From data acquisition to digital reconstruction: virtual restoration of the Great Wall’s Nine Eyes Watchtower

This article presents the virtual restoration of the Nine Eyes Watchtower, a significant cultural heritage site along the Great Wall. By applying the Seville Charter and digital technology, a detailed virtual restoration work-flow is developed. The methodology involves acquiring data from multiple sources, including physical evidence, historical data, and comparative data. Advanced survey technologies, architectural knowledge, historical research, and computer modelling techniques are integrated to accurately capture the architectural and historical significance of the Nine Eyes Watchtower. The virtual restoration process follows a systematic approach, combining evidence interpretation and explicit deduction steps. The main outcome is a comprehensive virtual restoration model that accurately represents the architectural features and historical context of the Nine Eyes Watchtower. The virtual scene includes environmental elements, with potential for immersive exploration. By bridging the gap between interpretation and deduction, this study advances the scientific understanding and presentation of virtual restorations. The project contributes to ongoing research, education, and appreciation of the Great Wall’s cultural legacy, ensuring its continued relevance for future generations


Introduction
The purpose of virtual restoration methods is to generate 3D reconstructions of ancient architectural structures that are partially or totally destroyed.This process starts with an evaluation of the 3D multiresolution representation of a site in its present state and information according to historical and archival sources.These sources often provide nonmetric properties and include both analogue and digital formats (Di Angelo et al. 2019).The starting point of the method is a 3D reality-based model, which is used to generate the first reconstruction, which is typically applied as a reference for the architectural structure.
In recent years, advances in data acquisition techniques and reality-based representations and the availability of low-cost digital modelling systems have led to an in-depth review of the methodologies for documenting cultural heritage (CH).Currently, two main 3D data acquisition methods are used to generate 3D realitybased models of CH objects and sites: range-based methods, in which range sensors are used, and image-based methods, in which imaging techniques are used (Martínez-Carricondo et al. 2020).Range and image data are often combined to generate 3D multiresolution representations, which are then used to derive different geometric levels of detail (LODs) of the given scene.Additionally, to completely describe CH objects, the 3D geometry and original location relative to the surrounding context must be constructed depending on the documentation and conservation purposes of the specific survey (Al-Ruzouq et al. 2018).Unmanned aerial vehicles (UAVs), i.e., model helicopters that fly autonomously, using amateur digital cameras and GPS systems can be used to obtain images of areas that are difficult to access.The result obtained is a point cloud or a 3D mesh model that can be used as a basis for creating building information models, which can then be used in CH management, analysis, reconstruction, and restoration (Valenti and Paternò 2019).
The digitisation and virtual restoration of CH sites have attracted substantial attention in recent years as crucial endeavours to preserve and promote the historical and cultural value of these sites, making them accessible and understandable to the public (Han et al. 2020).However, many virtual reconstruction projects have faced challenges in incorporating historical and scientific references, and these projects often rely on subjective assumptions and creative interpretations.This raises the fundamental question of how to present 3D virtual reconstructions as scientific processes supported by comprehensive data.
Virtual restoration, as an evolving discipline in archaeological and architectural research, aims to reconstruct artefacts and structures based on available evidence and historical knowledge (Demetrescu 2018;Pietroni and Ferdani 2021).The process begins with the collection of all relevant information about the CH site, including data gathered through surveys, excavations, ancient drawings, photographs, and historical documents (Stamati et al. 2022).Advanced digital technologies, such as hyperspectral imaging, laser scanning, and photogrammetry, are crucial for capturing detailed information and generating accurate digital representations of CH sites (Zhou et al. 2019;Hester, Shafer, and Feder 2009;Parcak 2009).In addition, evidence, including apparent attributes such as appearance and colour, and historical materials, sketches, and structures are important in the virtual restoration process, particularly in deducing missing parts of CH sites and objects (Shui et al. 2015;Lu et al. 2015;Chen S. 2018;Chen Y. 2018;Soto-Martin et al. 2020).
Despite recent initiatives and guidelines for scientific visualisation and reconstruction processes, more methodologically consistent and standardised solutions are needed in the field of virtual restoration.The complexity of digital acquisition, data analysis, and 3D modelling techniques and the various types of data used contribute to the inconsistencies among virtual restoration practices.Efforts such as the London Charter, Sevilla Principles, Venice Charter, Nara Document on Authenticity, and Burra Charter have sought to establish guidelines, but more comprehensive and standardised approaches would be beneficial for the field.
Among the sites in need of extensive restoration and preservation, the Great Wall of China remains one of the most iconic and significant architectural marvels in human history.The Great Wall holds immense historical research value, particularly in the fields of military architecture, technology, and art.Unfortunately, the Great Wall has been impacted by natural disasters and anthropogenic destruction, resulting in extensive damage and the loss of critical historical information (Liu et al. 2020;Wei 2012;Du et al. 2017).As a result, restoring comprehensive information about the Great Wall has become a pressing issue, especially considering the challenges posed by minimal intervention policies (Chen 2019).The virtual restoration methods and techniques needed for the Great Wall also hold substantial value for similar projects worldwide, given that there are more than 29,510 buildings spread along its length (Zhang et al. 2021).
This paper introduces a virtual restoration workflow for the Great Wall, focusing on the Nine Eyes Watchtower as a case study.The primary objective is to formalise the technical steps in the virtual restoration process specifically for the Nine Eyes Watchtower, enabling the tracking, management, and visualisation of the restoration process (Liu, Yang, and Zhang 2016;Chen S. 2018;Chen Y. 2018;Du et al. 2020).The proposed workflow incorporates various technologies to address challenges related to evidence organisation and classification, including data acquired by UAVs, terrestrial laser scanners (TLSs), and textual sources (Han et al. 2020).Additionally, the workflow employs a triple-scale evidence approach to deduce the original form and structure of the Great Wall, integrating information from multiple sources and perspectives (Chen et al. 2016;Soto-Martin, Fuentes-Porto, and Martin-Gutierrez et al. 2020).By utilising these advanced technologies and incorporating a comprehensive range of evidence, the virtual restoration workflow aims to more accurately determine missing parts and faithfully reproduce the original form and structure of the Nine Eyes Watchtower.
The significance of this research extends beyond the scope of the Nine Eyes Watchtower or the Great Wall itself.The proposed virtual restoration workflow can be extended and applied to other domains, such as archaeology and architectural heritage, enabling the preservation and restoration of CH sites worldwide.By formalising the technical steps and emphasising the importance of evidence-based restoration processes, this research contributes to enhancing the scientific integrity and quality of virtual restoration practices.

