正在加载图片...
1 WANG Hongyuan,et al:Efficient tracker based on sparse coding with Euclidean local structure-based constraint ·139. Table 1 Optimization for ELS constraint based SC(ELSSC) Input:Given N data points Y=[yY]ER,over-complete dictionary DR) Output:Sparse matrix C=[ce]R(x Parameters:Maxiumn iteration number J=10,neighborhood size K=5,all-zero vector co,a=0.01,B=0.5,y=0.001 1:For each point y,compute the nearest K neighborhoods N(y)and weights 2:Compute the SVD-decomposition of D UJVT,where VER()(2m) 3:compute ID'D‖,and set‖D'D|+2 B randomly 4:For i=1:N 5:For t=1:J 6:If lec)<,break inner iteration 7:0omie0=,6-,0=[Dy+290,-w+(y-28)c-DD,e-"],md0=VeR 8:Repesent”thparse colficient vector”,ie,optimize子l”-lke”tallo"l, 9:End 10:End Equation (5)is the objective function of our Eu- clidean local structure constraint-based SC and can be +7291a-ol-宁Ie-nl店 solved through iterative computation.In particular,at 1 lI3-6:D9〉+aIe0,-2pc9,-)+ the t-th iteration,for a single candidate y:in Y,Eq. (5)can be written as follows: 子"1-g-9e)+Y,81o+ 2 mineof(c()=min Iy:-De+ a ll e :+B lc)-0)(6) 〈De”,D〉-2ID,l3+BI-wI经= where.At the t-th iteration for the optimization of c,c,j is fixed.Therefore,we can 子1e"3-De)-29(c0,g-)- regard 1)as a constant.To solve Eq.(6),we intro- (y-2B)c.co+(Def Dco)+a llel+ duce the following surrogate function as presented in [11]: (兮%店+7291l+811》- 6o)=分1e-6l-71n0-l目 子Ic”-0+aI0,+R (7) (8) where is convex.According to Daubechies,when where AI-D'D is a strictly positive definite matrix,c,co) 1 =[D'y:+280+(y-28)c-D De-] 2 is strictly convex for any co with respect to c.Hence,in and our experiments,the constant A is set accordingly (A y-28;Table 1).Once the over-complete dictionary D is R=lxl+Y291o1-1l+ fixed,we can derive the following convex objective func- tion from Eq.(7): BIwI-子I” )lyDeare fixed at the t-th iteration.Thus,we can simplifyTable 1 Optimization for ELS constraint based SC(ELSSC) Input:Given N data points Y= [ y1… yN ]∈R m×N ,over⁃complete dictionary D∈R m×(n+2m) Output:Sparse matrix C= [c1… cN ]∈R (n+2m)×N Parameters: Maxiumn iteration number J = 10,neighborhood size K= 5,all⁃zero vector c0 ,α= 0.01,β = 0.5,γ = 0.001 1:For each point yi,compute the nearest K neighborhoods NK(yi) and weights wji 2:Compute the SVD⁃decomposition of D = UΣV T ,where V∈R (n+2m)×(n+2m) 3:compute ‖D TD‖, and set‖D TD‖+2β randomly 4:For i = 1:N 5: For t = 1:J 6: If ‖c (t) i - c (t-1) i ‖2 < τ , break inner iteration 7:Compute θ (t) i = ∑j wji c (t-1) j ,v (t) i = 1 γ [D T yi + 2βθi (t-1) + (γ - 2β)ci t-1 - D TDi c t-1 ] ,and x (t) i =Vv (t) i ∈R (n+2m)×1 8: Represent x (t) i with sparse coefficient vector c (t) i ,i.e.,optimize γ 2 ‖x (t) i -Vc (t) i ‖2 2 +α‖c (t) i ‖1 9: End 10:End Equation (5) is the objective function of our Eu⁃ clidean local structure constraint⁃based SC and can be solved through iterative computation. In particular, at the t⁃th iteration, for a single candidate yi in Y, Eq. (5) can be written as follows: minci (t) f(ci (t) ) = minci (t) ‖yi - Dci (t)‖2 2 + α ‖c (t) i ‖1 + β ‖c (t) i - θ (t-1) i ‖2 2 (6) where θ (t) i = ∑ jwji c (t-1) j . At the t⁃th iteration for the optimization of ci,cj,i ≠ j is fixed. Therefore, we can regard θ (t-1) i as a constant. To solve Eq. (6), we intro⁃ duce the following surrogate function as presented in [11]: ψ(ci,c0) = λ 2 ‖c (t) i - c0‖2 2 - 1 2 ‖Dc (t) i - Dc0‖2 2 (7) where λ is convex. According to Daubechies [13] , when λI - D TD is a strictly positive definite matrix, ψ(ci,c0) is strictly convex for any c0 with respect to ci . Hence, in our experiments, the constant λ is set accordingly ( λ = γ - 2β; Table 1). Once the over⁃complete dictionary D is fixed, we can derive the following convex objective func⁃ tion from Eq. (7): f(c (t) i ) = 1 2 ‖yi - Dc (t) i ‖2 2 + α ‖c (t) i ‖1 + β ‖c (t) i - θ (t-1) i ‖2 2 + γ - 2β 2 ‖c (t) i - c0‖2 2 - 1 2 ‖Dc (t) i - Dc0‖2 2 = 1 2 ‖yi‖2 2 - 〈yi,Dc (t) i 〉 + α ‖c (t) i ‖1 - 2β〈c (t) i ,θ (t-1) i 〉 + γ 2 ‖c (t) i ‖2 2 - (γ - 2β)〈c (t) i ,c0〉 + γ - 2β 2 ‖c0‖2 2 + 〈Dc (t) i ,Dc0 〉 - 1 2 ‖Dc0‖2 2 + β ‖θ (t-1) i ‖2 2 = γ 2 ‖c (t) i ‖2 2 - 〈yi,Dc (t) i 〉 - 2β〈c (t) i ,θ (t-1) i 〉 - (γ - 2β)〈c (t) i ,c0〉 + 〈Dc (t) i ,Dc0〉 + α ‖c (t) i ‖1 + ( 1 2 ‖yi‖2 2 + γ - 2β 2 ‖c0‖2 2 + β ‖θ (t-1) i ‖2 2 ) = γ 2 ‖c (t) i - v (t) i ‖2 2 + α ‖c (t) i ‖1 + R (8) where v (t) i = 1 γ D T yi + 2βθ (t-1) i + (γ - 2β)c t-1 i - D TDc t-1 [ ] and R = 1 2 ‖yi‖2 2 + γ - 2β 2 ‖c0‖ - 1 2 ‖Dc0‖2 2 + β ‖θ (t-1) i ‖2 2 - γ 2 ‖v (t) i ‖2 2 are fixed at the t⁃th iteration. Thus, we can simplify 第 1 期 WANG Hongyuan, et al: Efficient tracker based on sparse coding with Euclidean local structure⁃based constraint ·139·
<<向上翻页向下翻页>>
©2008-现在 cucdc.com 高等教育资讯网 版权所有