论文部分内容阅读
本文提出一种跨尺度分类方法,该方法立足于多尺度像斑模型,应用特征构造来实现跨尺度特征的构建,从而将最佳尺度选择问题隐含在特征构造中,不直接进行最佳尺度选择,避免了主观选择尺度的弊端。实验结果证明跨尺度分类方法一方面能减少特征维数空间,另一方面能充分利用尺度之间的纵向信息,较单一尺度分类能更准确地区分地物,提高分类精度。
A cross-scale classification method is proposed in this paper. Based on the multi-scale speckle model, this method uses the feature structure to construct the cross-scale feature, so that the optimal scale selection problem is implicit in the feature structure, Choice, to avoid the shortcomings of the subjective choice scale. The experimental results show that the cross-scale classification method can reduce the feature dimension space on the one hand, and can make full use of the vertical information between the scales, so as to distinguish the features more accurately than the single scale classification, and improve the classification accuracy.