Visual behavior characteristics of historical landscapes based on eye-tracking technology

ABSTRACT

The subjective encounter with architecture encompasses the particular presentation of architectural elements or the overall structure to individuals, taking into account their perception, cognition, and thought processes. The purpose of this research is to investigate the intricate relationship between visual attention directed towards historical architecture and the subjective experiences engendered during the process of observation. Utilizing eye-tracking technology within the realm of virtual reality, this study delves into the observation patterns exhibited by individuals when confronted with historical architecture, specifically focusing on the traditional courtyard residences found in Beijing’s hutongs. The panoramic images, meticulously modeled and rendered, are divided into six distinct areas of interest: base, ground, window&doors, walls, roof, and eaves. The eye-tracking data of 81 participants, who engaged with 10 virtual scenes through the employment of VR headsets, along with their responses to architectural style questionnaires, were systematically gathered. Through comprehensive analysis, encompassing the examination of total fixation duration (TFD), fixation count (FC), first fixation duration (FFD), and time to first fixation (TFF) across the areas of interest (AOIs), in addition to deviations in scores among different architectural areas and styles, notable insights emerged. Results indicate that both professionals and non-professionals allocate heightened attention, as evidenced by TFD, FC, TFF, and FFD, to elements that exhibit greater score deviations in the questionnaires as opposed to those with smaller discrepancies. Moreover, the TFD and FC metrics pertaining to the Windows&doors AOI collectively constitute 40.20% and 40.71%, respectively. Undoubtedly, these figures signify the preeminent focal points within all AOIs. This underscores the pivotal role played by doors and windows in shaping individuals’ judgments pertaining to historical architectural styles. This research provides designers with valuable understanding of the cognitive patterns exhibited by individuals in engaging with aspects of history, which enable them to identify and preserve key elements associated with historical styles. Additionally, it establishes a fundamental cornerstone for future quantitative research endeavors centered around the preservation of historical architectural styles.

KEYWORDS:

• Historic Buildings
• Beijing quadrangle
• VR
• Historic building protection

I. Introduction

Historical style research plays a pivotal role in protecting the integrity of historical streets and buildings, as it contributes to the restoration of their historical appearance and the preservation of their cultural heritage value and regional characteristics. Several studies provide valuable recommendations for protecting the original style and features of historical buildings and streets (Wang 2012), while striking a balance between the preservation of historical buildings and the demands of urban renewal (Knippschild and Zöllter 2021). Additionally, it enhances the environmental quality and promotes the sustainable development of historical buildings and streets (Kou et al. 2018).

Historical style research encounters various challenges and limitations. The distinctive nature of historical buildings and streets poses unique complexities (Boussaa 2017), and data collection can be a demanding task (Hung et al. 2017). Furthermore, the existing research methods of historical style may encounter some limitations (İ̇̇pekoğlu 2006).

The extraction of cognitive elements from historical streetscapes has gained increasing attention in the field of historical streetscape research, with a focus on qualitatively examining how individuals perceive and extract these elements. Various approaches have been employed in this regard. For instance, a Semantic Differential (SD) scale was developed to extract elements from historical streetscapes (Milligan 2007). Others employ the cognitive map overlay method, wherein participants are asked to draw street plans and conduct overlay analyses (Askarizad, He, and Khotbehsara 2022). Additionally, there are studies that involve participants sketching buildings to extract cognitive elements from historical streetscapes (Boschmann and Cubbon 2014).

With the advancement of modern technology, researchers have started incorporating quantitative analyses in the study of historical streetscapes (Benni et al. 2012). Utilize large datasets to examine the proportions of plan dimensions in historical neighborhoods and the color composition of historical buildings (Feliu et al. 2005). Moreover, leveraging historical streetscape images, some studies have combined street image technology with big data and applied machine learning techniques to determine architectural style and quality (Lee and Ostwald 2021; Obeso et al. 2017; L. Zhang et al. 2020). However, this approach necessitates data training and is constrained by the complexity of alley spaces and the accuracy of images, limiting its ability to fully capture subjective judgments. In contrast, eye-tracking technology can be utilized to extract attention to landscape elements in historical streetscapes (Najd et al. 2015).

In the realm of historical landscape preservation research employing eye-tracking technology, certain studies have conducted experiments utilizing a combination of SD semantic questionnaire and eye-tracking technology (Sektani et al. 2021). The experimental materials fall into three categories: the first involves conducting experiments in the field environment, employing eyeglasses-shaped eye-tracking equipment to capture real-time eye gaze data (De la Fuente Suárez 2020). This type of experimental scenario is more intricate and can potentially influence participants’ experiences. The second category employs historical street images displayed on a screen for participants to view, which is constrained to a two-dimensional format, limiting the amount of information people can obtain and thereby affecting the experiment’s effectiveness (Krupa et al. 2021; Lisińska-Kuśnierz and Krupa 2020; Spanjar and Suurenbroek 2020; Sussman and Ward 2019). The third category involves using modeling software to reconstruct three-dimensional scenes and conducting eye-tracking experiments within a laboratory setting (Rusnak and Szewczyk 2018; R. X. Zhang and Zhang 2021; L. M. Zhang et al. 2019). This experimental method yields more precise data and can eliminate certain interfering factors.

In the aforementioned historical block preservation studies employing eye-tracking, the focus primarily lies on individuals’ preferences for landscape elements (Dupont, Antrop, and Van Eetvelde 2014; N. Li et al. 2022). However, the relationship between these preferred elements and people’s historical cognition of architectural style has not been thoroughly explained. Eye-tracking experiments utilizing virtual reality (VR) technology have demonstrated that compared to two-dimensional images, people can acquire information more easily and comprehensively in a VR environment, thereby experiencing lower cognitive load (Wenk et al. 2021). This has the potential to reduce the impact of external factors on people’s judgments of historical landscapes.

In the study of historical urban landscapes, eye-tracking technology is utilized to identify the elements that capture people’s attention during scene observation, which serves as the foundation for preserving the historical character of a place (N. Li et al. 2022). The primary parameters of interest are the number of fixations and their duration, as they reflect the salience of different elements (Krucien, Ryan, and Hermens 2017). Glaholt and Reingold’s research posits that Time to First Fixation is intricately tied to scene information processing, with elements fixated upon earlier deemed more pertinent to the scene task (Glaholt and Reingold 2012). First Fixation Duration, on the other hand, is associated with the search target and, in visual behavioral tasks, is influenced by the similarity between the target and distractors (Reingold and Glaholt 2014). Objects that are more challenging to identify tend to exhibit longer durations, and crucial scene elements also evoke prolonged durations (G. Zhang et al. 2021).Numerous studies indicate noteworthy disparities in eye-tracking parameters and patterns between experts and novices, often leveraging eye-tracking technology (Dogusoy-Taylan and Cagiltay 2014; Gegenfurtner, Lehtinen, and Säljö 2011; Ooms, De Maeyer, and Fack 2014).

