Identify a Specified Fish Species by the Co-occurrence Matrix and AdaBoost - Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing - page 103

Ta b l e 1 Statistical evaluation result of single black bass body texture

Evaluation a b c d e Average

10 − 3 ] ASM [ × 0.819 0.790 0.824 0.756 0.786 0.795

CNT 2145 2694 2395 2748 2572 2511

CRR 0.319 0.315 0.317 0.325 0.331 0.322

SSQ 1567 1972 1743 2029 1932 1849

10 − 1 ]

IDM [ × 0.448 0.380 0.422 0.448 0.413 0.422

SAV 422 414 418 411 412 415

SVA 4158 5172 4616 5395 5122 4893

SEP 5.34 5.42 5.38 5.45 5.44 5.41

EPY 7.39 7.41 7.39 7.43 7.42 7.4

DVA 927 1140 1053 1197 1119 1087

DEP 4.51 4.63 4.56 4.62 4.59 4.58

Ta b l e 2 Statistical evaluation result of single Carp body texture

Evaluation a b c d e Average

10 − 3 ] ASM [ × 0.657 0.611 0.623 0.614 0.596 0.620

CNT 206 326 290 244 326 278

CRR 836 989 953 1102 956 967

SSQ 1283 1373 1300 1226 1317 1300

10 − 1 ] IDM[ × 0.718 0.600 0.631 0.617 0.594 0.632

SAV 336 290 302 315 307 310

SVA 3146 3646 3522 4203 3501 3604

SEP 5.34 5.40 5.39 5.47 5.39 5.40

EPY 7.43 7.48 7.47 7.48 7.50 7.47

DVA 66 105 93 78 100 88

DEP 3.39 3.61 3.56 3.47 3.60 3.52

peripheral image of the chest fin as shown by the red frame in Fig. 4. for each kind
of fish, the body texture pattern shown in Figures 4-6 are used. From the side around
the chest fin of each image, five pieces of the body texture image of 60 × 40 pixel

was cut from each image.
Tab le 1

3 show the quantified value of co-occurrence matrix for one single
black bass, crap, crucian body texture on different body positions respectively. The
last column is average value of different body position. The value in each column is
relatively stable. this means such parameters can be used for texture evaluation.
Tab le 4

∼

∼

6 show the quantified average value of co-occurrence matrix for each
five black bass, crap, crucian body texture on different body positions respectively.
Some of the parameters show the obviously difference for three kinds of fishes.
Table 7 shows the quantified feature value range of black bass, crap, crucian
respectively. And we can find a available feature value to identify the black bass
from two others.