Multi-Source Data and Semantic Segmentation: Spatial Quality Assessment and Enhancement Strategies for Jinan Mingfu City from a Tourist Perception Perspective

Abstract

In the context of cultural tourism integration, tourists’ spatial perception intention is an important carrier of spatial evaluation. In historic cultural districts represented by Jinan Mingfu City, tourists’ perceptual depth remains underexplored, leading to a misalignment between cultural tourism development and spatial quality needs. Taking Jinan Mingfu City as a representative case of a historic cultural district, while the living heritage model has revitalized local economies, the absence of a tourist perspective has resulted in misalignment between cultural tourism development and spatial quality requirements. This study establishes a technical framework encompassing “data crawling-factor aggregation-human-machine collaborative optimization”. It integrates Python web crawlers, SnowNLP sentiment analysis, and TF-IDF text mining technologies to extract physical elements; constructs a three-dimensional evaluation framework of “visual perception-spatial comfort-cultural experience” through SPSS principal component analysis; and quantifies physical element indicators such as green vision rate and signboard clutter index through street view semantic segmentation (OneFormer framework). A synergistic mechanism of machine scoring and manual double-blind scoring is adopted for correlation analysis to determine the impact degree of indicators and optimization strategies. This study identified that indicators such as green vision rate, shading facility coverage, and street enclosure ratio significantly influence tourist evaluations, with a severe deficiency in cultural spaces. Accordingly, it proposes targeted strategies, including visual landscape optimization, facility layout adjustment, and cultural scenario implementation. By breaking away from traditional qualitative evaluation paradigms, this study provides data-based support for the spatial quality enhancement of historic districts, thereby enabling the transformation of these areas from experience-oriented protection to data-driven intelligent renewal and promoting the sustainable development of cultural tourism.

1. Introduction

Under the context of deep cultural tourism integration, the spatial experience quality of cultural destinations has emerged as a critical determinant of tourist satisfaction [1]. Tourist perceptual imagery at destinations profoundly influences travel decision-making, consumption motivation, cultural identity formation, and recommendation intention [2]. This dynamic process, driven by perceived images and ultimately condensed into experience quality, is of vital importance to the development strategy of urban tourism [3]. Driven by the policy of “Cultivating Tourism through Culture and Showcasing Culture through Tourism,” Jinan Mingfu City—as a National 4A-grade Scenic Area and pivotal cultural landmark of the “Spring City” [4]—has seen a remarkable enhancement in its cultural and tourism appeal. The city’s annual tourist volume surged from 64 million to 116 million visits within a three-year period [5]. Notably, the Mingfu City–Daming Lake precinct, covering only 0.42 km², accommodates 17.2% of the city’s total tourist flow [6], with its single-day reception during the 2025 May Day holiday reaching 3.93 million tourists [7]. This exponential growth in visitation has precipitated massive volumes of real-time perception data, establishing an unprecedented empirical foundation for systematically analyzing tourists’ perceptual imagery towards spatial environments, cultural atmospheres, and service facilities [8]. Current conservation policies for historical blocks predominantly emphasize physical fabric and scholarly perspectives (as seen in research by Lv [9] and Zhao [10]), yet lack a quantitative evaluation and response mechanism for spatial quality based on tourist perception data, this study focuses on the tourist perspective to explore the optimization path of spatial quality in historical blocks.

Tourists, as the core experiencers and evaluation subjects of cultural tourism value in historical blocks [11], have spatial perception serving as a key indicator for measuring the effectiveness of block vitalization [12]. In the field of theoretical research on tourists’ spatial perception, Urry’s [13] ‘The Tourist Gaze’ theory emphasizes the initiative of tourist interpretation; Ai et al. [14] established a tourist perception evaluation system; Ma [15] explored the relationship between tourism experience and tourist satisfaction; Stankov, U. et al. [16] conducted empirical research further proving that social media comments can effectively capture tourists’ embodied perceptions. In terms of technological application, text mining and sentiment analysis have emerged as core methodologies for analyzing tourist-generated content (TGC), demonstrating empirical efficacy in representing collective tourist perceptions. The evolution of text mining technology has traversed critical developmental phases: Manning, C.D. et al. [17] established the theoretical foundations for structured data extraction from textual sources and analyzed interactive models between users and web interfaces; Hearst, M. A. [18] promoted the evolution from “script-based crawlers” to “systematic crawlers”; Rilof, E. et al. [19] pioneered foundational research; Hatzivassiloglou, V. et al. [20] initiated the construction of core sentiment lexicons and affective classification frameworks. The field concurrently witnessed expanding applications, exemplified by Alonso, M. A. et al. [21] analyzing the polarity–intensity dimensions of disinformation effects. Domestically, Li et al. [22] employed SnowNLP for Weibo comment sentiment analysis, while Peng et al. [23] integrated sentiment analytics with LDA modeling to construct a Sentiment-Aware Collaborative Filtering algorithm. These established technical frameworks have been extensively deployed across domains, including destination image perception [12], service quality bench-marking, and experience value identification [14], thereby furnishing robust analytical instruments for capturing tourist perceptions.

In quantitative spatial quality assessments of historical blocks, multi-source data fusion methodologies—spanning Space Syntax and social media sentiment data—have catalyzed a paradigm shift toward quantified research frameworks. For example, Karimi, K. [24] coupled spatial syntax (integration and choice) with social media sentiment data, revealing the impact of street topological structures on tourist pleasure. Maniei, H. et al. [25] demonstrated enhanced pedestrian space safety and sociability potential in traditional alleyway circuits through integrated Space Syntax and behavioral mapping. Concurrently, the integration of streetscape imagery and machine learning technologies has pioneered novel pathways for quantifying spatial quality. For example, Ye et al. [26] used machine learning to calculate the greenery visibility from street view images and conducted an integrated analysis with street accessibility. Wang et al. [27] combined image data with deep learning techniques to establish the relationship between street view elements and perceptual evaluation values. Wang [28] constructed a PSPNet-based semantic segmentation model, combining POI data to quantify street cultural perception and achieve an objective evaluation of spatial quality. Huang et al. [29] measured the correlation between green vision rate and public satisfaction through street view semantic segmentation, within quantitative metrics of spatial quality. Dai et al. [30] proposed a composite index of “green vision rate–street openness–interface enclosure” (weighted 4:3:3), which comprehensively evaluates comfort and optimizes facility layout by integrating tourist review data. Zheng et al. [31] analyzed streetscape data using AI and proposed a green visual index, sky openness index, and color atmosphere index, which are used to quantify the visual coordination between the natural and artificial environment of historic districts. Gu [32] focused on the cultural dimension, introducing the Heritage Conservation Impact Degree and Cultural Element Diversity index. Fang et al. [33] analyzed the impact of historic district texture on street aesthetics through indicators such as Street Height–Width Ratio and Store Signage Count. Han et al. [34] adopted indicators, including Visual Entropy and the Color Richness Index, to assess street comfort in historic cultural districts.

Collectively, these studies converge on an integrated paradigm, synthesizing tourist perception analytics, computational data processing, and multi-dimensional spatial performance indicators. This corpus provides critical theoretical grounding, methodological frameworks, and technical infrastructures for the evidence-based optimization of spatial quality in historical blocks through the tourist perspective.

Two persistent limitations remain: First, standardized analytical frameworks for multi-source data synergy are underdeveloped. Prevailing studies predominantly employ singular data sources, failing to establish correlative analytics between affective semantics and physical spatial elements. Affective semantic data often remain confined to qualitative descriptions, exhibiting insufficient integration with quantified metrics such as the Green View Index and Interface Enclosure Ratio. Second, the collaborative optimization mechanism for multi-source data is not yet mature, where technical tools such as Space Syntax and street view semantic segmentation remain weakly coupled with tourist evaluation frameworks, impeding precise identification of spatial quality thresholds. Addressing these gaps, this research adopts Jinan Mingfu City as an empirical case study. Targeting its high-traffic yet suboptimal experiential outcomes, it integrates Python web crawlers, SnowNLP sentiment analysis, and OneFormer semantic segmentation into a “sentiment element extraction-physical space quantification-indicators impact analysis” workflow, establishing pixel-level associations between tourists’ affective orientations towards material elements and 150 physical elements, thereby resolving the subject–object data dichotomy prevalent in conventional research. A double-blind scoring mechanism involving urban planners and tourists, validated via Spearman’s rank correlation analysis, to establish statistically validated key performance indicators across three dimensions: visual perception–spatial comfort–cultural experience, providing operational quantitative standards for spatial quality evaluation. By synthesizing multi-source data and quantitative analysis, this study constructs a living heritage-aligned spatial quality evaluation system, offering a scientific basis for the sustainable development of historic districts under cultural tourism integration and fostering dynamic equilibrium among protection, development, and tourist needs.