Cognitive theories can help explain the undWerlying processes involved in the perception of urban landscapes. When we observe a visual environment, certain objects automatically capture our attention in a “bottom-up” manner, meaning that they stand out from their surroundings (Fernandez-Duque and Johnson 1999; Friedenberg and Silverman 2012; Gazzaniga, Ivry, and Mangun 2009; Lachter, Forster, and Ruthruff 2004). According to attentional filtering models and perceptual load theory, focused visual attention acts like a rapidly moving “spotlight,” allowing only selected information to reach higher levels of processing. Many bottom-up attentional control models are based on the concept of a saliency map, which is a two-dimensional map that encodes the salience of objects in a visual scene. Within this map, competition between neurons determines the winning position that corresponds to the next target of interest, and this position is automatically suppressed to enable the system to attend to the next most salient location (Sussman and Hollander 2015).

The question arises: do people have a similar saliency map when perceiving the historical character of courtyard buildings? What cognitive processes do people employ to judge the historical character of urban landscapes, and which elements are more salient and therefore more likely to be processed first? These questions remain subjects of investigation and further research in the field of historical urban landscape perception and preservation.

The purpose of the research is to investigate the relationship between visual attention and individuals’ assessments of historical and non-historical elements. The study focuses on the classification of architectural elements found in traditional courtyard houses in Beijing. Using eye-tracking data, a quantitative analysis of visual attention patterns was performed to identify how attention is allocated. Interviews and subjective questionnaires were also conducted to provide further insights into these attention allocation patterns. The findings from this research offer recommendations for identifying elements that require preservation and enhancing the evaluation system for the conservation of historic urban areas. By understanding the relationship between visual attention and judgments of historical elements, this study contributes to the development of effective strategies for the protection and preservation of historic urban environments.

II. Research method and experiment design

II.1. Study area

The prototype of this study is situated in San Yan Jing Hutong, Jingshan Houjie, Dongcheng District, Beijing. This hutong has a historical significance as it was part of the imperial city during the Qing Dynasty. Initially named San Yan Jing Hutong during the reign of Emperor Qianlong, it derived its name from a well that had three wells within the alley. Later, during the reign of Emperor Xuantong, it was known as San Yan Jing.

The houses in this area are characterized by their brick-wood structure bungalows. Throughout the renovation process, efforts were made to preserve the original urban texture, maintain the scale and proportion relationship between the hutongs and buildings, and retain the authentic style of the courtyard buildings in the block. The renovation project aimed to safeguard the original architectural style, color schemes, basic layout, and artistic characteristics of the courtyards.

Over time, the area has undergone various transformations, resulting in a mix of historical and non-historical buildings within the same block. These buildings can be classified into different styles of renovation, including conservative, self-transformed, and radical approaches (Figure1). This diverse range of styles within the block provides an ideal setting for studying the relationship between visual attention and people’s judgments of historical and non-historical elements.

Figure 1. Courtyard houses in the study area (from left to right: self-reconstructed, conservative reconstruction, radical reconstruction).

II.2. Experiment preparation

II.2.1. Selection of research object and samples

As shown in Figure 2, in this study, three forms of transformation were established to represent different styles of renovation: Radical reconstruction, Conservative reconstruction, and Residents reconstruction.

1. Radical reconstruction: This form of transformation represents an architect-led retrofit design. The prototypes for this style were selected from the Design Cases website, which showcases innovative and contemporary retrofit designs. The Radical reconstruction approach aims to introduce modern elements and design concepts while preserving the historical character of the buildings.

2. Conservative reconstruction: This form of transformation represents renovation based on the design specifications outlined in the Beijing Old City Conservation Housing Repair Manual. The manual provides guidelines and standards for the preservation and renovation of historical buildings in Beijing. The Conservative reconstruction approach focuses on maintaining the original architectural features and traditional design elements of the buildings.

3. Residents reconstruction: This form of transformation represents the renovations carried out by the residents themselves. The prototype for this style was selected based on research conducted on the current situation of an alley. The Residents reconstruction approach reflects the individual preferences and design choices of the residents, which may vary and differ from the other two styles.

Figure 2. Rendered images of the experimental scene model (in order of conservative reconstruction, residential reconstruction, and radical reconstruction).

To create the scene models for each transformation style, Rhino software was used for modeling, Lumion for rendering, and the resulting 360° images were exported. These 360° images provide a comprehensive and immersive representation of the renovated scenes, allowing for detailed analysis and evaluation of visual attention patterns during the eye-tracking experiments.

The courtyard in the scene has been designed with dimensions of 14 m × 21 m, featuring a single building measuring 9 m × 5 m with a height of 4.5 m. The selection of rendering materials for the scene is based on current situation photos and case photos, ensuring a realistic representation of the renovated environments.

For the eye-tracking experiments, the observation point is set at the center of the courtyard, providing a central vantage point for participants. The viewing height is set to 1.6 meters, which corresponds to the average eye level of most individuals (Leyrer et al. 2015).

The partial transformation focuses on two dimensions: radical-conservative and specific parts of the building, namely roof, facade, and ground. This approach allows for the evaluation of visual attention patterns in response to specific renovation styles and areas of interest.

There are a total of six scenes created, representing different combinations of radical and conservative renovations applied to the roof, facade, and ground areas. In each scene, only the corresponding part is altered, while the rest of the courtyard remains in its original residents’ construction style.

The exported 360° images have a resolution of 4096 × 2048 pixels at 300 dpi, ensuring high-quality visuals with an aspect ratio of 1:2. This resolution enables detailed analysis and evaluation of visual attention patterns during the eye-tracking experiments.

II.2.2. Participants

Eighty-one healthy participants (45 experts [27 females and 18 males]; 21–22 years old with the mean age of 21.55 ± 1.08 years old and 36 nonexperts [21 females and 15 males]; 20–45 years old with the mean age of 22.31 ± 4.55 years old) participated in the experiment. All of them had normal or corrected-to normal vision, and none of them had a history of neurological or psychiatric disorders. All participants completed a consent form to participate in this study. They volunteered to participate in a study to investigate the effect of architectural expertise on Visual behavior characteristics of historical landscapes based on eye-tracking technology. The participants included: (a) current students, including first-year undergraduate students (not enrolled at the time of the experiment) and fourth-year students,(b)construction practitioners，(c) residents of the hutong area.

