A Fuzzy Background Modeling Approach for Motion Detection in Dynamic Backgrounds - Multimedia and Signal Processing - page 183

Table 2. The statistica al results of GMM, T2-FGMM and the proposed method

FPR PBC F-measure

GMM 0.75 1 11.387% 0.937

T2-FGMM 0.03 9 2.251% 0.988

Our method 0.034 4 1.510% 0.992

Frame 2315 Frame 2444 Frame 2535 Frame 2728 Frame 2887

Input

Ground
truth

GMM

T2-
FGMM

Our
method

ult of “overpass”, the 1 st row is the input frames, the 2 nd row is
he GMM's output, the 4 th row is the T2-FGMM's output, the

Fig. 1. The representative resu
ground truth, the 3 rd row is th
row is our method's output

s the
e 5 th

5.2 Water Rippling

We use the “canoe” seque
Table. 3 are the experime
method can perform bette
evaluations.

nce to test the water rippling situation. Fig. 2, Fig. 3
ental results. Similar to the waving trees situation,
er than GMM and T2FGMM in both local and ove

and
our
erall

Table 3. The statistica al results of GMM, T2-FGMM and the proposed method

FPR PBC F-measure

GMM 0.90 5 33.219% 0.792

T2-FGMM 0.044 4 2.968% 0.985

Our method 0.03 2 2.397%

0.988