High-resolution satellite images provide extensive spectral

shape

and textural information of ground objects; as such

these images have been widely used in many fields. As by products of images

shadows affect the visual interpretation and automatic identification of landscape objects. Nevertheless

shadows reveal additional useful information

such as shape

height

surface characteristics

and relative position of targets. Therefore

studies should be conducted to develop methods for detecting shadows.To extract shadows accurately

researchers should consider the necessary pre-conditions; such methods should also be readily available for further utilization. This study proposed a shadow detection method based on object-oriented method and established characteristic components for high-resolution satellite images. To analyze the spectral characteristics of shadows

we determined several components

such as color invariant C3

brightness I

first principal component（ PC1）

and RATIOb

n

ir. We then used these components to highlight shadow areas in images. RATIOb

n

irindex

which can be used to distinguish shadow and water efficiently

was d eveloped by considering the Rayleigh scattering of different wavelengths in shadow and non-shadow areas. However

such c omponents are difficult to analyze comprehensively because different construction methods have revealed various value ranges. To overcome this problem

we used a linear normalization method and transformed the pixel values of images in the same range of 0 to1. Object-oriented method

which comprised segmentation and information extraction

was used to extract shadow areas in the enhanced images. Brightness I and PC1

which contained relatively clear boundary information

were chosen as the main data source for multi-resolution segmentation based on the characteristics of high-resolution images. C3and RATIOb

n

irindices were also used the main data source in the subsequent classification. Several characteristics

such as mean value

maximum difference

standard deviation

area

and gray-level co-occurrence matrix

indicated the difference between shadow and non-shadow objects; as such

these characteristics were selected to extract shadow areas from images. Shadows in 20 QuickBird images were extracted using the proposed method and two contrast experiments. Data revealed that the average total accuracy of the proposed method was 97%

the average producer accuracy was 96%

and the average Kappa index was 0. 94. The combined characteristic component-based and object-oriented methods could be used to obtain shadows with perfect shapes but without fragmentation compared with pixel-based method. The combined method also exhibited higher accuracy than the object-oriented method based on original optical i mages alone. Considerable experiments and statistically high-precision results of the proposed method showed that the combined characteristic component-based method and object-oriented method could be used efficiently to enhance the contrast of shadows a gainst other features. Furthermore

this combined method could be used to ensure the complete extraction of shadow areas. The proposed method could be applied not only to QuickBird imagery used as test data in this study but also to other high-resolution satellite images.