II.2.3. Experiment equipment and environment

The experimental platform used in this study is Tobii Pro Lab, which is capable of integrating with a 360° image and recording eye movement data. This platform allows for precise tracking and analysis of participants’ visual attention within the virtual environment.

To provide an immersive experience for participants, virtual reality (VR) equipment, specifically the HTC VIVE Pro head-mounted display (HMD), is utilized. This VR setup allows participants to view and interact with the virtual scenes in a realistic and engaging manner.

The experimental site is located at the Beijing Old Town Conservation and Renewal Base of Beijing University of Architecture, situated in an alley in Xicheng District, Beijing. This site provides a suitable environment for conducting research on historical urban landscapes and the perception of architectural transformations.

During the experiments, subjects are seated on a rota-table computer chair, which offers flexibility and ease of movement while wearing the VR equipment. The rotating chair enables participants to freely rotate and explore the 360° scene, enhancing their ability to observe and engage with the virtual environment. The experimental environment is shown in Figure 3.

Figure 3. Laboratory layout.

II.2.4. Questionnaire design

The questionnaire used in this study aims to capture participants’ subjective judgments of architectural style. It includes both overall assessments and specific evaluations of different aspects of the scene. The Likert scale is employed, which allows participants to rate their traditional-modern feelings on a scale consisting of seven points.

To gather participants’ opinions, the questionnaire is presented in a “text + serial number” format within the VR environment. This format is chosen to avoid any potential suggestive influence of numerical ratings on participants. Instead, participants are instructed to verbally communicate their responses to the experimenter while viewing the questions in VR.

The questionnaire begins with an inquiry about the participants’ overall feelings towards the scene. Following that, three separate questions assess their impressions of the three modification sites within the scene. Finally, participants are asked to provide a subjective ranking of the architectural elements they find most significant or interesting.

By utilizing this questionnaire, the study aims to capture participants’ subjective assessments of architectural style, evaluate the perceived sense of history in the scenes, and understand participants’ preferences and priorities regarding specific architectural elements.

II.3. Procedure

The experimental session had a duration of approximately 35 minutes. The participants were initially asked to provide basic information, completed an informed consent form and then instructed to wear the VR headset. They were given time to adjust the headset for optimal comfort and to calibrate the pupil distance (IPD), which ensures more accurate eye movement data. This calibration process typically took around 5 minutes and contributes to the accuracy of the eye-tracking measurements (Murray, Hunfalvay, and Bolte 2017).

Once the calibration was completed, the participants proceeded to view the scenes. Each scene was observed for a duration of 60 seconds. Subsequently, the questionnaire page appeared in text format within the VR environment, and the participants verbally communicated their answers to the experimenter. A 60-second transition break was provided to allow participants to complete the questionnaire.

After viewing all 10 scenes (Table 1), the participants were asked to complete the remaining questions on the questionnaire (Table 2). The order of the scenes was as follows: one real-life photo, six partial transformation scenes, and three overall scenes. This randomization of scene order was implemented to avoid any potential order effects.

Table 1. Experimental scene.

	conservative reconstruction	radical reconstruction
overall
Ground transformation
Facade transformation
Roof transformation
residential reconstruction
Pre-experiment

Table 2. Questionnaire design.

Number	Question content
Question 1	Do you think this place is modern or traditional, please let me know the degree.
		Traditional-T3-T2-T1-0-M1-M2-M3-Modern
Question 2	Do you think these parts are modern or traditional, please let me know the degree
	Ground	Traditional-T3-T2-T1-0-M1-M2-M3-Modern
	Facade	Traditional-T3-T2-T1-0-M1-M2-M3-Modern
	Roof	Traditional-T3-T2-T1-0-M1-M2-M3-Modern
Question 3	Which parts do you think attract your attention? (windows&doors, walls, floors, pedestals, roofs, and eaves) Please let me know your ranking.
	1-> 2- > 3- > 4- > 5- > 6-

To ensure that participants were familiarized with the experimental procedure, they were informed of the process after viewing the first scene and answering the associated questions. During the viewing of the scenes, participants were tasked with evaluating the attributes of wind and landscape. The disordering of the six local scenes and three overall scenes further mitigated any potential order effects.

The study was approved by the local ethics protocols of the Beijing University of Civil Engineering and Architecture (No. BUCEA21092002).

II.4. Data analysis

The raw eye movement data obtained from Tobii Pro Lab software were processed to extract the relevant indicators for analysis. The eye movement data, along with the questionnaire scores, were then imported into SPSS 27.0, a statistical analysis software, for further analysis.

In order to ensure data quality, samples with an eye movement sampling rate higher than 90% were selected for analysis (Ouzts and Duchowski 2012). This criterion helps to ensure that the eye movement data used in the study is reliable and provides accurate insights into participants’ visual attention patterns. The eye-tracking data undergoes filtration employing the Tobii I-VT Fixation Filter. The fundamental purpose of a fixation filter is to sift through fixations within the raw eye-tracking data. In the Tobii software context, it encompasses a comprehensive set of procedures for identifying fixations in the raw data. Within the Tobii Infrastructure for Visualizations and Tobii Fixation Filter, the procedural steps include gap fill-in, eye selection, noise reduction, velocity calculation, I-VT classification, merging adjacent fixations, and discarding brief fixations (Olsen 2012).

Once the data was prepared and filtered based on the specified criteria, various statistical analyses could be conducted to examine the relationships between eye movement patterns, questionnaire responses, and other variables of interest. The explanation of the analysis indicators in this study is shown in Table 3.

Table 3. Explanation of eye movement technical index terms.

