The International Archives of the Photogrammetry. Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
1362
Tablel. Accuracy results for object-based classification with topographic data
Master Landscape Type
No Classification Category
Conifer
Plantation
Evergreen
Broad-leaved
Forest
Deciduous
Broad-leaved
Forest
Bamboo
Grove
Grassland
Wetland
Vegetation
Paddy Field
Bare Ground
Rural
Residential
Urban
Residential
Open Water
Total
Point
Match
Point
User's
Accuracy
1 Conifer Plantation
219
2
66
40
20
1
0
16
4
2
1
371
219
59%
2 Evergreen Broad-leaved Forest
35
32
56
23
6
0
0
6
6
8
0
172
32
19%
3 Deciduous Broad-leaved Forest
64
7
445
35
7
1
0
11
2
2
0
574
445
78%
4 Bamboo Grove
130
0
132
229
8
0
0
4
15
5
0
523
229
44%
5 Grassland
11
1
54
10
144
62
125
12
6
35
6
466
144
31%
6 Wetland Vegetation
5
0
4
2
20
200
240
3
3
0
0
477
200
42%
7 Paddy Field
1
0
2
0
7
20
429
0
1
0
1
461
429
93%
8 Bare Ground
2
2
2
1
11
0
0
281
9
23
0
331
281
85%
9 Rural Residential
79
16
125
54
77
0
0
167
242
174
0
934
242
26%
10 Urban Residential
7
1
6
3
23
0
0
68
24
518
0
650
518
80%
11 Open Water
0
0
0
0
0
0
0
0
0
0
40
40
40
100%
Total
553
61
892
397
323
284
794
568
312
767
48
4999
2779
Produser's Accuracy
40%
52%
50%
58%
45%
70%
54%
49%
78%
68%
83%
*K = Class Kappa Index
Overall Accuracy = 56%
Overall Kappa Index = 0.51
Table2. Accuracy results for object-based classification only (Kamagata et al., 2006)
Master Landscape Type
No Classification Category
Conifer
Plantation
Evergreen
Broad-leaved
Forest
Deciduous
Broad-leaved
Forest
Bamboo
Grove
Grassland
Wetland
Vegetation
Paddy Field
Bare Ground
Rural
Residential
Urban
Residential
Open Water
ft
Match
Point
User's
Accuracy
1 Conifer Plantation
217
2
67
39
24
8
12
13
4
2
4
392
217
55%
2 Evergreen Broad-leaved Forest
31
32
55
22
7
7
7
2
5
5
1
174
32
18%
3 Deciduous Broad-leaved Forest
67
7
448
36
9
7
36
11
2
2
0
625
448
72%
4 Bamboo Grove
130
0
133
232
12
1
22
4
15
5
0
554
232
42%
5 Grassland
8
1
50
5
104
34
48
6
4
23
2
285
104
36%
6 Wetland Vegetation
4
3
8
4
65
173
34
38
16
48
1
394
173
44%
7 Paddy Field
8
0
5
2
5
1
357
31
6
7
0
422
357
85%
8 Bare Ground
2
2
2
1
11
6
93
273
6
17
0
413
273
66%
9 Rural Residential
79
13
123
52
63
42
155
138
234
154
0
1053
234
22%
10 Urban Residential
7
1
2
4
23
5
30
52
20
504
0
648
504
78%
11 Open Water
0
0
0
0
0
0
0
0
0
0
40
40
40
100%
Total
553
61
893
397
323
284
794
568
312
767
48
5000
2614
Produser's Accuracy
39%
52%
50%
58%
32%
61%
45%
48%
75%
66%
83%
*K = Class Kappa Index
Overall Accuracy = 52%
Overall Kappa Index = 0.47
(a) IKONOS
(b) Landscape Map
FigurelO. Comparison of landscape map with
3. RESULTS AND DISCUSSIONS
The image results generated by the two classification methods
are shown in Figure 6 (object-based using topographic data),
and 7 (object-based only); and the classification accuracy for
these two methods are shown respectively in Table 1 and 2.
In terms of overall classification accuracy, the object-based
using topographic data results (56%) scored higher than object-
based alone (52%). In terms of overall Kappa index as well,
object-based with topographic (0.51) outscored the object-based
only (0.47).
The results using only spectral characteristics generated a
producer’s accuracy for Rural Residential of 75%. In contrast,
the user’s accuracy for this same landscape type was only 22%.
This large gap may be attributed to the fact that the Rural
Residential area also included small patches of forest, barren
land and agriculture area, which in terms of spectral
characteristics overlap substantially with other landscape types
such as forest and Bare Ground. Misclassification of Paddy