In High Spatial Resolution Remote Sensing（ HSRRS） images

targets in the same class have different shapes. The d escription at one scale or one template is inadequate to describe target shapes from the same class. In this study

a multiscale shape model based on wavelet transform and Fourier descriptors is constructed. A new object-oriented method for target recognition in HSRRS images is also developed. The model uses wavelet approximation coefficients with successively decreasing scale to represent the target shape from top to bottom approximately. Approximate shapes are described quantitatively using Fourier d escriptors.The final recognition results are obtained using the semantic rule to synthesize recognition results at multiple scales. This method can reduce the effect arising from the broken objects segmented at a small scale and the underidentification of small objects at a large scale. Aircrafts and buildings in HSRRS images are identified

and the comparison results show that the method proposed in this paper has higher identification accuracy.