	Abbreviation	Significance	Data processing
Total fixation duration (s）	TFD	The longer the total fixation duration, the more attention participants pay to the corresponding area, indicating that the area is more attractive.	SPSS
Fixation count	FC	The more frequently participants fixate on an area, the more they attend to it, indicating that the area is more attractive.
Time to first fixation (s)	TFF	Time to first fixation (TFF) refers to the time it takes for the eyes to move to an area of interest (AOI). If TFF is short, it is easier to notice a particular element, and it is used to measure visual salience.
First fixation duration (s)	FFD	The longer the FFD, the longer it takes to recognize or discover AOI, indicating that it is more attractive or more difficult.
Areas of interest	AOI	Areas of interest (AOIs) are parts of stimuli that are created based on the semantic information of the stimuli.	Defining

II.4.1. Eye-tracking technology and principles

Eye-tracking technology is a relatively recent and innovative research tool employed within the domains of design and environmental psychology. It serves to objectively capture individuals’ observations of a given scene and quantitatively measure their visual attention and cognitive processes. Consequently, it enables the objective representation of evaluators’ subjective perceptions through the utilization of eye-tracking indicators (Duchowski and Duchowski 2017; Holmqvist et al. 2011). In comparison to methods such as verbal reporting by participants, eye-tracking technology boasts enhanced objectivity and offers the advantages of quantitative research and direct evaluation, thus establishing itself as a potent research instrument in the realm of visual studies (Carter and Luke 2020).

The process of eye-tracking involves the monitoring of ocular movements by precisely measuring the position of the eye’s gaze point or its relative displacement in relation to the head. The fundamental nature of eye movements lies in the active or passive allocation of attentional resources, facilitating the selection of information that is more pertinent or visually appealing (Salvucci and Goldberg 2000).

Within the context of this study, an eye-tracking apparatus (HTC Vive Pro EYE) integrated into a virtual reality (VR) system was employed. The eye-tracking parameters encompassed the following: gaze data output frequency (binocular): 120Hz; accuracy: 0.5°-1.1°; calibration: 5 points; tracking field of view: 110°; data output (eye information): timestamps (device and system), gaze origin, gaze direction, pupil position, pupil size, and eyelid status.

II.4.2. Defining AOI

AS shown in Figure 4, The investigation of eye movement behavior and its relationship with the visual environment involved the utilization of Areas of Interest (AOIs). AOIs refer to specific regions within the stimulus space that are deemed relevant and crucial to the experimental design of the study (Orquin, Ashby, and Clarke 2016). In the context of monitor-based eye tracking, AOIs are typically defined by screen pixel boundaries. In our study, we categorized the elements within the scene into distinct AOIs, including building parts: windows & doors, walls, floors, pedestals, roofs, and eaves. These areas of interest were delineated and mapped using Tobii Pro Lab software. To account for oculomotor bias, the boundaries of the AOIs were slightly extended beyond the actual analyzed areas. Subsequently, the software employed an algorithm to assign gaze points to the corresponding AOIs (Benjamins, Hessels, and Hooge 2018).

Figure 4. AOI of different samples.

III. Results

III.1. Subjective questionnaires

III.1.1. Reliability analysis

The test of questionnaire effectiveness includes reliability and validity tests. The reliability test is known from the Cronbach reliability coefficient test on the questionnaire that a coefficient of all the questionnaire data is 0.892 > 0.8 (Table 4) This finding indicates that the questionnaire is useable and highly reliable, indicating that the questionnaire is of good validity (Yan and Hao 2017).

Table 4. Reliability statistics.

Cronbach’s alpha	Standardization items Cronbach’s Alpha	Item number
.892	.893	4

The Cronbach coefficient is less than 0.7, which is the minimum acceptable level of questionnaire reliability, between 0.7–0.8 indicates that the questionnaire reliability is average, 0.8–0.9 considers the reliability to be relatively good, and above 0.9 indicates that the reliability is very good. The reliability of this experiment is in the better interval.

III.1.2. Score for different scenes

The questionnaire employed relative scoring, whereby the score of the reference scenario (resident -reconstructed) was subtracted from the score of each scenario to ascertain the impact of the modifications on the traditional-modern score (Table 5). SPSS was utilized to calculate the mean and standard deviation. Subjects subjectively perceived the experimental scenarios designed in accordance with the guidelines as traditional, while the prototype-based scenarios were perceived as modern. It is worth noting that the partial transformation scenes, despite being anticipated to elicit changes in traditional modernity, did not yield significant results. The overall scores, part scores and heat maps of all scenes are shown in Table 6.

Table 5. Score of each scenario.

Scene name	arithmetic mean	Standard deviation
Pre-experiment	−2.12	2.421
Facade - Conservative	−0.56	3.302
Facade - radical	0.52	2.932
Roof - Conservative	−0.68	2.736
Roof - radical	1.88	2.589
Ground - Conservative	0.28	2.673
Ground - radical	1.72	2.905
resident - reconstruction	0	2.451
conservative reconstruction	−1.64	3.070
radical reconstruction	3.32	2.081

Table 6. Score of each scenario.

Scene name	Heat map	Score Chart of Questionnaire
Pre-experiment
Facade - Conservative
Facade - radical
Roof - Conservative
Roof - radical
Ground - Conservative
Ground - radical
residential reconstructed
conservative reconstruction
radical reconstruction

III.1.3. Subjects’ attention sequencing

Regarding the ranking results (Table 7), subjects allocated a score of 6 to the part they deemed most influential in their overall sense of history, while assigning a score of 1 to the part they perceived as having the least impact. According to participants’ subjective judgment, the window and door areas were considered the most influential elements in terms of historical significance, followed by the walls, both of which are facade components. These findings align with the subjective questionnaire results and are consistent with the objective eye movement data results.

Table 7. Sorting results of different parts.

	pedestals	floors	Windows & doors	eaves	walls	roofs
Score mean	2.31	2.85	5.02	3.41	4.03	3.38
ranking	6	5	1	3	2	4

III.1.4. The effect of different part scores on the overall score

The questionnaire data exhibited a non-normal distribution; thus, the non-parametric Wilcoxon Signed Rank Test was employed to examine significant differences in the impact of each site on the overall wind score. The results indicate the highest variability between the roof and the overall score, while the least variability was observed between the elevation score and the overall score (Table 8).

Table 8. Significance test.

tests statistic a	Ground - Overall	Facade - Overall	Roof - Overall
Z	−2.484	−1.154	−8.142
Asymptotic Significance(2-sided)	.013	.249	<.001

III.2. Eye-tracking data analysis

The software platform facilitated the export of each indicator into an Excel table. For the analysis in this study, four indicators were initially selected: Total Fixations Time (TFT), Fixations Number (TN), First Fixations Time (FFT), and First Fixations Duration (FFD). TFT and TN provide insights into the extent of human observation of the elements, while FFT reveals the order of element observation, and FFD reflects the level of difficulty and interest in observing the elements (Fischer and Weber 1993; Skaramagkas et al. 2021). Subsequently, a normality test was conducted for each Eye-tracking Metric within the AOI. Independent samples t-tests and one-way Analysis of variance (ANOVA) were then performed.

