多尺度分割的高分辨率GIS影像建筑物提取模型

发布时间：2018-06-05 02:20

本文选题：高分辨率影像 + 建筑物提取　；参考：《科技通报》2016年12期

【摘要】：高分辨率GIS影像在建筑物提取时由于高分图像细节丰富也增加了一定的噪声,使得提取精度下降。本文提出了一种基于多尺度分割的高分辨率GIS影像建筑物提取模型。首先结合图像的纹理特征分别对形状信息的权重值和光谱信息的权重值进行设置,并以此进行多尺度分割,然后将纹理、形状、空间特征引入到最近邻监督分类的特征空间中进行分类,接着采用知识规则将分割图像进行细分类,最后采用Sobel算子对建筑物轮廓进行边缘检测。通过仿真实验表明,本文提出的改进算法相比较目前常用的算法具有更高的建筑物提取精度。
[Abstract]:The precision of high-resolution GIS image is reduced because of the rich details of high-resolution GIS image. This paper presents a building extraction model for high resolution GIS images based on multi-scale segmentation. Firstly, combining the texture features of the image, the weight of the shape information and the spectral information are set, and then the texture, the shape, the texture and the shape are segmented at multi-scale. The spatial feature is introduced into the feature space of the nearest neighbor supervised classification, then the segmented image is subdivided by knowledge rules, and the building contour is detected by Sobel operator. The simulation results show that the proposed improved algorithm has higher building extraction accuracy than the commonly used algorithms.
【作者单位】：厦门理工学院土木工程与建筑学院;
【基金】：厦门理工学院高层次人才项目基金(YKJ13030R)
【分类号】：TU198;TP391.41

【相似文献】