View in Article

    ϒ(dM)=12n=1NdMnF2(9)

    View in Article

    Γ(S)=12n=1NSn-IF2(10)

    View in Article

    Di,c=1-λSIDxi,xc×tanSAM(xi,xc)+λix-cx2+iy-cy2,(11)

    View in Article

    SAM(xi,xc)=cos-1xi,xcxixc(12)

    View in Article

    SIDxi,xc=l=1L(pil-pcl)logpilpcl(13)

    View in Article

    IK1KTQan+-IK0KTRWn=0K1KTJ(14)

    View in Article

    L=12yn-(MSn+dMn)an22+μn(A)2Qan+RWn(A)+Vn(A)-JF2+μn(A)2an-Tn(A)+Gn(A)F2+α2Tn(A)1+lR+(Wn(A)),(15)

    View in Article

    an=(MSn+dMn)T(MSn+dMn)+μn(A)(QTQ+IK)-1(MSn+dMn)Tyn+μn(A)QT(J-RWn(A)-Vn(A))+Tn(A)-Gn(A),(16)

    View in Article

    WnA=maxan+Vn(A),0K(17)

    View in Article

    Tn(A)=softan+Gn(A),αμn(A)(18)

    View in Article

    VnA=VnA+Qan+RWn(A)-J(19)

    View in Article

    Gn(A)=Gn(A)+an-Tn(A)(20)

    View in Article

    Sn=Sn,1000Sn,2000Sn,KPn=Sn,1Sn,2Sn,K(21)

    View in Article

    L=12y˜l-m˜l(PA)-δ˜l22+μl(M)2m˜l-Wl(M)+Vl(M)F2+β2k=1Km˜lGk22+lR+(Wl(M))+n=1NlR+(Wl(M)Sn+dMn)(22)

    View in Article

    m˜l=y˜l-δ˜lPAT+μl(M)Wl(M)-Vl(M)PAPAT+βk=1KGkGkT+μl(M)IK-1(23)

    View in Article

    wl,kM=maxm˜l+Vl(M)k,0,max-dM1,l,kS1,k,,-dMN,l,kSN,k(24)

    View in Article

    Vl(M)=Vl(M)+m˜l-Wl(M)(25)

    View in Article

    L=12yn-(MSn+dMn)an22+μn(dM)2MSn+dMn-Wn(dM)+Vn(dM)F2+γ2dMnF2+lR+(Wn(dM)).(26)

    View in Article

    dMn=(yn-MSnan)anT+μn(dM)(Wn(dM)-Vn(dM)-MSn)ananT+(γ+μn(dM))IK-1.(27)

    View in Article

    wn,kdM=maxMSn+dMn+Vn(dM)k,0K(28)

    View in Article

    Vn(dM)=Vn(dM)+MSn+dMn-Wn(dM)(29)

    View in Article

    L=12yn-(MSn+dMn)an22+μn(S)2Sn-Wn(S)+Vn(S)F2+η2Wn(S)-IF2+lR+(Wn(S))+lR+(MWn(S)+dMn).(30)

    View in Article

    μn(S)MTM-1Sn+SnananT=MTM-1MTynanT-MTdMnananT+μn(S)(Wn(S)-Vn(S)).(31)

    View in Article

    wn,k(S)=maxη+μn(S)(Sn,k+Vn,k(S))μn(S)+η,0,max-dMn,1,kM1,k,,-dMn,L,kML,k(32)

    View in Article

    Vn(S)=Vn(S)+Sn-Wn(S)(33)

    View in Article

    aRMSE=1Nn=1N1Pk=1Kak,n-a^k,n2(34)

    View in Article

    mSAD=1N1Kn=1Nk=1Karccosmn,kTm^n,kmn,kTm^n,k(35)

    View in Article

    yRMSE=1Nn=1N1Ll=1Lyl,n-y^l,n2(36)

    View in Article

    Ning WANG, Wenxing BAO, Kewen QU, Wei FENG. Hyperspectral unmixing with shared endmember variability in homogeneous region[J]. Optics and Precision Engineering, 2024, 32(4): 578
    Download Citation