III.2.1. Data analysis of the overall transformation scenario

Among the different areas of interest, the windows and doors section attract more attention (Figure 5). The facade comprises two components: windows&doors and walls. These two elements collectively capture 50% of the total attention time, suggesting that individuals allocate more attention to process the information conveyed by the window and door walls (Figure 6). This finding aligns with the subjective questionnaire results, confirming their consistency. The proportion of AOIs in the picture uses the pixel counting method.

Figure 5. Eye-tracking heat map and AOI proportion of the overall renovation scene. conservative reconstruction. radical reconstruction. residential reconstructed.

Figure 6. Proportions of AOIs in four indicators in the overall renovation scene.

The formula for calculating the proportion is as follows:

$\left[\mathit{P}=\frac{I_i}{{\sum }_{i=1}^n{T}_i}\right]$

$I_i$ represent the indicator value for the $\mathit{i}$-th AOI

$T_i$be the total of indicators for the $\mathit{i}$-th AOI.

When comparing the percentage of Eye-tracking metrics across all scenes, the highest values are consistently observed in the door&window and walls areas of interest (AOI). This finding suggests that individuals tend to allocate a greater amount of attention to the facade when observing quadrangle buildings (Table 9).

Table 9. Percentage of indicators in different parts (%).

	TFD	FC	TFF	FFD	Proportion to the area of the picture
pedestals	12.61%	11.61%	17.63%	17.44%	17.58%
floors	10.39%	10.34%	24.50%	15.82%	42.57%
Windows&doors	40.20%	40.71%	2.64%	19.19%	24.50%
walls	15.69%	15.44%	11.57%	16.96%	31.37%
roofs	9.32%	9.75%	23.85%	14.94%	11.36%
eaves	11.79%	12.14%	19.81%	15.64%	12.73%

The dispersion of the four indicators (F values) was compared across different areas of interest in different renovation scenes to assess differences in attention allocation (Table 10). The highest F value was observed in the radical renovation scene (70.329 > 58.108, 53.934), indicating that variations in attention allocation were more significant when individuals observed this particular scene.

Table 10. F-value comparison.

Metrics	conservative transformation		residential transformation		radical transformation
	F	Sig.	F	Sig.	F	Sig.
TFD	58.108	0.01	53.934	0.01	70.329	0.01
FC	71.08	0.01	86.096	0.01	99.178	0.01
TFF	12.958	0.01	14.702	0.01	13.181	0.01
FFD	1.83	0.106	1.93	0.089	1.334	0.249

The F-value is the statistic of the F-test, which is the ratio of the sum of squared deviations between and within groups to the degrees of freedom, and significance is the level of significance corresponding to the F-statistic, F = between-group variance/within-group variance = MSB/MSE~F(c-1, nT-1).

In order to account for the differences in the visual field occupied by different regions, we normalized the fixation points by the respective area of the visual field. After this normalization, the window and door area remained the most densely fixated area, indicating that it continued to attract the highest concentration of attention from participants.

III.2.2. Data analysis of the partial transformation scenario

III.2.2.1. Ground transformation

Within the realm of ground renovation (Figure 7), two distinct approaches can be identified: courtyard paving and platform formation. When analyzing the areas of interest related to the ground, it becomes evident that there exists a notable disparity in mean fixation time (TFT) between radical and conservative renovations (6698 > 3295, sig. = 0.001, p < 0.05). Similarly, the mean fixation count (FN) for radical renovation (35) surpasses that of conservative renovation (16) with statistical significance (sig. = 0.001, p < 0.05). Moreover, the number of saccades (FFT) for radical renovation (3727) is significantly lower than that of conservative renovation (13769) (sig. = 0.001, p < 0.05). These compelling findings suggest that radical renovations elicit heightened attention within these specific areas in comparison to conservative renovations.

Figure 7. Eye movement heat map of ground transformation scene (conservative reconstruction - radical reconstruction).

When scrutinizing the areas of interest related to the platform, no substantial disparity is observed between the two groups in the context of local renovations (sig. = 0.202, p > 0.05). This could potentially be attributed to individuals directing their attention towards alternative regions, consequently leading to a reduced allocation of attention to the platform area. Furthermore, this implies that the renovation of the platform component may not be directly intertwined with the criteria utilized for discerning historical architectural styles.

Table

Ground transformation - floors						Ground transformation - pedestals
Metrics	Scenes	mean	SD	SEM	Sig.	Metrics	Scenes	mean	SD	SEM	Sig.
TFD	conservative	3295.34	2675.14	367.46	0.001	TFD	conservative	4381.26	3598.75	460.772	0.202
	radical	6698.78	4257.79	549.68			radical	7005.84	4708.58	602.872
FC	conservative	16.91	11.69	1.61	<0.001	FC	conservative	20.98	14.46	1.851	0.524
	radical	35.13	19.53	2.52			radical	32.87	15.88	2.033
TFF	conservative	13796.36	14685.41	2017.20	<0.001	TFF	conservative	6682.10	7429.42	975.531	<0.001
	radical	3727.05	5266.75	679.93			radical	2011.45	5092.76	657.473
FFD	conservative	180.00	129.96	17.85	0.190	FFD	conservative	176.10	99.25	13.033	0.054
	radical	175.15	87.17	11.25			radical	225.88	134.44	17.356

III.2.2.2. Facade transformation

Facade renovation involves the alteration of door&window and walls (Figure 8). Within the AOI related to door&window, it is evident that scenes featuring radical renovations demonstrate a considerably greater fixation time, fixation count, and first fixation duration when compared to conservative renovations (TFD: 17553 > 9063; FC: 94 > 67; FFD: 369 < 990, sig. = 0.001, p < 0.05). In contrast, within the AOI associated with walls, the relationship among these three metrics is reversed and exhibits significant disparities.

Figure 8. Eye movement heat map of facade transformation scene (conservative reconstruction - radical reconstruction).

