相对传统的中低分辨率遥感数据而言,高空间分辨率遥感影像同一地物内部丰富的细节得到表征,空间信息更加丰富,地物的尺寸、形状以及相邻地物的关系得到更好的反映,但其光谱统计特性不如中低分辨率影像稳定,类内光谱差异较大,而传统分类方法仅依据像元的光谱值,因此在高分辨率影像分类中,传统方法往往不能获得好的结果。在此背景下,提出了一种多尺度空间特征融合的分类方法,旨在利用不同尺度的空间邻域特征弥补传统方法的不足。其基本过程是：首先针对不同尺度特点,用小波变换压缩空间邻域特征,并结合支持向量机得到不同尺度下的分类结果,然后根据尺度选择因子为每个像元选择最佳的类别。文中QuickBird和IKONOS影像实验证明该算法能有效提高高分辨率影像解译的精度。

A new classification algorithm for high spatial resolution remotely sensed mi agery is proposed, which integrates neighborhood information ofmultiscale such as 2×2, 4×4, 8×8 and 16×16 window sizes around the central pixe.l In order to compress the information of the multiscale spatial features, a wavelet coefficients fusion algorithm is employed to reduce the dmi ension but retain the spatial information at the same tmi e. After the stage ofmultiscale neighborhood feature extraction, a good tool ofpattern recognition: SVM is employed to process the multiscale features, in this algorithm, four groups of spatial features based on four scales produce four classification maps. And then, these maps, which represent multiscale classification results, are fused by a scale selection parameter. The final fusion map is the result ofmultiscale features classification and shows an obvious adaptability to objects of different scales. Expermi ents ofQuickBird and Ikonos show that the proposed classification algorithm ofmultiscale features fusion can achieve better results and better accuracies than the conventional perpixel multis pectral method.