Page 138 - JCAU-7-2

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

Figure 1. Network image feature quantification framework
Source: Diagram by the authors.
word frequencies based on standard deviation,
intensity value, and typological generalization.
(iv) Image and text two-factor comprehensive analysis:
• By overlaying the first-label type of images with the
word frequency types of texts, the study identified
the spatial cumulative distribution of images.
• This method reflects the focal points and hotspots
of the town image, offering insights into content
emphasis and high-interest areas (Liang & Pan,
2015; Marine, 2017).

3.3. Data collection
Figure 2. Example of Tencent Cloud artificial intelligence (AI) image
identification results The research data consists of two primary types: population
Source: Tencent Cloud AI. data and internet data. The internet data were obtained
from the Bing search engine, which offers higher image
package for the social sciences (SPSS), was then quality and less redundant information compared to other
applied to determine the types and spatial layout search engines available in China. The accompanying
of primary image factors. text information for the images was sourced from news
(iii) Text word frequency analysis: websites, enterprise portals, and new media platforms such
• Textual comments and descriptions as Xiahongshu (小红书; translated as Little Red Book) and
accompanying the images were analyzed using Weibo. Using the directory of central towns published
SPSS, ROSTCM6, and GIS software. by the Guangdong Provincial Government in 2002 as
• The analysis focused on overall text word a reference, this study utilized Python programming to
frequency, co-occurrence values, and top-ranked search for the names of central towns in the PRD through

Volume 7 Issue 2 (2025) 4 https://doi.org/10.36922/jcau.5733