Information Technology Reference
In-Depth Information
Ta b l e 6
Average statistical evaluation result of five crucians body texture
Evaluation
a
b
c
d
e
10
−
2
]
ASM[
×
0.141
0.084
0.108
0.118
0.173
CNT
77
103
62
76
47
CRR
0.836
0.900
0.904
0.803
0.862
SSQ
258
521
326
196
174
IDM
0.182
0.128
0.152
0.144
0.191
SAV
242
251
146
356
310
SVA
953
1984
1227
706
649
SEP
4.66
5.13
4.91
4.64
4.55
EPY
6.79
7.23
7.01
6.93
6.62
DVA
40
41
24
31
20
DEP
2.81
3.05
2.81
2.91
2.65
Ta b l e 7
Average statistical evaluation result of three kinds of fishes
Evaluation
Black bass
Crap
Crucian
ASM[
×
10
−
3
]
0.8
∼
0.5
0.9
∼
0.6
1.7
∼
0.80
CNT
2626
∼
1129
551
∼
88
103
∼
47
CRR
0.534
∼
0.254
0.892
∼
0.835
0.903
∼
0.802
1849
∼
840
1765
∼
343
521
∼
174
SSQ
IDM[
×
10
−
1
]
0.42
∼
0.24
1.08
∼
0.63
1.90
∼
1.28
SAV
415
∼
291
310
∼
102
1984
∼
649
SVA
5121
∼
2225
6659
∼
1281
1984
∼
649
SEP
5.58
∼
5.20
5.68
∼
4.93
5.13
∼
4.54
EPY
7.65
∼
7.40
7.56
∼
7.12
7.22
∼
6.62
DVA
1087
∼
430
222
∼
30
41
∼
20
DEP
4.61
∼
4.21
3.84
∼
2.97
3.05
∼
2.65
From Table 7, the CNT value of black bass is more than 1000 which is larger
than crap and crucian. CRR value is smaller than 0.6 which is little than the others.
Furthermore, the DEV, DEP value of black bass is larger than crap and crucian,
and IDM is small. We also can combine some of the parameters for a more stable
identification. Or if we only need to separate the black bass from crucian, the SSQ,
SVA, SEP and EPY also can be used.
5Con lu ion
In this study, we propose a specific method of fish species by co-occurrence matrix
and AdaBoost, it was verified in a ideal condition. The fish area detection learning
file for AdaBoost is made originally, and it is work well. For identify the target
species from others, the feature amount of fish body texture is calculated from co-
occurrence matrix. And by setting a threshold value, it is possible to identify the
Search WWH ::
Custom Search