2. Materials and Methods

2.1. Spatial Characteristics and Cultural Heritage Distribution of Jinan Mingfu City

Jinan Mingfu City, situated in Lixia District, Jinan City, Shandong Province, constitutes a vital component of Jinan’s historic urban core and serves as a key area of Jinan’s historical and cultural city. Enclosed by the Moat and Daming Lake, it covers a total area of 3.2 km² [35]. The city features two quintessential historical and cultural blocks: Jiang Junmiao historical blocks and Fu Rong Street–Bai Huazhou historical blocks. These blocks preserve the majority of the district’s heritage buildings, spring channels, and cultural landscapes, retaining the architectural features and spatial configurations of the Ming and Qing dynasties (Figure 1).

• Fu Rong Street–Bai Huazhou historical blocks (bounded by Quancheng Road to the south, Daminghu Road to the north, Zhenchi Street to the east, and Gong Yuan qianggen Street to the west) exemplify Jinan’s traditional commercial culture and folk customs, hosting clusters of time-honored eateries, handicraft stores, and cultural and creative boutiques. Baihuazhou, with its natural landscapes and historical and cultural heritage, has become a premier destination for tourists to experience Jinan’s spring water culture and historic urban ambiance.
• Jiang Junmiao historical blocks (delimited by Quancheng Road to the south, Daminghu Road to the north, Bianzhi Lane to the east, and Taiping Temple Street and Xichenggen Street to the west) are distinguished by their religious heritage and historical architecture, including landmarks such as the Guandi Temple and Fu Xue Confucian Temple. These ancient buildings not only showcase the artistic style of Jinan’s ancient architecture but also reflect the city’s profound historical and cultural heritage.

To comprehensively document the historical and cultural resources of Mingfu City, a historical resource database was constructed. The database encompasses distribution information of heritage conservation units and intangible cultural heritage. Through collecting and organizing relevant materials, detailed records of cultural relics protection units within Mingfu City were meticulously documented. These records include attributes such as name, location, historical period, architectural style, and other pertinent information. Meanwhile, intangible cultural heritage sites were cataloged, with metadata recorded on aspects such as name and type. Spatial analysis of this dataset enables a clear understanding of the distribution patterns and conservation status of heritage resources, providing a scientific basis for subsequent research and conservation work (Figure 2).

• Heritage architecture: A total of 276 traditional buildings, including 11 provincial-level conservation units such as Fu Xue Confucian Temple, Chen Mian’s Zhuangyuan Mansion, and 30 municipal-level conservation units such as Lao She’s Former Residence, Hou Zaimen Christian Church;
• Intangible cultural heritage sites: National-level intangible cultural heritage sites such as Jinan Shadow Puppetry Theater, Lu Embroidery, Cloisonné Workshops, and the birthplace of Qushui Liushang culture;
• Characteristic spaces: The spring–alley–courtyard spatial fabric, such as Furong Spring, Wangfu Pool, Ming-Qing dynasty streets and alleys, such as Qushuiting Street, Jiang Junmiao Street, and spring water folk activities, such as the Lotus Festival and Dragon Lantern Fair.

2.2. Research Framework

This study employs Python 3.8.7 web crawlers to collect 9000 tourist reviews (2019–2023) from platforms including Mafengwo, Xiaohongshu, and Ctrip. Emotional word segments were extracted via the SnowNLP sentiment analysis module. A sentiment intensity matrix was generated through −3 to 3 standardization processing. This quantifies the public’s emotional tendencies (positive/negative/neutral) toward high-frequency physical elements in the block, such as spring water, historic streets, and architecture. Concurrently, TF-IDF text mining identifies core semantic features. Through SPSS principal component analysis (KMO = 0.85, Bartlett’s sphericity test p < 0.001) with maximum variance rotation, the vocabulary was merged into three dimensions: visual perception, spatial comfort, and cultural experience. Building on literature reviews and expert consultations, an evaluation framework was formulated:

• Visual dimension: Prioritizes aesthetic coherence;
• Comfort dimension: Focuses on spatial accessibility and facility convenience;
• Cultural dimension: Emphasizes historical symbol preservation and narrative immersion.

An initial index pool containing 10 observed variables was constructed, with calculation algorithms further derived for each indicator. Based on Space Syntax theory, DepthmapX 10-0.6.0 software was used to calculate the integration value and connectivity value of 36 main streets in Mingfu City. Combined with Heritage Visibility Density (HVD), five typical streets were selected as study samples. Using ArcMap, 71 sampling points were set at 30 m intervals to systematically capture Baidu Street View panoramic images. The OneFormer semantic segmentation framework (pre-trained on the ADE20K dataset) was employed for pixel-level analysis, quantifying proportional distributions of 150 physical elements, such as buildings, vegetation, and signage. This approach transforms abstract concepts, such as the signboard clutter index and street enclosure ratio, into computable metrics, overcoming the subjectivity inherent in manual surveys.

To ensure the objectivity and reliability of evaluation results, a human–machine collaborative evaluation mechanism was implemented: Automated scoring via semantic analysis and element proportion algorithms, and double-blind human evaluations by experts and tourists. Through Spearman rank correlation analysis, the quantitative correlations between machine indicators and manual scores were revealed (Figure 3).

3. Data and Methodology

3.1. Street Potential Assessment and Screening Based on Space Syntax

To obtain a detailed block network map of the study area, the research first utilized OSM (OpenStreetMap, https://www.openstreetmap.org, accessed on 17 December 2024) road network data based on 2024 China’s administrative divisions to acquire the initial road network of Jinan Mingfu City. Subsequently, the initial road network was refined using historical district conservation blueprints released by the Jinan Municipal Bureau of Resources, and a CAD road network map was drafted using AutoCAD 2014 software. Considering the small-scale characteristics of streets in historical blocks, this study defined road intersections as the basis for street segmentation. After completing the CAD road network map, it was imported into DepthmapX software for further analysis (Figure 4 and Figure 5).

In the geographic research progress of spatial syntax, Luo [36] and Tao Wei explored the relationship between the built environment and the pedestrian path selection behavior, which indicates that the shortest path and the topological path are the two most discussed indicators. Consequently, the integration value and connectivity value were selected as analytical indicators within the Space Syntax framework. The integration value reflects the accessibility of a space within a region by analyzing the topological relationships and connectivity patterns among spatial elements [37]. It serves as an indicator of spatial centrality and aggregation [38]. The connectivity value refers to the frequency of the shortest topological distance connecting two independent nodes in the spatial system [39]. It represents the collection of the shortest topological distances from any element to others in the system, indicating the likelihood of a spatial unit being traversed by other shortest paths [40].

Regarding the determination of high and low value ranges for indicators, scholars’ criteria vary. Typically, based on the color assignment in DepthmapX visualization results, warm colors (red/yellow) represent high values, while cool colors (blue) represent low values [39,41,42,43]. Some studies set percentage thresholds (e.g., top 25% [44], 20% [45], or 10% [46]), but there is no unified standard. More research papers adopt the mean value as the criterion for distinguishing high and low levels [47,48]. In the numerical calculations of this study area, the values are relatively small, making the mean method suitable as the basis for judgment. Furthermore, due to the differing numerical scales of integration value and connectivity value, normalization processing (to the [0–1] interval) was performed, resulting in a mean integration value of 0.5631 and a mean connectivity value of 0.4482.

