Vision Target Tracker Based on Incremental Dictionary Learning and Global and Local Classification

and Applied Analysis 3 Affine transformation element to b set a = In t+1 Mapping from a a b X = {I1 t+1, I2 t+1, . . . , In t+1} I = {I1 (t+1)(x0,y0), . . . , I 1 (t+1)(xn,yn), . . . , I n (t+1)(xn,yn)} b = {I1 (t+1)(x,y), . . . , I1 (t+1)(x,y)} Figure 2: Affine transformation from X to I. the transfer of the probability state of the affine transformation parameters is obtained, the function of motion model is as follows: P (X t+1 | X t ) = N (X t+1 | X t , σ) , (1) whereN(X t+1 | X t ) is modeled independently by a Gaussian distribution, σ is a covariance diagonal matrix, and the elements of the diagonalmatrix are the variance of each of the affine parameters. {X1 t+1 , X 2 t+1 , . . . , X n t+1 } is a group of affine parameter sets which are randomly generated by function (1), in current frame, and {I1 t+1 , I 2 t+1 , . . . , I n t+1 } is area of the target that may occur (candidate image area) which can be constructed by affine transformation from {X1 t+1 , X 2