Multiscale Geoscene Segmentation for Extracting Urban Functional Zones from VHR Satellite Images

Abstract

Urban functional zones, such as commercial, residential, and industrial zones, are basic units of urban planning, and play an important role in monitoring urbanization. However, historical functional-zone maps are rarely available for cities in developing countries, as traditional urban investigations focus on geographic objects rather than functional zones. Recent studies have sought to extract functional zones automatically from very-high-resolution (VHR) satellite images, and they mainly concentrate on classification techniques, but ignore zone segmentation which delineates functional-zone boundaries and is fundamental to functional-zone analysis. To resolve the issue, this study presents a novel segmentation method, geoscene segmentation, which can identify functional zones at multiple scales by aggregating diverse urban objects considering their features and spatial patterns. In experiments, we applied this method to three Chinese cities-Beijing, Putian, and Zhuhai-and generated detailed functional-zone maps with diverse functional categories. These experimental results indicate our method effectively delineates urban functional zones with VHR imagery; different categories of functional zones extracted by using different scale parameters; and spatial patterns that are more important than the features of individual objects in extracting functional zones. Accordingly, the presented multiscale geoscene segmentation method is important for urban-functional-zone analysis, and can provide valuable data for city planners.

Context-Enabled Extraction of Large-Scale Urban Functional Zones from Very-High-Resolution Images: A Multiscale Segmentation Approach

“…Accordingly, the larger d i,j is, the larger the segmentation scale will be. For example, for an object to be merged, its neighboring object with minimal heterogeneity increase is first found, then the adaptive scale used for these two objects are determined by Equation (8). If their heterogeneity increase is less than the calculated scale S, they will be merged; otherwise, nothing is done for this object and another object will be chosen for further processing.…”

“…Image segmentation techniques should be addressed to generate UFZs directly from VHR images since pixels, image tiles and road blocks are inadequate to represent UFZs [8,[17][18][19][20][21][22]. Existing widely used segmentation methods include object segmentation [23][24][25][26][27][28] and semantic segmentation [29][30][31][32].…”

“…However, the method is only designed to identify different urban area categories rather than segmenting UFZs. Zhang et al [8] presented a multi-level aggregation approach considering both object features and spatial pattern features to achieve geoscene segmentation. However, this approach requires conducting image object segmentation and classification.…”

Residential land extraction from high spatial resolution optical images using multifeature hierarchical method

“…We assume that fðx; yÞ is a digital image, with a size of M × N and gray level of N g . Then the GLCM satisfying a certain spatial relationship is shown in E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 1 ; 1 1 6 ; 7 1 1 pði; jÞ ¼ #½ðx 1 ; y 1 Þ; ðx 2 ; y 2 Þ ∈ M · Njfðx 1 ; y 1 Þ ¼ i; fðx 2 ; y 2 Þ ¼ j: (1) where # denotes the number of elements in the set, p is a matrix with a size of N g × N g , ðx 1 ; y 1 Þ, and ðx 2 ; y 2 Þ are two pixels in the image with gray tones i and j, respectively. If the distance between ðx 1 ; y 1 Þ and ðx 2 ; y 2 Þ is d, and the angle between them and the horizontal axis of the coordinates is θ; then, the GLCM pði; j; d; θÞ of various spacings and angles can be obtained.…”

“…The places where people partake in different socioeconomic activities are divided into different functional zones. 1 Residential land (RL), as a typical functional zone, is the most extensive land use type in cities, and it can reflect the situation of land use to a certain extent. Moreover, it is one of the most transformed land use/land cover (LULC) types in the process of urbanization in China.…”