By evaluating two key spatial syntax metrics—the integration value and connectivity value—the streets of Mingfu City were classified into four categories:

• High-efficiency streets (In > 0.5631, Cn > 0.4482): High integration value and connectivity value, serving as the core streets with significant cultural tourism potential.
• Potential streets (In > 0.5631, Cn ≤ 0.4482): High connectivity value but low integration value. These streets could be transformed into culturally valuable corridors through targeted improvements.
• Isolated streets (In ≤ 0.5631, Cn ≤ 0.4482): Low integration value and connectivity value, largely disconnected from the main street network.
• Transitional streets (In ≤ 0.5631, Cn > 0.4482): Intermediate integration value and connectivity value, exhibiting moderate cultural tourism potential.

In historic districts, the planning of internal road networks requires a balanced approach that harmonizes heritage conservation with spatial–functional coordination. Simply pursuing topological indicators such as integration value and connectivity value from Space Syntax is not applicable to such special regions. Due to the presence of numerous protected heritage assets within the block, solely pursuing high integration values and connectivity values may lead to heritage damage. Therefore, by overlaying the Heritage Visibility Density (HVD), the streets were further classified for conservation, enabling the proposal of targeted protection measures. The street potential assessment methodology based on Space Syntax integrates the integration value and connectivity value through spatial overlay analysis (Figure 6). Combined with Heritage Visibility Density (HVD), to construct a Composite Potential Index (CPI). Five typical streets were subsequently identified for in-depth analysis.

The HVD is defined as the number of heritage assets (historic buildings, inscriptions, etc.) within every 100 m of the visual field, which is identified and counted via semantic segmentation of heritage-related labels (

Table 1. Heritage Visibility Density of key streets.

Street Name	Heritage Visibility Density	Street Length
Furong Street	5.2 sites/100 m	Approx. 420 m
Bai Huazhou Street	4.5 sites/100 m	Approx. 300 m
Jiang Junmiao Street	3.8 sites/100 m	Approx. 280 m
Hou Zaimen Street	1.5 sites/100 m	Approx. 200 m
Shuang Zhongci Street	1.2 sites/100 m	Approx. 150 m
Dong Huaqiangzi Street	2.0 sites/100 m	Approx. 250 m
…	…	…

). (See details in the Published List of Provincial-Level Cultural Relics Protection Units in Shandong Province http://whhly.shandong.gov.cn/art/2023/6/8/art_315161_10323830.html, accessed on 23 December 2024 and the Directory of Municipal-Level Cultural Relics Protection Units in Jinan City https://data.sd.gov.cn/portal/catalog/d132a73610c74f2495146964ee0e721f, accessed on 25 December 2024).

$$\mathrm{Heritage} \mathrm{Visibility} \mathrm{Density}={\displaystyle \frac{\mathrm{Number} \mathrm{of} \mathrm{Cultural} \mathrm{Heritage} \mathrm{Sites}}{\mathrm{View} \mathrm{Length} (\mathrm{in} \mathrm{hundreds} \mathrm{of} \mathrm{meters})}}$$

The CPI is calculated through weighted integration value (In), connectivity value (Cn), and the HVD using the formula:

CPI = 0.4 × In + 0.3 × Cn + 0.3 × HVD

Based on the overlay analysis of the CPI, Space Syntax metrics (integration value and connectivity value), and the HVD, five streets (Qu Shuiting Street, Shuang Zhongci Street, Jiang Junmiao Street, Hou Zaimen Street, and Shengfu Front Street) were selected as typical samples for further analysis. These streets encompass the primary street typologies of high-efficiency, isolated, and potential types (with no transitional streets included in these historic districts). They exhibit significant variations across CPI values (ranging from 0.327 to 2.067), heritage density (0 to 5.2 sites/100 m), and spatial topological characteristics. This diversity ensures a comprehensive reflection of the distinct spatial features and problem types present in the street network of the study area, thereby guaranteeing the typicality and persuasiveness of the case selection. Furthermore, considering the workload associated with field surveys and multi-source data collection, selecting five streets enhances the operational feasibility of this study while maintaining representativeness. The specific characteristics of these streets are detailed in

Table 2. Composite Potential Index of streets.

Street	Comprehensive Potential Index (CPI)	In	Cn	HVD (Sites/100 m)	Typological Characteristics
Qu Shuiting Street (High-efficiency)	2.067	0.846 (High)	0.561 (High)	5.2 (High)	Core cultural tourism corridors
Jiang Junmiao Street (Isolated)	1.411	0.5310 (Low)	0.194 (Low)	3.8 (High)	Conflict between heritage conservation and accessibility
Shuang Zhongci Street (High-efficiency)	0.858	0.797 (High)	0.598 (High)	1.2 (Low)	Local cultural deficiency
Hou Zaimen Street (Potential)	0.784	0.596 (High)	0.318 (Low)	1.5 (Low)	Marginal revitalization blind spots
Shengfu Front Street (Potential)	0.327	0.789 (High)	0.037 (Low)	0	Traffic-Induced deprivation of heritage space

.

3.2. Street View Image Data Acquisition

In conducting street view data collection for the five selected streets, the sampling point configuration critically determines data representativeness and analytical accuracy. For high-efficiency streets, a 30 m equidistant sampling interval was implemented. For isolated streets with HVD, a dense sampling strategy was adopted to holistically capture spatial characteristics and support data-driven interventions for enhancing tourism attractiveness. This dual-strategy framework yielded 71 sampling points across the study area.

Leveraging the Baidu Map API endpoint, Python crawlers systematically retrieved Baidu Street View images, each embedded with metadata including size, location, heading, and pitch. At each sampling point, street view images were captured in four directions (0°, 90°, 180°, and 270°) (Figure 7 and Figure 8), resulting in 284 validated street view images for subsequent analysis.

3.3. Public Perception Element Extraction

With the rapid development of Web 3.0 [49], social media platforms have become a crucial channel for tourists to express experiential perceptions [50]. Research has proven that TGC from travel review websites holds significant value for the exploration and analysis of destination image perception [51]. Most scholars prioritize platforms such as Ctrip, Qunar, and Mafengwo as key sources for user reviews [3,52,53,54,55,56]. To investigate tourists’ perceptions and experiences in Jinan Mingfu City, a multi-source data acquisition system was set up, encompassing major cultural tourism social platforms such as Xiaohongshu, Qunar, Mafengwo, and Weibo. Utilizing Python web crawlers and the Octopus 8 software for targeted and collected review data, 8985 raw data entries were collected, including user profiles, textual descriptions, image links, and rating details. During the data preprocessing phase, duplicate reviews (127 entries), invalid data (215 entries unrelated to Mingfu City), and low-quality texts (389 entries with fewer than 10 characters) were sequentially removed to form the foundational corpus.

The research referred to the spatial element classification system of Jinan Mingfu City proposed by Li [57] and combined the collation of historic district material environmental components by Du [58], Zhao [59], etc., to summarize a material element keyword repository specific to Mingfu City, covering 136 core terms such as “spring water”, “park”, “lakeside”, “lotus”, and “architecture”. Early applications of SnowNLP sentiment analysis used a scoring range of 0–1 [60,61]. As technology advanced, an increasing number of studies began to further refine the granularity of sentiment scores. Scholars such as He et al. [53] and Luo et al. [56] utilized the sentiment analysis functionality of RostCM6, introducing three intensity levels—neutral, moderate, and high—within the categories of positive and negative sentiment. Zheng [62] categorized sentiment into evaluative positivity, evaluative attitude, evaluative intensity, and evaluative focus in her thesis. To achieve more specific emotional tendencies for physical elements, the Python code was enhanced to incorporate extremely strong emotional biases, implemented through adverbial–noun combinations, assigning sentiment scores of +3 and −3.

In the sentiment semantic analysis phase, a Python sentiment sentence processing model (Material_Environment_Classifier) was first developed to filter 15,648 valid sentiment text segments. Subsequently, a dual-dimensional sentiment polarity–intensity model was constructed using natural language processing (NLP) toolkits (NLTK [63] + SnowNLP). This involved the following steps:

• Sentence segmentation and tokenization of preprocessed texts.