Table

Facade transformation - windows&doors						Facade transformation - walls
Metrics	Scenes	mean	SD	SEM	Sig.	Metrics	Scenes	mean	SD	SEM	Sig.
TFD	conservative	9063.77	5271.24	674.913	<0.001	TFD	conservative	4978.11	3751.79	480.368	<0.001
	radical	17558.07	5510.25	705.516			radical	2313.70	1967.97	251.972
FC	conservative	67.90	26.87	3.441	<0.001	FC	conservative	22.95	12.09	1.548	0.003
	radical	94.16	28.35	3.629			radical	12.80	9.29	1.190
TFF	conservative	990.51	2072.37	265.339	<0.001	TFF	conservative	2903.58	6164.86	802.596	<0.001
	radical	369.75	825.38	105.679			radical	10208.86	10530.11	1370.904
FFD	conservative	212.65	136.84	17.665	<0.001	FFD	conservative	161.95	107.04	13.935	0.852
	radical	178.26	136.99	17.540			radical	170.54	105.03	13.673

III.2.2.3. Roof transformation

Roof renovation involves modifications to both the roof surface and the eaves (Figure 9). Within the areas of interest related to the roof, it is evident that scenes featuring radical renovations demonstrate significantly higher fixation time, fixation count, and first fixation duration when compared to conservative renovations (TFD: 6589 > 2637; FC: 36 > 14; FFD: 3347 < 10893, sig. = 0.001, p < 0.05). On the other hand, within the areas of interest associated with the eaves, only the fixation time metric shows a significant difference, with higher values observed in conservative renovations compared to radical renovations (4869 > 3206, sig. = 0.007, p < 0.05), while the other three metrics do not exhibit significant disparities. It is worth noting that in this experiment, the sole difference between conservative renovations and resident renovations in the roof scenes was the lifting and folding of the roof truss, yet it still resulted in statistically significant differences. Following the experiment, interviews were conducted with the participants, and both the professional and non-professional groups encountered cases where they were unable to identify certain elements, suggesting the possibility of random occurrences.

Figure 9. Eye movement heat map of roof transformation scene (conservative reconstruction - radical reconstruction).

Table

Roof transformation - roofs						Roof transformation - eaves
Metrics	Scenes	mean	SD	SEM	Sig.	Metrics	Scenes	mean	SD	SEM	Sig.
TFD	conservative	2687.07	1872.95	239.806	<0.001	TFD	conservative	4867.90	3514.71	450.012	0.007
	radical	6589.89	4078.89	522.248			radical	3206.95	2258.47	289.168
FC	conservative	14.31	9.85	1.261	<0.001	FC	conservative	26.62	18.19	2.329	0.027
	radical	36.13	20.99	2.687			radical	18.28	11.30	1.447
TFF	conservative	10893.19	9443.72	1240.022	<0.001	TFF	conservative	7789.23	7425.95	958.687	0.285
	radical	3847.61	4201.03	546.927			radical	5382.03	6017.35	783.393
FFD	conservative	167.02	89.03	11.690	0.896	FFD	conservative	166.83	106.27	13.720	0.379
	radical	176.93	86.45	11.255			radical	160.80	96.22	12.527

III.3. Differences in each indicator between professional and non-professional groups

III.3.1. Eye-tracking data analysis

III.3.1.1. Differences in indicators across various AOIs

As delineated in Table 11, during the observation of AOIs encompassing pedestals and floors, the TFF emerges notably larger for both professional and non-professional cohorts (5969.11 > 4058.80, sig. = 0.03 < 0.05; 10829.35 > 7089.52, sig. = 0.06). Nevertheless, within the Windows & doors AOI, it is the professional group that demonstrates a comparatively abbreviated TFF (702.50 < 1398.75, sig. = 0.04 < 0.05). Non-professional participants exhibit an extended TFD in the floors, Windows & doors, and walls AOIs, whereas in the eaves and roofs AOIs, it significantly diminishes in comparison to the professional cohort.

Table 11. Aois for indicators of professional and non-professional groups in the overall transformation scenario.

		pedestals		floors		Windows & doors		eaves		walls		roofs
		mean	Sig.	mean	Sig.	mean	Sig.	mean	Sig.	mean	Sig.	mean	Sig.
TFD	non-professional	4273.00	0.47	3301.76	0.73	12382.56	0.27	3860.50	0.38	2867.24	0.44	2969.28	0.13
	professional	4294.03		2774.70		12215.22		4550.00		2713.20		3469.49
FC	non-professional	20.65	0.39	16.89	0.25	65.61	0.03	20.07	0.80	15.83	0.95	16.20	0.18
	professional	19.37		13.30		57.81		22.39		14.19		18.02
TFF	non-professional	4058.80	0.03	7089.52	0.06	1398.75	0.04	5960.94	0.42	9952.30	0.87	6571.13	0.18
	professional	5969.11		10829.35		702.50		5271.82		9658.17		6530.78
FFD	non-professional	205.53	0.63	192.20	0.60	179.00	0.08	192.94	0.61	191.96	0.87	161.20	0.68
	professional	190.12		163.74		217.06		175.79		174.24		151.24

III.3.1.2. Proportions of various indicators in professional and non-professional groups

The proportion of indicators remains highest within the Windows & doors AOI, regardless of whether it is the professional or non-professional group (Table 12). In the non-professional group, the TFD has a higher percentage in the floors AOI compared to the professional group, while it is lower in the walls AOI.

Table 12. Professional and non-professional groups percentage of indicators in different parts (%).

		pedestals	floors	Windows & doors	walls	roofs	eaves
TFD	non-professional	14.41%	11.13%	41.76%	13.02%	9.67%	10.01%
	professional	14.31%	9.24%	40.69%	15.16%	9.04%	11.56%
FC	non-professional	13.30%	10.88%	42.26%	12.93%	10.20%	10.44%
	professional	13.35%	9.17%	39.85%	15.43%	9.78%	12.42%
TFF	non-professional	11.59%	20.24%	3.99%	17.02%	28.41%	18.76%
	professional	15.32%	27.79%	1.80%	13.53%	24.79%	16.76%
FFD	non-professional	18.30%	17.12%	15.94%	17.18%	17.10%	14.36%
	professional	17.73%	15.27%	20.24%	16.40%	16.25%	14.11%

III.3.2. Values indicating the degree of dispersion across the three scenarios

The non-professional group exhibits significant differences in various renovation styles, with the F-value for the conservative style being lower than that of the radical scenario (Table 13). However, in the professional group, such differences are not pronounced, and none of the three renovation styles show significant distinctions. This trend is evident across the TFD, FC, TFF, and FFD indicators.

Table 13. F-value comparison.

