基于多源遥感图像和复杂网络的城市主城区识别方法

Main Urban District Identification Method Based on Multi-source Remote Sensing Images and Complex Network

针对现有的采用遥感图像的城市主城区提取方法仅利用数据光谱特征,未考虑城市的空间结构特征的情况,本文提出了一种基于多源遥感图像和复杂网络模型的城市主城区提取新方法。首先,采用八邻域极值法识别夜光图像的特征点,通过最佳路径成本法识别特征点的连接边,从而构建夜光图像复杂网络;其次,利用特征向量中心度改进复杂网络模型核心边缘结构识别方法,弥补其无法顾及节点邻域特征的不足;最后,依托分形网络演化算法获取Landsat-8图像的图像对象,构建夜光图像复杂网络节点与图像对象的空间映射关系,实现城市主城区提取。本文以沈阳、成都、西安3个城市为研究区域,以2019年的NPP/VIIRS夜光图像和Landsat-8图像为数据源,实验结果表明本文识别方法的总体精度为0.880 0,Kappa系数为0.741,较传统方法分别提高了0.119 8和0.255 9。

The existing methods of main urban district extraction using remote sensing images only use the spectral characteristics of the data,without considering the spatial structure characteristics of the city.Therefore,this paper proposes a new method of main urban district extraction based on multi-source remote sensing images and complex network model.Firstly,the feature points of nighttime light images are identified by the eight-neighborhood extremum method,and the connection edges of feature points are identified by the optimal path cost method,so as to construct the complex network of nighttime light images.Then,the core-edge structure recognition method of complex network model is improved by using eigenvector centrality to optimize the deficiency that it can not take into account node neighborhood features.Finally,the image object of Landsat-8 image was obtained by using fractal network evolutionary algorithm,and the spatial mapping relationship between nodes and image object of complex network of nighttime light image was constructed to realize the extraction of main urban district.In this paper,Shenyang,Chengdu and Xi′an were selected as the research area,and NPP/VIIRS nighttime light images and Landsat 8 images of 2019 were used as data sources.Experimental results show that the overall accuracy of the proposed method is 0.880 0,and the Kappa coefficient is 0.741,which is 0.119 8 and 0.255 9 higher than the traditional method,respectively.