Sentiment lexicon tagging based on the HowNet Sentiment Lexicon to identify affective lexemes.

• Weighted scoring algorithm assigning sentiment scores for each review:
  Positive: +3 to +1
  Negative: −1 to −3
  Neutral: 0

A theme-element coupling extraction method was then employed, utilizing the TF-IDF algorithm to perform term frequency statistics on material element keywords. Low-frequency irrelevant terms were filtered using mutual information (MI) and the chi-square test (χ²) [64]. Through a cross-analysis of sentiment scores and element term frequencies, the correlation strength between material elements and emotional tendencies in tourist experiences was revealed (

Table 3. Sentiment statement filtering and scoring.

Sentiment Segment	Sentiment Score
Walking on Furong Street felt like being completely pushed along by the crowd—it was packed with people everywhere.	−3
Furong Street is Jinan’s characteristic old street and food street, named after the famous spring “Furong Spring” on the street.	3
Bordered by Mashi Street to the east, connecting to Qi Fengqiao Street, Xiangfeng Lane, and Furong Lane; adjacent to Yuhuanquan Street to the west, leading to Shengfu East Street.	0
After admiring Qifeng Bridge, you can wander through the surrounding alleys to experience the charm of old Jinan.	0
As a tourism educator and nature enthusiast, I personally believe that artificial attractions compromising natural landscapes are counterproductive.	−1
The Qushuiting, Baihuazhou, and Mingfu City area south of the road represents Jinan’s spring water cultural characteristics and is the most worthwhile section to visit.	1

and

Table 4. Word Frequency Statistics and Sentiment Scores of Physical Elements.

Physical Element	Word Frequency	Sentiment Score
Spring Water	870	0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 3, 3, 3, 3, 0, 0, 3, 3, 3, 3, 3, 0, 0, …
Lakeside	615	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 1, 0, 0, 0, …
History	455	0, 0, −1, −1, 0, 0, 0, 0, 1, 1, 1, 0, 0, 1, 0, 0, 1, 3, 3, 1, 0, 0, 1, 1, 1, 0, 0, …
Ancient Street	267	1, 0, 1, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1, 1, 1, …
Decorated Archway	15	1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, −1
Balustrade	3	0, 1, 1
Temple of Literature	239	0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, −1, 0, −1, 1, 0, 0, 1, 1, …
Lao She	223	0, 0, 0, 0, 0, 1, 0, 3, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
…	…	…

).

4. Results

4.1. Development of a Three-Dimensional Indicator Framework

Drawing upon the research dimensions established by scholars in the spatial quality conservation of historic districts, the 136 physical elements were categorized and processed. Principal component analysis (PCA) in SPSS was employed to extract key features. For instance, Zhang [65] emphasized the need for a synergistic relationship among physical spatial morphology, functional adaptability, and cultural value continuity in historic urban conservation. Li et al. [66] highlighted the significance of the “natural landscapes–historic streets and alleys–cultural activities” triad in their study of the Ciqikou Historic District. Wang, Y. et al. [67] based on cultural landscape gene theory, analyzed the perceptual differences between tourists and residents across the dimensions of “material landscapes–human actions–place meanings.” Jiang, S. Y. et al. [68] utilized street view image data to evaluate historic district spatial quality from the three dimensions of “vitality, safety, and landscape.” Building on the research of these scholars, core variables—including total word frequency, mean sentiment score, positive frequency, negative frequency, and sentiment variance—were extracted from the 136 physical element terms derived from word frequency statistics and sentiment scoring, thereby constructing a multi-dimensional dataset.

A multi-dimensional dataset was constructed by extracting core variables—including total word frequency, mean sentiment score, positive/negative frequency, and sentiment variance—from the 136 material element keywords derived through word frequency statistics and sentiment scoring (

Table 5. Core variables in principal component analysis (PCA).

Physical Element	Total Word Frequency	Mean Sentiment Score	Positive Frequency	Negative Frequency	Sentiment Variance
Spring Water	870	0.36	196	30	0.7933199771055678
Characteristic	637	1.42	459	4	0.9097815458958181
Lakeside	615	0.28	143	18	0.6894500687623522
Park	582	0.28	125	39	0.7774605564306287
Lotus	520	0.24	112	12	0.578430722378848
History	455	0.31	117	14	0.657227382560551
Culture	405	0.39	114	6	0.7095662231009212
Tourism	403	0.36	117	12	0.7068955872188105
Quancheng Square	392	0.23	68	15	0.660073757758151
Ancient Street	267	0.56	120	6	0.74885764957086
…	…	…	…	…	…

). Principal component analysis (PCA) was subsequently conducted using SPSS 25 software [69], following these standardized procedures: First, the original variables were standardized to eliminate dimensional differences; the suitability of the data for factor analysis was confirmed through the KMO test (KMO = 0.82) and Bartlett’s sphericity test [70] (p < 0.001); then, the principal component method was used to extract common factors with eigenvalues greater than 1, resulting in three principal components with a cumulative variance explanation rate of 68.72%; finally, varimax rotation was applied to obtain a clear factor loading matrix.

Based on the rotated factor loading matrix, variables were categorized according to their loading coefficients (absolute value > 0.4) across principal components (

Table 6. Principal component factor scores.

Visual Perception Elements	Factor 1	Spatial Comfort Elements	Factor 2	Cultural Experience Elements	Factor 3
Spring Water	5.03821	Lively Atmosphere	2.27207	Signboards	2.88051
Park	3.99731	Trail	2.12973	Huixianlou Restaurant	2.05298
Lakeside	3.30701	Hutong	2.00042	Listen to Crosstalk	1.66994
Lotus	2.52627	Stalls	1.85029	Ancient Trees	1.65663
Architecture	0.78775	Bustling	1.75446	Former Residence of Lu Dahuang	1.65663
Pearl Spring	0.62592	Affordable	1.6487	Folk Performance	1.65663
Photography	0.88237	Ancient Street	1.23603	Market	1.65663
Flowers and Plants	−0.5027	Stone Bench	−1.28178	Historical Site	1.65663
Blue Sky	−0.48927	Bench	−1.2777	Cultural and Artistic	1.19729
Trees	0.44884	Shade of Trees	−1.2777	Former Residence of Lao She	1.03973
Stream	0.44837	Leisure	1.02609	Storefront	0.99678
Pond	−0.40417	Stroll	1.00449	Teahouse	0.95303
Open Space	−0.39282	Quiet	0.97863	Lotus Festival	0.84724
Advertisement	−0.51878	Decorated Archway	0.96219	Archway	−1.29944
Street Lamps	−0.49488	Sculpture	−1.27906	Qifeng Bridge	−0.81796

).

The first principal component (Factor 1) was characterized by significantly higher loading values from visual-related elements such as “spring water” (5.038), “park” (3.997), “lotus” (2.526), and “architecture” (0.787). These variables collectively reflect the visual attributes of the physical environment, including color, form, vegetation, and water bodies. This dimension is classified as the “visual perception” dimension.

The second principal component (Factor 2) emphasized spatial amenity indicators such as “lively atmosphere” (2.272) and “leisure” (1.026), as well as facility-related terms such as “hutong” (2.000), “stalls” (1.850), and “stone bench” (−1.281) highlighting functional convenience, physiological comfort, and pedestrian activity patterns, and these are categorized as the “spatial comfort” dimension.

The third principal component (Factor 3), historical architecture-related terms, such as “Huixianlou Restaurant” (2.052) and “Lao She’s Former Residence” (1.039), exhibit strong correlations with intangible cultural practices such as “listen to crosstalk” (1.669) and “folk performance” (1.656). This cluster focuses on the cultural symbol-bearing and living experience of historical sites, thus being classified as the “cultural experience” dimension.