Metrics		conservative transformation		residential transformation		radical transformation
		F	Sig.	F	Sig.	F	Sig.
TFD	non-professional	13.985	0.01	8.995	0.01	26.041	0.01
	professional	45.401	0.01	45.801	0.01	40.635	0.01
FC	non-professional	18.697	0.01	13.439	0.01	36.846	0.01
	professional	53.220	0.01	72.469	0.01	54.317	0.01
TFF	non-professional	2.408	0.042	6.972	0.01	5.378	0.01
	professional	10.859	0.01	10.952	0.01	8.477	0.01
FFD	non-professional	0.326	0.896	1.132	0.349	1.853	0.110
	professional	2.043	0.073	1.393	0.228	1.775	0.118

III.3.3. Subjective questionnaires

III.3.3.1. Score for different scenes

The non-professional group categorizes only conservative reconstruction and the Pre-experiment scenario as traditional, perceiving other partial renovations as modern (Table 14). In contrast, the professional group regards Facade – Conservative as aligned with tradition, showcasing substantial differences from the non-professional cohort. Furthermore, in alternative scenarios, the score differentials between the two contrasting transformation styles are more pronounced. For instance, in conservative transformation, the scores for the non-professional and professional groups are −1.23 and −1.49, respectively, whereas in radical transformation, the scores are 2.77 and 4.37, signifying that the professional group associates higher traditional scores with conservative-style transformations and higher modern scores with radical-style transformations.

Table 14. Score of each scenario.

Scene name	non-professional		professional
	arithmetic mean	SD.	arithmetic mean	SD.
Pre-experiment	−1.94	2.744	−2.10	2.260
Facade - Conservative	0.00	3.734	−1.26	2.886
Facade - radical	0.53	3.077	2.51	2.814
Roof - Conservative	0.53	2.910	0.74	2.608
Roof - radical	1.48	2.728	2.23	2.597
Ground - Conservative	0.42	2.921	1.02	2.595
Ground - radical	1.36	3.199	2.37	2.780
resident - transformation	0.00	2.818	0.00	2.223
conservative transformation	−1.23	3.430	−1.49	2.847
radical transformation	2.77	2.503	4.37	1.881

III.3.3.2. Subjects’ attention sequencing

The professional group regards Windows & doors as the most crucial, followed by walls and roofs, with minimal disparity between floors and eaves (Table 15). In contrast, the non-professional group considers walls to be more important, albeit with a slight difference compared to Windows & doors. The most significant distinction from the professional group lies in the eaves, where the non-professional group sees it as a representation of traditional style. Both groups unanimously consider pedestals to be the least important.

Table 15. Sorting results of different parts.

		pedestals	floors	Windows & doors	eaves	walls	roofs
non-professional	Score mean	1.89	2.17	4.44	4.11	4.61	3.78
	ranking	6	5	2	3	1	4
professional	Score mean	2.49	3.14	5.26	3.12	3.79	3.21
	ranking	6	4	1	5	2	3

III.3.3.3. The effect of different part scores on the overall score

The non-professional group manifests the most minimal score discrepancy in ratings for Facade – Overall, while the professional group displays the slightest score variations in Ground – Overall and Facade – Overall, with Roof – Overall standing out as the most substantial difference (Table 16).

Table 16. Significance test.

	tests statistic a	Ground - Overall	Facade - Overall	Roof - Overall
non-professional	Z	−2.094b	−1.078c	−6.590b
	Asymptotic Significance (2-sided)	0.036	0.281	0.001
professional	Z	−.184b	−.340b	−11.037b
	Asymptotic Significance (2-sided)	0.854	0.734	0.001

IV. Conclusions

IV.1. Identification of historical architectural character through structural components and overall architectural style

When examining the correlation between style scores of the ground, facade, and roof components and the overall score, it becomes evident that the facade score carries greater weight in influencing the overall score, regardless of whether the setting is modern or traditional. The evaluation of whether the architectural courtyard aligns with historical architectural character primarily relies on the architectural style exhibited in the facade. This can be attributed to the fact that facade elements are situated closer to the visual center of human perception, making them more easily noticeable and observable (N. Li et al. 2022; Z. Li et al. 2021). This correlation may also be linked to the structural characteristics of the eyeball. The central fovea, being most receptive to high spatial frequencies, is specialized for object recognition and detailed analysis (Pointer and Hess 1989; Shapley and Lennie 1985; Wright and Johnston 1983). In contrast, the peripheral retina is most responsive to low spatial frequencies, facilitating behaviors such as fixating on both peripheral and central foveal areas (Chung, Legge, and Tjan 2002; Davis, Yager, and Jones 1987; McKee and Nakayama 1984), thereby contributing to this phenomenon.

IV.2. Eye gaze patterns and identification of historical architectural character

Based on the subjective questionnaires and participants’ initial fixation times, it was consistently observed that when observing a siheyuan (quadrangle courtyard), participants directed their attention primarily towards the main building. Furthermore, there was a specific sequence of gaze directed towards individual architectural elements: windows and doors were attended to first, followed by walls, and finally pedestals and eaves, with subsequent focus on the ground and roof. This gaze order indicates the varying levels of attraction that these elements hold for participants, highlighting the heightened noticeability of windows and doors. These findings align with the weightings assigned in the subjective questionnaires and may also be influenced by the proximity of structural components to the visual center, which is consistent with previous research findings (Tatler 2007; Tseng et al. 2009).

Elements located near the visual center possess greater perceptual salience, making them more likely to capture attention. As a result, these visually prominent areas serve as the foundation for evaluating whether a building adheres to its historical architectural character. The areas that attract visual focus subsequently form the basis for subjective judgments when assessing architectural character.

IV.3. Attention allocation and identification of historical architectural character

Upon analyzing the duration and number of fixations, it becomes apparent that participants predominantly focus their observation on windows and doors, as well as the walls-two prominent elements of the facade. This observation suggests that participants tend to redirect their attention to windows and doors after sequentially examining other areas, allocating more time to observe these particular regions. The longer the fixation duration, the more information is acquired, which also influences the weighting of judgments.

It is well-established that visual attention is naturally drawn to spatial elements that possess unique characteristics (Berto, Massaccesi, and Pasini 2008; Franěk, Petružálek, and Šefara 2019; Hollander et al. 2020; J. Li et al. 2020; Nordh, Hagerhall, and Holmqvist 2013). Within the same scene context, radical renovations elicit longer fixation durations and a greater number of fixations compared to conservative renovations. This finding aligns with existing research, suggesting that elements deviating from cognitive norms tend to captivate attention (Askari 2009; Utaberta et al. 2012).

