论文部分内容阅读
高分辨率遥感影像在不同的尺度下表现出不同的特征,根据这一特性,本文提出了一种基于多层次特征的航空影像规则建筑物提取的新方法。该方法先从利用大尺度特征—方向梯度直方图(histogramsoforientedgradient,HOG)特征对建筑物进行识别,然后提出一种小尺度特征—光谱和纹理融合特征,该特征能够有效的将HOG特征识别结果中的道路、草地等非建筑物剔除,最终获取建筑物边缘信息。实验结果表明该方法不仅对矩形建筑物有较好的提取效果,同时对结构复杂的规则建筑物也有较好的提取效果。
According to this characteristic, high resolution remote sensing images show different characteristics at different scales. In this paper, a new method for extracting the regular aerial buildings based on multi-level features is proposed. The method firstly identifies the building by using the features of large scale histogramsoforiengragradient (HOG), and then presents a small-scale feature-spectrum and texture fusion feature that can effectively identify HOG features As a result, non-buildings, such as roads and grasslands, are eliminated and the edge of the building is finally captured. The experimental results show that this method not only has a good extraction effect for rectangular buildings, but also has a good extraction effect for the regular buildings with complicated structures.