Based on Li’s “Spring-Alley-Courtyard” contextual framework for Mingfu City [57] and integrating Du “Resident-Tourist Symbiosis” theory [58] with domestic and international research on street space quality, ten core indicators were systematically selected across three dimensions: visual perception, spatial comfort, and cultural experience, with the indicators closely linked to the factors derived from the three principal components (Table 7 and Table 8).

• The visual dimension builds upon Fang, Z.G.’s quantitative research on street aesthetic indicators [33], focusing on the visual attributes of the physical environment impacting tourist experience: green vision rate, sky visibility rate, and water visibility ratio. Additionally, the signboard clutter index was introduced to reflect conflicts between commercial interfaces and historic landscapes. Special terms such as “Spring Water,” “Lakeside,” “Pearl Spring,” and “Stream” in Factor 1 are rarely found in other scholars’ indicators because of the spring city characteristics of Jinan Mingfu City. Therefore, the water visibility ratio was added.
• The spatial comfort dimension is bench-marked against Han, J.W.’s “Street Comfort” model [34], quantifying the effective pedestrian clear width, shading facility coverage, and street enclosure ratio. Terms such as “stone bench” and “bench” in Factor 2 indicate tourists’ emphasis on resting spaces. Thus, the public seating density indicator was added to quantify resting convenience.
• The cultural dimension integrates Gu, W.Q.’s “Cultural Heritage Transmission Pathway” theory [32]. Combining cultural activity spaces such as “listen to crosstalk,” “folk performance,” “Lotus Festival,” and “market” identified in Factor 3, indicators for the cultural activity space ratio and informal vending space occupancy were established, translating “Intangible Cultural Heritage (ICH) Activation” into measurable street vitality indicators.

Specific indicators and computational methods are as follows:

Table 7. Indicator definitions and computational methods.

Dimension	Indicator	Scholar	Conceptual Definition	Algorithmic Formulation
Visual Dimension	Green Vision Rate	Fang, Z.G. [33] Huang, J.X. [34] Zheng, Y. [31] Pei, Y. [71]	Pixel share of vegetation (trees, grass) in the field of view and identification of vegetation areas by semantic segmentation.	$\mathrm{G}\mathrm{r}\mathrm{e}\mathrm{e}\mathrm{n} \mathrm{V}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n} \mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e}=\left({\displaystyle \frac{\mathrm{V}\mathrm{e}\mathrm{g}\mathrm{e}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n} \mathrm{P}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{l} \mathrm{C}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}}{\mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l} \mathrm{P}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{l} \mathrm{C}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}}}\right)\times 100\%$
	Sky Visibility Rate	Fang, Z.G. [33] Yang, J.Y. [72] Han, J.W. [34]	Sky visible area share, reflecting the sense of spatial oppression, and semantic segmentation to extract sky pixels.	$\mathrm{Sky} \mathrm{Visibility} \mathrm{Rate}=\left({\displaystyle \frac{\mathrm{S}\mathrm{k}\mathrm{y} \mathrm{P}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{l} \mathrm{C}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}}{\mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l} \mathrm{P}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{l} \mathrm{C}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}}}\right)\times 100\%$
	Water Visibility Ratio	Fang, Z.G. [33] Xiong, X [73]	The proportion of water bodies, such as springs and rivers, that are visible in the streetscape, and semantic segmentation to recognize water body pixels.	$\mathrm{W}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{V}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y} \mathrm{R}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}=\left({\displaystyle \frac{\mathrm{Water} \mathrm{Body} \mathrm{Pixel} \mathrm{Count}}{\mathrm{Total} \mathrm{Pixel} \mathrm{Count}}}\right)\times 100\%$
	Signboard Clutter Index	Fang, Z.G. [33] He, Z.Y. [74]	Percentage of commercial advertisements with non-historical and cultural elements, semantic segmentation identifies billboard pixels, and calculates density.	$\mathrm{S}\mathrm{i}\mathrm{g}\mathrm{n}\mathrm{b}\mathrm{o}\mathrm{a}\mathrm{r}\mathrm{d} \mathrm{C}\mathrm{l}\mathrm{u}\mathrm{t}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{I}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x}=\left({\displaystyle \frac{\mathrm{A}\mathrm{d}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{t}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{n}\mathrm{g} \mathrm{S}\mathrm{i}\mathrm{g}\mathrm{n} \mathrm{P}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{l} \mathrm{C}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}}{\mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l} \mathrm{P}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{l} \mathrm{C}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}}}\right)\times 100\%$
Comfort Dimension	Effective Pedestrian Clear Width	Han, J.W. [34] Zhang, Y.C. [75]	Percentage of passable width after subtracting obstacles, semantic segmentation identifies occupying facilities (electrical boxes, stalls), and calculates the available space.	$\mathrm{E}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e} \mathrm{P}\mathrm{e}\mathrm{d}\mathrm{e}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{i}\mathrm{a}\mathrm{n} \mathrm{C}\mathrm{l}\mathrm{e}\mathrm{a}\mathrm{r} \mathrm{W}\mathrm{i}\mathrm{d}\mathrm{t}\mathrm{h}=\left({\displaystyle \frac{\mathrm{Usable} \mathrm{Width} \mathrm{Pixel} \mathrm{Count}}{\mathrm{Total} \mathrm{Width} \mathrm{Pixel} \mathrm{Count}}}\right)\times 100\%$
	Shading Facility Coverage	Han, J.W. [74] Feng, X.X. [76]	Percentage of shaded areas, such as pergola, shade, etc., and semantic segmentation to identify shaded areas (shadows + amenities).	$\mathrm{S}\mathrm{h}\mathrm{a}\mathrm{d}\mathrm{i}\mathrm{n}\mathrm{g} \mathrm{F}\mathrm{a}\mathrm{c}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y} \mathrm{C}\mathrm{o}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{g}\mathrm{e}=\left({\displaystyle \frac{\mathrm{Shade} \mathrm{Area} \mathrm{Pixel} \mathrm{Count}}{\mathrm{Total} \mathrm{Pixel} \mathrm{Count}}}\right)\times 100\%$
	Public Seating Density	Han, J.W. [34] Fang, H.Y. [77]	Percentage of street seats, etc., semantic segmentation identifies seats and counts them.	Public Seating Density = $\left({\displaystyle \frac{\mathrm{Number} \mathrm{of} \mathrm{Seats}}{\mathrm{Street} \mathrm{Length} (\mathrm{in} \mathrm{hundreds} \mathrm{of} \mathrm{meters})}}\right)\times 100\%$
	Street Enclosure Ratio	Han, J.W. [34] Fang, Z.G. [33]	Evaluate the degree of interface enclosure and street openness, composed of elements such as walls, buildings, and pavements, affecting the privacy and openness of the space.	$\mathrm{Street} \mathrm{Enclosure} \mathrm{Ratio}=\left({\displaystyle \frac{{\mathrm{P}}_{\mathrm{w}\mathrm{a}\mathrm{l}\mathrm{l}}+{\mathrm{P}}_{\mathrm{b}\mathrm{u}\mathrm{i}\mathrm{l}\mathrm{d}\mathrm{i}\mathrm{n}\mathrm{g}}}{{\mathrm{P}}_{\mathrm{s}\mathrm{i}\mathrm{d}\mathrm{e}\mathrm{w}\mathrm{a}\mathrm{l}\mathrm{k}}}}\right)\times 100\%$
Cultural Dimension	Cultural Activity Space Ratio	Gu, W.Q. [32] Jiang, L. [78]	Proportion of public space available for cultural activities (plazas, stages, bazaars) to the total area of the neighborhood, and semantic segmentation to identify activity spaces.	Cultural Activity Space Ratio = $\left({\displaystyle \frac{\mathrm{Activity} \mathrm{Space} \mathrm{Pixel} \mathrm{Count}}{\mathrm{Total} \mathrm{Pixel} \mathrm{Count}}}\right)\times 100\%$
	Informal Vending Space Occupancy	Gu, W.Q. [32] Fang, Z.G. [33]	Proportion of pavement area occupied by temporary stalls (snacks, handicrafts) in the street, semantic segmentation to identify stall areas.	Informal Vending Space Occupancy = ${\displaystyle \frac{\mathrm{Stall} \mathrm{Area} \mathrm{Pixel} \mathrm{Count}}{\mathrm{Sidewalk} \mathrm{Area} \mathrm{Pixel} \mathrm{Count}}}\times 100\%$

Table 7. Indicator definitions and computational methods.

