The Use of Landscape Metrics and Transfer Learning to Explore Urban Villages in China

Abstract

Urban villages (UVs), the main settlements of rural migrant workers and low-income groups in metropolitan areas of China, have become of major concern to city managers and researchers due to the rapid urbanization in recent years. A clear understanding of their evolution and spatial relationships with the city is of great importance to policy formulation, implementation and assessment. In this paper, we propose a new framework based on landscape metrics and transfer learning for the long-term monitoring and analysis of UVs, and we apply it to Shenzhen and Wuhan, two metropolitan cities of China, with high-resolution satellite images acquired from 2003-2012 and 2009-2015, respectively. In the framework, landscape metrics are used for identifying the UVs and quantifying their evolution patterns on the basis of a city-UV-building hierarchical landscape model. Transfer learning is also introduced to use the samples and features across the spatial and temporal domains, which reduces the time and labor cost, as well as improves the mapping accuracies by 3-10%. The results show that the total area of UVs has decreased by less than 6% in Shenzhen and more than 45% in Wuhan. Moreover, we observe significant spatial correlations in the development of UVs in Shenzhen. By contrast, no strong spatial correlations are found in Wuhan's UVs, indicating that their development is largely independent of the spatial location. The results reveal two typical strategies, i.e., demolition and renovation, towards the redevelopment of UVs in China.

Urban villages as transfer stations for dengue fever epidemic: A case study in the Guangzhou, China

“…Urban space has undergone dramatic transformation and reconstruction during China’s rapid urbanization, which has caused a large number of rural villages that were originally on the edge of the city to be gradually surrounded or semi-enclosed by urban land [16–20]. A lack of overall planning and scientific management of UVs has resulted in a large number of irregular buildings scattered in urban areas, with subsequent poor sanitation, lack of infrastructure and serious environmental pollution [21, 22]. These characteristics of UVs, combined with their perennial humidity and relatively low temperature [23, 24], provides an ideal living environment for the breeding of Aedes albopictus , the sole vector of dengue transmission in Guangzhou.…”

WorldView-2 Data for Hierarchical Object-Based Urban Land Cover Classification in Kigali: Integrating Rule-Based Approach with Urban Density and Greenness Indices

“…Previous studies investigated various methods for improving image classification accuracy including: (i) the use of a knowledge-based and rule-based approach [19,23,24]; (ii) integrating original spectral bands with texture feature measures [10,[25][26][27]; (iii) combining original image bands with transformed bands using principle components analysis (PCA) and/or intensity-hue-saturation (HIS) [28,29]; (iv) classifying generated segments using hybrid approaches [30], and (v) the fusion of multispectral images with synthetic aperture radar (SAR) images [19,24]. Other studies proposed either the application of edge detection algorithms [31,32] for accurately depicting linear features such as road networks, coupling landscape metrics and transfer learning as a framework for urban dynamics monitoring [33] or performing multi-scale hierarchical classification in high spectral dimensional feature space [34,35]. Machine learning methods have been further applied, specifically for assessing the potential of high-resolution data in producing accurate urban land cover information.…”

The Scope of Earth-Observation to Improve the Consistency of the SDG Slum Indicator

“…Without knowing the best feature combination in characterizing the heterogeneous configuration of deprived areas at the city scale, one could either choose features through trial-and-error or include a full set of empirically effective features to make the choice less arbitrary [53]. Thus, mapping may be conducted in a poorly designed feature space and subject to the curse of dimensionality [56], reducing the generalization capability and transferability of feature sets [57]. Feature selection allows reducing dimensionality, working with the most significant features, and it improves accuracy [58]; however, it still depends on the initial conceptualization of deprived living conditions and features fed into the feature selection approach.…”