In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
resolution (Blue, Green, Red and NIR) and a panchromatic 
band with 0.61m resolution. In this study, the multispectral and 
panchromatic images were fused to produce a four-band 
pan-sharpened multispectral image with pixel size of 0.61m. 
The image fusion was carried out using the Gram-Schmidt 
procedure (Laben and Brower, 2000). A subset of the 
pan-sharpened multispectral image with size of 1800 x 2800 
pixels was finally used in the study (Figure 1). The image 
subset covers a portion of the suburban area. The land cover 
types in the area include tree, grassland, soil and impervious 
surface (building, road). 
Figure 1 Quickbird image of study area (Bands 3, 4, 2 as R, G, B) 
4. RESULTS AND DISCUSSION 
The impervious surface was extracted using the proposed 
method and the method based on traditional SVM, respectively. 
Table 1 shows the impervious surface mapping results. From 
the table, both overall accuracy and Kappa coefficient of the 
proposed method are higher than those of the method using the 
SVM. The producer’s accuracy and user’s accuracy of the 
impervious surface are also acceptable. In particular, the higher 
user’s accuracy of the impervious surface indicates that the 
proposed method produced less commission error that the 
existing method using the SVM. From Figure 1, although the 
results from two methods showed very similar appearance, the 
result from the proposed method are more homogeneous inside 
the class. 
Table 1 Impervious surface mapping results using different methods (all in %) 
OA 
Kappa 
Accuracies for impervious surface 
PA 
UA 
Multi-class 
SVM 
83.93 
73.20 
91.68 
79.34 
OCSVM 
88.56 
77.27 
81.87 
96.51 
OA, Overall accuracy; PA, producer’s accuracy; UA, user’s accuracy.