Dimension	Indicator	Scholar	Conceptual Definition	Algorithmic Formulation
Visual Dimension	Green Vision Rate	Fang, Z.G. [33] Huang, J.X. [34] Zheng, Y. [31] Pei, Y. [71]	Pixel share of vegetation (trees, grass) in the field of view and identification of vegetation areas by semantic segmentation.	$\mathrm{G}\mathrm{r}\mathrm{e}\mathrm{e}\mathrm{n} \mathrm{V}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n} \mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e}=\left({\displaystyle \frac{\mathrm{V}\mathrm{e}\mathrm{g}\mathrm{e}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n} \mathrm{P}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{l} \mathrm{C}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}}{\mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l} \mathrm{P}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{l} \mathrm{C}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}}}\right)\times 100\%$
	Sky Visibility Rate	Fang, Z.G. [33] Yang, J.Y. [72] Han, J.W. [34]	Sky visible area share, reflecting the sense of spatial oppression, and semantic segmentation to extract sky pixels.	$\mathrm{Sky} \mathrm{Visibility} \mathrm{Rate}=\left({\displaystyle \frac{\mathrm{S}\mathrm{k}\mathrm{y} \mathrm{P}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{l} \mathrm{C}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}}{\mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l} \mathrm{P}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{l} \mathrm{C}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}}}\right)\times 100\%$
	Water Visibility Ratio	Fang, Z.G. [33] Xiong, X [73]	The proportion of water bodies, such as springs and rivers, that are visible in the streetscape, and semantic segmentation to recognize water body pixels.	$\mathrm{W}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{V}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y} \mathrm{R}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}=\left({\displaystyle \frac{\mathrm{Water} \mathrm{Body} \mathrm{Pixel} \mathrm{Count}}{\mathrm{Total} \mathrm{Pixel} \mathrm{Count}}}\right)\times 100\%$
	Signboard Clutter Index	Fang, Z.G. [33] He, Z.Y. [74]	Percentage of commercial advertisements with non-historical and cultural elements, semantic segmentation identifies billboard pixels, and calculates density.	$\mathrm{S}\mathrm{i}\mathrm{g}\mathrm{n}\mathrm{b}\mathrm{o}\mathrm{a}\mathrm{r}\mathrm{d} \mathrm{C}\mathrm{l}\mathrm{u}\mathrm{t}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{I}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x}=\left({\displaystyle \frac{\mathrm{A}\mathrm{d}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{t}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{n}\mathrm{g} \mathrm{S}\mathrm{i}\mathrm{g}\mathrm{n} \mathrm{P}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{l} \mathrm{C}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}}{\mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l} \mathrm{P}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{l} \mathrm{C}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}}}\right)\times 100\%$
Comfort Dimension	Effective Pedestrian Clear Width	Han, J.W. [34] Zhang, Y.C. [75]	Percentage of passable width after subtracting obstacles, semantic segmentation identifies occupying facilities (electrical boxes, stalls), and calculates the available space.	$\mathrm{E}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e} \mathrm{P}\mathrm{e}\mathrm{d}\mathrm{e}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{i}\mathrm{a}\mathrm{n} \mathrm{C}\mathrm{l}\mathrm{e}\mathrm{a}\mathrm{r} \mathrm{W}\mathrm{i}\mathrm{d}\mathrm{t}\mathrm{h}=\left({\displaystyle \frac{\mathrm{Usable} \mathrm{Width} \mathrm{Pixel} \mathrm{Count}}{\mathrm{Total} \mathrm{Width} \mathrm{Pixel} \mathrm{Count}}}\right)\times 100\%$
	Shading Facility Coverage	Han, J.W. [74] Feng, X.X. [76]	Percentage of shaded areas, such as pergola, shade, etc., and semantic segmentation to identify shaded areas (shadows + amenities).	$\mathrm{S}\mathrm{h}\mathrm{a}\mathrm{d}\mathrm{i}\mathrm{n}\mathrm{g} \mathrm{F}\mathrm{a}\mathrm{c}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y} \mathrm{C}\mathrm{o}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{g}\mathrm{e}=\left({\displaystyle \frac{\mathrm{Shade} \mathrm{Area} \mathrm{Pixel} \mathrm{Count}}{\mathrm{Total} \mathrm{Pixel} \mathrm{Count}}}\right)\times 100\%$
	Public Seating Density	Han, J.W. [34] Fang, H.Y. [77]	Percentage of street seats, etc., semantic segmentation identifies seats and counts them.	Public Seating Density = $\left({\displaystyle \frac{\mathrm{Number} \mathrm{of} \mathrm{Seats}}{\mathrm{Street} \mathrm{Length} (\mathrm{in} \mathrm{hundreds} \mathrm{of} \mathrm{meters})}}\right)\times 100\%$
	Street Enclosure Ratio	Han, J.W. [34] Fang, Z.G. [33]	Evaluate the degree of interface enclosure and street openness, composed of elements such as walls, buildings, and pavements, affecting the privacy and openness of the space.	$\mathrm{Street} \mathrm{Enclosure} \mathrm{Ratio}=\left({\displaystyle \frac{{\mathrm{P}}_{\mathrm{w}\mathrm{a}\mathrm{l}\mathrm{l}}+{\mathrm{P}}_{\mathrm{b}\mathrm{u}\mathrm{i}\mathrm{l}\mathrm{d}\mathrm{i}\mathrm{n}\mathrm{g}}}{{\mathrm{P}}_{\mathrm{s}\mathrm{i}\mathrm{d}\mathrm{e}\mathrm{w}\mathrm{a}\mathrm{l}\mathrm{k}}}}\right)\times 100\%$
Cultural Dimension	Cultural Activity Space Ratio	Gu, W.Q. [32] Jiang, L. [78]	Proportion of public space available for cultural activities (plazas, stages, bazaars) to the total area of the neighborhood, and semantic segmentation to identify activity spaces.	Cultural Activity Space Ratio = $\left({\displaystyle \frac{\mathrm{Activity} \mathrm{Space} \mathrm{Pixel} \mathrm{Count}}{\mathrm{Total} \mathrm{Pixel} \mathrm{Count}}}\right)\times 100\%$
	Informal Vending Space Occupancy	Gu, W.Q. [32] Fang, Z.G. [33]	Proportion of pavement area occupied by temporary stalls (snacks, handicrafts) in the street, semantic segmentation to identify stall areas.	Informal Vending Space Occupancy = ${\displaystyle \frac{\mathrm{Stall} \mathrm{Area} \mathrm{Pixel} \mathrm{Count}}{\mathrm{Sidewalk} \mathrm{Area} \mathrm{Pixel} \mathrm{Count}}}\times 100\%$

4.2. Machine Scoring and Human Scoring

Semantic segmentation, a deep learning technique, enables pixel-level recognition and delineation of distinct objects and scenes within images. Employing a fully convolutional network (FCN-8s) trained on the ADE20K dataset, street view images undergo semantic segmentation to classify pixels into 150 urban elements, including road, sky, and car (Figure 9). For each sampling point, the pixel ratio of a specific element is calculated by averaging its proportion across four-direction street view images (0°, 90°, 180°, and 270°), yielding the final pixel ratio of that element at the sampling point.

Through deep learning semantic segmentation technology, pixel-level analysis of street view images is conducted. Based on indicator calculation formulas, indicators such as the green vision rate and signboard clutter index are quantified (Figure 10). The scatter plots of the proportion of each indicator are shown as follows (Figure 11):

Machine scoring was derived by normalizing semantic segmentation-derived metric percentages through min–max linear scaling, where the observed range of each metric (minimum to maximum percentage) was proportionally mapped to a standardized interval of [−3, +3], yielding the machine-calculated scores for individual indicators (Figure 12).

