Page 137 - JCAU-7-2

Journal of Chinese
Architecture and Urbanism Urban features of PRD in online image

image structure. Liu (1994) identified iconic images of areas and smaller townships (Li et al., 2022). They serve
historical and cultural villages, including ancestral halls, as key nodes in China’s township system, bridging the gap
towers, trees, squares, ponds, water bridges, roofs, and between small towns and larger urban centers (Dong &
gables. Fan & Wang (2010) analyzed the rural landscape Gao, 2011).
image of the PRD using questionnaire surveys and cognitive In 2023, the urbanization rate of the PRD region
maps, emphasizing the importance of regional landscape reached 87.5%, making it the most densely populated and
characteristics for the reconstruction of rural images.
economically urbanized region in China (Office of the
In recent years, the influence of mobile internet has Seventh National Population Census Leadership Group
transformed city image research. Perceiving subjects, of the State Council, 2022). The spatial quality of towns
objects, modes, presentations, and feedback mechanisms within the PRD significantly impacts the region’s high-
have evolved, reflecting tendencies toward planarization, quality urban agglomeration development.
symbolization, and digitalization (Wang & Zhen, 2015; Over the years, the composition of central towns in the
Yu & Chen, 2024). Scholars have increasingly studied PRD has undergone notable changes. Compared to the list
network-based imageability through street view analysis of central towns announced by the Guangdong Provincial
and public perception data derived from media platforms.
For example, Zhou et al. (2014) used web image crawling Government in 2002, some towns have expanded into small
to summarize city image types. Zhao et al. (2015) analyzed cities, while others have lagged behind in development.
internet images of 21 cities retrieved from Google. Long & This study identifies central towns with strong imageability
in internet media, offering insights into the general
Zhou (2017) leveraged open internet data to evaluate urban developmental trends of central towns in the PRD.
images, proposing the concept of “picture urbanism.” He &
Li (2021) employed statistical methods, including principal 3.2. Analytical framework and methodology
component analysis and regression, to explore typological
characteristics and relationships between image elements. The readability of town images and texts serves as a
Wang (2023) conducted a comparative study of internet foundation for understanding the spatial perception of
images of Changsha from official websites and social towns. As American poet Ezra Pound observed, “Image
networking platforms, exploring the metaphorical is the complex of thoughts and feelings in a moment”
ideologies behind the images. (Pound, 1986, p. 152). To address the dynamic nature of
urban images, which evolve over time and space, this study
Despite advancements in urban image research, most collected images and accompanying text content associated
studies focus on singular data types, such as cognitive maps, with town names from internet search engines over a single
street view recognition, or internet texts. Such approaches day in 2024. This approach aims to provide an objective basis
risk incomplete or biased evaluations. As information for analyzing town images from an internet perspective.
technology continues to advance, integrating multiple data
types is imperative for comprehensive analyses. In this This study seeks to construct a quantitative framework
study, we analyzed images and texts of central towns in the for analyzing network image features by evaluating their
PRD retrieved from internet search engines. This approach characteristics and commonalities. The methodology
highlights similarities and differences between small-town involves the following steps (Figure 1 for the network
and city image elements. Our findings will provide valuable image feature quantification framework):
insights into perceptions of town landscape styles, identify (i) Overall images and text recognition:
landscape style types under network environments, explore • A brightness histogram was utilized to analyze
the development potential of small towns, and contribute overall image brightness, reflecting disparities
to establishing positive town images. between urban and rural image content for each
town.
3. Overview of PRD areas and research • Semantic network analysis was utilized to analyze
methodology co-occurring word pairs associated with towns,
highlighting public attention and focus areas.
3.1. Overview of central towns in the PRD (ii) Image type recognition:
Central towns refer to towns within smaller settlements • Images for each town were analyzed using
that possess advantageous locations and robust economic Tencent Cloud AI image recognition (Figure 2).
strength. These towns benefit from the economic and • After manually categorizing and recording the
infrastructural influence of nearby large- and medium- semantic labels for each image, the top picture
sized cities while also playing a significant role in gathering labels for each town were counted. Principal
resources and radiating development to surrounding rural factor analysis, conducted using a statistical

Volume 7 Issue 2 (2025) 3 https://doi.org/10.36922/jcau.5733