Remote sensed imagery with very high resolution (VHR) contains substantial spatial information. Multi-scale segmentation algorithm was usded to construct the hierarchical system of objects. The relationships between all features, such as spectral, texture, topography and shape,were analyzed based on canonical correlation analysis (CCA). A difference image was developed by using multivariate alteration detection technology (MAD). The propsed object-oriented double-windows flexible pace searching (OB-DFPS) method achieved accuracy of 93.2%, with Kappa coefficiant of 0.862. After compared with traditional pixel-based method, the proposed OB-DFPS method was more robust. 