To enhance the objectivity and scientific rigor of spatial quality assessment in Jinan Mingfu City historic district, this study introduces a “double-blind manual scoring” mechanism, integrating the perspectives of planning experts and tourists to conduct a refined assessment of the three dimensions of “Visual Perception-Spatial Comfort-Cultural Experience” through street view images as the medium.

• Methodological Protocol:

Data Acquisition: High-definition streets view images of five representative streets were captured via big data crawling, constructing a standardized image repository comprising a total of over 71 images. To facilitate comparison of scoring data, the images were numbered from 0 to 70, based on sampling points.

• Scoring Process:

Expert Group: Including 10 certified urban planners.

Visitor Group: Including 50 participants (local residents and non-local tourists).

Double-Blind Design: Participants independently scored randomly assigned images through a uniform scoring interface without knowledge of the evaluation targets or each other’s identities, effectively mitigating subjective biases arising from regional preferences or professional standpoints.

The manual scoring results are shown in Figure 13 and Figure 14:

These data indicate that expert scores across all three dimensions of the historic district were consistently higher than those of tourists (Table 8). This highlights that current protection and development policies overly focus on the physical spatial ontology and expert perspective [9,10], lacking authentic tourist experience data. Consequently, it is difficult to enhance the tourist-perceived spatial quality.

Table 8. Mean scores of expert and tourist ratings across three dimensions.

Dimension	Expert Group	Visitor Group	Differentiation Characteristics
Visual Dimension	1.28	1.18	Expert scores are slightly higher (+8.5%)
Comfort Dimension	1.12	0.86	Expert scores are significantly higher (+30%)
Cultural Dimension	1.05	0.64	Expert scores are substantially higher (+64%)

Table 8. Mean scores of expert and tourist ratings across three dimensions.

Dimension	Expert Group	Visitor Group	Differentiation Characteristics
Visual Dimension	1.28	1.18	Expert scores are slightly higher (+8.5%)
Comfort Dimension	1.12	0.86	Expert scores are significantly higher (+30%)
Cultural Dimension	1.05	0.64	Expert scores are substantially higher (+64%)

The visual perception dimension exhibited high consistency (e.g., Qu Shuiting Street experts 2.74 versus tourists 2.67), with negligible ambiguity. However, significant differences emerged in the cultural experience dimension, exemplified by Jiang Junmiao Street (experts 1.35 vs. tourists 0.18, a 650% difference) and Shuang Zhongci Street (experts 0.68 vs. tourists 0.26, a 161% difference). This discrepancy stems from experts prioritizing architectural details (e.g., decorated archways and brick carvings) and traditional activities, whereas tourists emphasize overall cleanliness, bustling atmosphere, and comfort (influenced by pedestrian flow), coupled with insufficient cultural symbol cognition. Therefore, this study will integrate the dual perspectives of experts and tourists for correlation analysis of the machine scoring, ensuring the scoring results reflect both professional conservation requirements and capture the tourists’ spatial perception experience.

4.3. Analysis of Indicator Weighting Adjustment and Data Calibration Based on Human–Machine Scoring Correlations

Data format consistency between machine-generated scores and double-blind human evaluations was verified [80]. Rigorous validation confirmed alignment in sample IDs (ranging from 0 to 70) and street groupings, including Shengfu Front Street, Hou Zaimen Street, etc. Regarding data normalization, both human ratings (visual/comfort/cultural dimensions) and machine scores (10 indicators) employed a unified [−3, +3] scale, eliminating the necessity for additional normalization. Employing Spearman’s rank correlation analysis [81], combined with effect size (Cohen’s d) [82] to determine the magnitude of impact, this study analyzed the associations between the ten machine-evaluated indicators and the three human-rated dimensions (Figure 15 and

Table 9. Effect size of machine-derived metrics on human ratings.

Machine Indicator	Dimensions	Mean of Compliant Group	Mean of Non-Compliant Group	Mean Difference	Effect Size (Cohen’s d)
Green Vision Rate	Visual Score	2.04	0.20	+1.84	1.28 (Large Effect)
Shading Facility Coverage	Visual Score	2.04	0.20	+1.84	1.28 (Large Effect)
Effective Pedestrian Clear Width	Visual Score	1.85	0.31	+1.54	0.99 (Large Effect)
Street Enclosure Ratio	Visual Score	−0.38	1.60	−1.98	−1.44 (Large Negative Effect)
Green Vision Rate	Comfort Score	1.50	0.49	+1.01	0.69 (Moderate Effect)
Effective Pedestrian Clear Width	Comfort Score	1.54	0.47	+1.07	0.70 (Moderate Effect)
Shading Facility Coverage	Comfort Score	1.58	0.44	+1.13	0.77 (Moderate Effect)
Street Enclosure Ratio	Comfort Score	0.04	1.33	−1.29	−0.91 (Large Negative Effect)

):

• Visual perception dimension: The green vision rate and shading facility coverage demonstrated strong positive correlations with visual perception scores, with correlation coefficients of 0.656 and 0.671. This indicates that vegetation coverage and shading infrastructure critically enhance visual experiences in urban streets and parks. Conversely, the street enclosure ratio exhibited a strong negative correlation (−0.604), suggesting excessive enclosure induces visual oppressiveness and necessitates calibrated enclosure control in urban design. Indicators such as the signboard clutter index, cultural activity space ratio, and informal vending space occupancy showed negligible correlations (|r| < 0.1), indicating minimal impact on visual evaluation.
• Spatial comfort dimension: The green vision rate and shading facility coverage maintained moderate positive correlations with comfort scores, with correlation coefficients of 0.442 and 0.485. The street enclosure ratio has a negative correlation (−0.469) with comfort evaluation scores, meaning that high enclosure is unfavorable for people to obtain a comfortable experience. Indicators such as the water visibility ratio and signboard clutter index exhibited weak associations (|r| < 0.2), implying limited influence on comfort.
• Cultural experience dimension: Most machine-evaluated indicators displayed statistically insignificant correlations with cultural scores.

The green vision rate, shading facility coverage, and effective pedestrian clear width exhibit large positive effects on visual perception scores (d = 1.28, 1.28, 0.99) and moderate positive effects on spatial comfort scores (d = 0.69, 0.77, 0.70). This indicates that indicators within the visual perception and spatial comfort dimensions demonstrate high consistency between machine-derived and manual evaluations. Such alignment stems from significant physiological and psychological commonalities in human responses to visual and comfort-related stimuli, coupled with the intuitive nature of these indicators, which are readily assessable through direct visual observation and bodily experience. Moreover, they are also amenable to accurate machine quantification via measurable parameters.

In contrast, machine–manual correlations for cultural dimension indicators (e.g., cultural activity space ratio, and informal vending space occupancy) are generally weak. This discrepancy arises primarily from the highly subjective interpretation of cultural connotations, which varies substantially based on individual backgrounds, experiences, and emotions. Furthermore, cultural elements (e.g., historical heritage, folk customs, and artistic ambiance) are inherently complex and often intangible, posing challenges for machine-based capture and quantification (e.g., ordinary corners imbued with historical memory). Additionally, machine scoring typically fails to account for contextuality and narrativity, whereas manual evaluations prioritize emotional resonance embedded in cultural scenarios. Therefore, for the assessment of the cultural experience dimension, it is necessary to integrate manual subjective evaluation (such as questionnaires and expert assessment) based on machine quantification data, and commit to deeper analysis of cultural elements, explore methods for translating intangible culture into quantifiable characteristics, and develop algorithms capable of recognizing contextuality and narrativity to incorporate into the scoring system.

5. Discussion

Based on the three-dimensional evaluation results of “visual perception-spatial comfort-cultural experience” and indicator impact analysis, this study identified that green vision rate, effective pedestrian clear width, shading facility coverage, and street enclosure ratio are the core driving indicators. For these indicators, the mean values across streets were compared (averaging machine-calculated metrics and manual double-blind scores) to obtain indicator proportion and score data for each street (

Table 10. Mean proportion of key impact indicators versus mean three-dimensional ratings along streets.