IV.4. Influence of architectural style on attention allocation

Regarding the three overall renovation scenes (radical, residential, conservative), participants’ evaluations resulted in categorizing these scenes into three styles: modern, neutral, and traditional. The dispersion of eye-tracking indicators within these three styles exhibited a gradient. Specifically, scenes characterized by a more modern style demonstrated a greater dispersion in attention allocation, with eye-tracking indicators among the six architectural elements displaying more variation. Conversely, scenes embodying a traditional style showcased a more balanced distribution of attention among the elements, with smaller numerical differences.

As a traditional architectural form characterized by balanced and unified composition, the Beijing courtyard has undergone historical sedimentation. Consequently, participants had a visually harmonious experience during the viewing process. The balanced attention allocation not only reflects the harmonious visual perception but also provides a comfortable visual experience.

IV.5. Renovation inclinations and recognition of architectural style

In the design process of renovations aimed at preserving historical features, careful consideration should be given to people’s gaze sequence and attention allocation. Since windows and doors have the greatest impact on the assessment of historical style, it is crucial to preserve their architectural form, color, and details as faithfully as possible. Other areas can be subject to appropriate renovations based on practical living needs.

Furthermore, the evenness of attention allocation can serve as a novel objective indicator for evaluating historical style. Renovations should strive to achieve a balanced distribution of attention among different architectural elements, ensuring that each element receives a sufficient level of visual focus. This approach can contribute to the preservation and enhancement of the historical character of the architectural design.

IV.6. Differences in historical style perception between professional and non-professional groups

In the comparative analysis of visual behavior between professional and non-professional groups, both cohorts exhibit a predominant focus on Windows & doors. However, the disparity lies in the non-professional group’s heightened attention to floors, whereas the professional group directs a more pronounced focus towards eaves and roofs in the AOIs. The professional group displays a more equitable distribution of attention. In terms of indicator values and F values, the non-professional group manifests more substantial differences, while the professional group demonstrates smaller disparities.

Regarding the assessment of scene styles, the professional group’s observational behavior is characterized by uniformity, displaying consistent attention to each area with minimal variations. In contrast, the non-professional group tends to be captivated by specific regions. In the subjective judgment process of scene styles, the professional group exhibits greater sensitivity. In comparison to the non-professional group, the professional cohort perceives scenes with conservative transformation as imbued with a stronger historical sense, while scenes with radical transformation are deemed more modern.

In the hierarchy of importance for judgment criteria, the professional group’s scores are more balanced, considering multiple renovation locations simultaneously. This inference is drawn from the correlation results.

V. Discuss

This study utilizes eye-tracking technology to examine the visual behavioral characteristics of participants in tasks related to the recognition of architectural styles. By employing virtual reality technology and manipulating the renovation variable across different architectural elements, the study measures participants’ eye-tracking indicators, subjective assessments of architectural style, and subjective gaze sequences.

The contributions of this study can be summarized as follows:

1. Introduction of a novel approach that combines subjective questionnaires and eye-tracking indicators to explore architectural style recognition and evaluate the stylistic features of architectural scenes. The study identifies elements associated with the preservation of historical styles based on the division of architectural elements.

2. Proposal of even attention allocation as an indicator for assessing scene styles.

3. Innovative suggestion of a methodology that integrates subjective descriptions of attentional areas with objective eye-tracking Areas of Interest (AOIs), thereby enhancing the accuracy and effectiveness of data analysis.

4. When ascertaining the historical style of a building, the Windows & doors and walls of the Facade are deemed to better encapsulate the overall historical essence, constituting over 40% of the Total Fixation Duration and Fixation Count.

The findings of this study can inform the prioritization and intensity of renovation in different architectural elements during the design process. By providing architects with a non-professional perspective on experiencing the visual process of architectural heritage, the study contributes to the optimization of renovation designs.

This study has certain limitations. Firstly, it focuses solely on exploring the renovation styles of architectural elements and does not quantify the impact of specific elements’ renovation intensity on architectural styles. Future research could conduct experimental studies to examine the effects of varying renovation intensities on specific elements, such as comparing the impact of ornate carvings and painted decorations on windows and doors.

Secondly, the questionnaire design has inherent limitations. The understanding of terms like “traditional/modern” can be influenced by professional backgrounds and cultural contexts. Therefore, it is recommended to provide participants with references and establish clear definitions for the extreme values of traditional and modern styles prior to their completion of the questionnaire. For example, participants could be presented with images of buildings identified by experts as representing extreme examples of traditional/modern styles. This approach can help mitigate the influence of individual differences when participants express their opinions using the Likert scale questionnaire.

The third section: The experiment primarily features the front facade of a courtyard building, limiting visibility from alternative orientations. Future experiments might consider employing a mobile experimental approach within the scene.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The work was supported by the Young Scientists Fund of the National Natural Science Foundation of China [52108004]; Beijing Municipal Natural Science Foundation [8222016, 8242008]; Humanities and Social Science Youth Foundation of Ministry of Education of China [20YJCZH041]; the Key Program of Social Science Fund of Beijing [19JDSRA001]

Notes on contributors

Shimeng Hao

Dr. Shimeng Hao is an associate professor at Beijing University of Civil Engineering and Architecture and a WELL AP. Her main research focuses on exploring the mechanisms through which the built environment affect human behavior and psychological characteristics.

Rui Hou

Rui Hou, master’s student, studying at School of Architecture and Urban Planning ,Beijing University of Civil Engineering and Architecture

Jie Zhang

Professor Jie Zhang is the Honorary Dean of the School of Architecture and Urban Planning at Beijing University of Civil Engineering and Architecture, and a Professor and Doctoral Supervisor at the School of Architecture at Tsinghua University. He is recognized as a National Master of Engineering Survey and Design, and a First-Class Registered Architect. Professor Zhang has long been engaged in research, teaching, and practical work in the fields of urban and rural heritage conservation, renewal, and sustainable development, as well as urban characteristic construction. He has achieved pioneering research results in the field of cultural cognition and overall protection of traditional Chinese cities and settlements. Professor Zhang has authored 20 monographs and published over a hundred papers in domestic and international journals, including SSCI and Chinese core journals.

Yang Shi

Dr. Yang Shi is an associate professor at Beijing University of Civil Engineering and Architecture, a nationally registered urban and rural planner. His main research interests focus on the preservation and utilization of historic cultural districts and traditional villages.

Yisong Zhang

Yisong Zhang, master’s student, studying at School of Architecture and Urban Planning, Beijing University of Civil Engineering and Architecture.

Chen Wang

Chen Wang, master’s student, studying at School of Architecture and Urban Planning, Beijing University of Civil Engineering and Architecture.