Street	Green Vision Rate	Shading Facility Coverage	Effective Pedestrian Clear Width	Street Enclosure Ratio	Visual Score	Comfort Score	Cultural Score
Qu Shuiting Street	0.286	0.256	0.096	0.314	2.533	1.694	0.858
Jiang Junmiao Street	0.076	0.070	0.052	0.528	0.121	0.639	0.479
Shuang Zhongci Street	0.051	0.047	0.056	0.543	0.049	0.410	0.578
Hou Zaimen Street	0.073	0.062	0.047	0.556	0.631	0.485	0.970
Shengfu Front Street	0.192	0.134	0.062	0.374	1.634	1.504	0.534

). This data-driven analysis served as the basis for proposing targeted enhancement strategies.

As shown in

Table 11. Core deficiencies and strengths of streetscapes.

Street	Core Deficiencies	Strengths
Jiang Junmiao Street	low green vision rate (0.076), extremely low visual score (0.121)	relatively high street enclosure ratio (0.528)
Shuang Zhongci Street	lowest green vision rate (0.051), lowest shading facility coverage (0.047), lowest comfort score (0.41)	highest street enclosure ratio (0.543)
Hou Zaimen Street	lowest effective pedestrian clear width (0.047), low shading facility coverage (0.064), low comfort score (0.485)	highest cultural score (0.97)
Shengfu Front Street	relatively low shading facility coverage (0.134), low effective pedestrian clear width (0.062)	acceptable visual/comfort score

above, Qu Shuiting Street scored well across all dimensions and exhibited a relatively low street enclosure ratio. In contrast, the other streets demonstrated significant deficiencies in certain aspects. Therefore, to address these issues and consider the substantial disparities observed in the cultural dimension, targeted and highly feasible improvement recommendations are proposed for specific streets across three domains: visual landscape optimization, facility layout adjustment, and cultural scenario implementation. These recommendations aim to enhance the spatial quality enhancement strategies for Jinan Mingfu City’s historic cultural district, thereby achieving a dynamic balance between preservation and revitalization.

5.1. Visual Landscape Optimization Strategies

Shuang Zhongci Street and Jiang Junmiao Street exhibit significant deficiencies in the green vision rate, resulting in monotonous streetscape visuals that lack vitality and aesthetic coherence, thereby failing to deliver a comfortable visual experience. It is recommended to implement strategic greening interventions across roadside areas, building walls, and public spaces along these streets. For example, street trees are planted on both sides of the road, and tree species that are adapted to the local climate and soil conditions, with beautiful shapes and lush branches and leaves, such as Sophora japonica and Ginkgo biloba, are selected; vertical greening is set up on the walls of buildings, and climbing plants such as ivy and creepers are planted to enrich the street facade landscape (Figure 16).

5.2. Facility Layout Adjustment Strategies

Regarding the indicators of shading facility coverage and effective pedestrian clear width, Shengfu Front Street, Hou Zaimen Street, and Shuang Zhongci Street performed poorly. Shuang Zhongci Street exhibited the weakest shading facility coverage (0.047), which makes pedestrians lack shade when walking on the street, especially in hot summer weather, which greatly reduces their comfort. Hou Zaimen Street had the narrowest effective pedestrian clear width (0.047), which is not conducive to the smooth passage of pedestrians and also restricts the development of public activities on the street. Therefore, facility layout adjustment and optimization within the district were undertaken based on practical conditions. For example, widen the sidewalks by appropriately compressing the width of the roadway and adjusting the outdoor space of shops on both sides of the street. In the widening process, pay attention to maintaining the historical style of the street and use paving materials that match the original street materials, such as bluestone slabs and permeable bricks, to improve the comfort and safety of pedestrians. Furthermore, public seating was rationally installed along the sidewalks, and shading facilities were added. Traditional pavilions and pergolas can be used to meet the shade requirements and blend in with the style of the district (Figure 17).

5.3. Cultural Scenario Implementation Strategies

In the cultural experience dimension, although Hou Zaimen Street achieved a relatively high cultural score of 0.97, the scores for Shengfu Front Street (0.534286), Jiang Junmiao Street (0.479444), and Shuang Zhongci Street (0.578667) indicate an overall deficiency in cultural ambiance cultivation. Based on the analysis of cultural indicator correlation differences and tourist experience demands, this study proposes the following cultural scenario implementation strategies to enhance the perception of cultural atmosphere in the district. These strategies primarily aim to deeply strengthen the district’s cultural narrative continuity, construct an innovative ecosystem for the transmission of living heritage, and achieve organic integration of historical-cultural resources with modern cultural tourism, thereby deepening tourists’ cultural perception of the district (Figure 18).

• Water narrative axis: Taking the Qushui River as the vein, it is proposed to plan a full-process experiential chain of “water drawing-tea tasting-lantern releasing” to recreate the classical literati ambiance of Qushui Liushang. Tourists engage in spring water tea brewing rituals, embodying millennia of spring culture. Then, at nightfall, they can participate in lantern-floating ceremonies, comprehending the poetic charm of Jinan amid sparkling ripples.
• Alley memory wall: Building upon the historical depth of Jiang Junmiao Street and referencing the implemented example of Fuzhou’s Three Lanes and Seven Alleys (Sanfang Qixiang) Historical and Cultural District AR Metaverse Experience [83,84] (which has achieved digital restoration and display of the Ming-Qing era alley layout), the plan is to use AR (augmented reality) technology to dynamically project scenes of late Qing dynasty market life onto the mottled walls. Tourists can trigger a time-traveling interactive experience by touching the walls, thus making historical scenes “come alive” in the present.
• Courtyard revitalization: Consider transforming the Jinju Lane into communal hubs for intangible cultural heritage, creating open-access ICH transmission spaces. Regular activities such as cloisonné craftsmanship experiences and Shandong cuisine culinary classes are held, allowing tourists to appreciate the charm of intangible cultural heritage through hands-on practice and promoting the intergenerational inheritance of traditional culture.

6. Conclusions

This study established a technical path of “data crawling, factor aggregation, and human-computer collaborative optimization”. By integrating multi-source data and semantic segmentation technologies, it achieved quantitative assessment and empirical optimization of the spatial quality in Jinan Mingfu City Historic District. The developed three-dimensional indicator system—“visual perception, spatial comfort, and cultural experience”—confirmed that indicators such as the green vision rate, shading facility coverage, and street enclosure ratio significantly influence tourists’ spatial perception, providing operational quantitative standards for spatial quality evaluation in historic districts. This closed-loop model of “data diagnosis–indicator quantification–strategy generation” provides a replicable technical paradigm for transforming historic districts from empirical protection to intelligent renewal.

Although the current research has broken through the limitations of traditional qualitative evaluation, machine learning algorithms still have limitations in capturing tourists’ subjective perception of historical and cultural connotations (e.g., artistic expression of architectural details, emotional memory of the district). Quantifying the cultural experience dimension requires continued integration with the double-blind scoring mechanism. Furthermore, the existing indicator system has limited ability to quantify tourists’ dynamic behaviors (e.g., length of stay and path selection). In the future, GPS trajectory and eye-tracking data can be integrated to improve the evaluation dimensions. Future research should refine the multi-source data collaborative optimization paradigm, such as integrating real-time perception data from eye-tracking and physiological indicators to establish a more comprehensive tourist experience evaluation system. It should also explore the combination of cultural heritage semiotics and place attachment theory with semantic segmentation technology to develop a bidirectional mapping model of “physical space-cultural meaning”, thereby enhancing the quantification capacity for subjective perceptions such as historical ambiance and cultural environment.

This study provides technical support for the transition of historic districts from experience-oriented conservation to data-driven intelligent renewal, helping to realize the dynamic balance between conservation, development, and tourists’ needs in the context of cultural and tourism integration. The methodology can provide quantitative tools for similar districts, such as Nanluogu Lane in Beijing, Pingjiang Road in Suzhou, etc., and help to realize the dynamic balance between cultural heritage preservation and sustainable development of the city.