The Nine Eyes Watchtower of the Great Wall
The Nine Eyes Watchtower, located in Sihai town (formerly known as the Sizhen and Xuanfu areas), Yanqing District, Beijing, was constructed during the early Ming Dynasty.Positioned atop a steep mountain ridge along the Great Wall, it stands at an elevation of 1,255 m (Fig. 1).The name of the watchtower references the nine distinctive holes present on each side, resembling eyes; hence, it is commonly referred to as the Nine Eyes Watchtower (Hua 1988).Unfortunately, due to prolonged natural erosion, the watchtower has partially collapsed over time (Fig. 2a).Historical records and legends indicate that the original structure of the Nine Eyes Watchtower included three sections: a base, a first floor, and a second floor.However, presently, only the base and first floor remain intact (Yanqing County Gazetteer Editorial Committee 1993).Figure 3 depicts the missing parts of the Nine Eyes Watchtower, including the second floor and upper sentry house.
Located at the junction of the three military defence zones of Jizhen, Xuanfuzhen and Changzhen, the Nine Eyes Watchtower is the eastern starting point of the Great Wall and connects two defence lines.Furthermore, as the watchtower is situated at the intersection of three mountain ridges, enemy incursions may occur on all sides, increasing the importance of military observations and early warning.
According to an inscription (Wang 2018), the watchtower at the site was initially a sturdy solid body until 1466.However, due to its minimal defensive effect in this form, in 1544, during the Jiajing period of the Ming Dynasty, it was transformed into a hollow watchtower, which was initially created by Qi Jiguang, a famous military general of the Ming Dynasty, as a highly distinctive type of military engineering to defend against the Huns.Following the hollow enemy watchtower construction ideas proposed by Qi Jiguang, the nine-hole building was converted into a single facade with 9 windows for arrows, in addition to the 37 arrow windows of the hollow brick building.To date, the Nine Eyes Watchtower is a unique case with the largest number of arrow windows, the thinnest arrow windows and the largest window-wall area ratio along the Ming Great Wall.Based on Qi Jiguang's construction philosophy and the military needs of the Nine Eyes Watchtower, this form aligns with historical records from the mid-16th century, including documents such as 'Four Towns and Three Passes' (Liu 2018), 'The Great Wall Ji Town Map of Ming Dynasty' (Yang 1987), and' Actual Record of Military Training' (Qi 1985).
According to the military analysis of the Nine Eyes Watchtower by Prof. Zhang Yukun of Tianjin University (Fig. 2b), the Nine Eyes Watchtower had to satisfy the needs of soldiers who monitored the hillside on all sides and support the observation of the tall mountains on the periphery (more than 10 degrees in elevation).Thus, the Nine Eyes Watchtower was designed to have a large number of thin arrow windows (the width and height of the windows were approximately 600 mm and 1750 mm, respectively).This design not only meets the surveillance needs of soldiers patrolling the surrounding corridor but also explains why the Nine Eyes Watchtower has a large number of thin arrow windows.In summary, according to the design of a hollow enemy building proposed by Qi Jiguang, as well as the unique morphological

Research stages
To virtually restore the Nine Eyes Watchtower, a comprehensive methodology including survey technology, architectural knowledge, and computer modelling is employed.This approach, illustrated in Fig. 4, consists of three main steps: data acquisition, virtual interpretation, and virtual restoration model development.
The first step, data acquisition, involves the collection of digitised geometric information on the remaining structure of the watchtower.In addition, historical records and information from similar CH sites are acquired.This process utilises advanced survey technologies, such as UAVs and TLSs, to obtain precise and detailed measurements of the existing structure (Rodríguez-Moreno et al. 2016).Additionally, extensive research on the relevant literature and comparable watchtowers is conducted.This comprehensive data acquisition approach ensures the collection of accurate geometric data of the remaining structure and reliable sources for the missing components, laying the foundation for the complete virtual restoration of the Nine Eyes Watchtower.
The second step, virtual interpretation, focuses on interpreting the morphology and historical context of the watchtower.In this step, historical records, ancient drawings, photographs, and other available sources are analysed and studied to determine the original form and structure of the watchtower.By integrating architectural knowledge and conducting in-depth historical research, the virtual interpretation process aims to reconstruct the missing parts of the watchtower and generate intelligible representations of its past state.This step is crucial for ensuring the accuracy and authenticity of the virtual restoration.
The final step, model development, utilises computer modelling techniques to create a virtual restoration model of the Nine Eyes Watchtower.Specialised software, such as 3ds MAX, is employed to integrate the acquired geometric data and virtual interpretations into a cohesive virtual representation.The model development process includes the reconstruction of missing components, the restoration of architectural features, and the addition of textures and materials to enhance the visual realism of the virtual restoration.This step aims to faithfully recreate the original appearance of the watchtower and provide viewers with an immersive and accurate virtual experience.
By following this methodology, the virtual restoration of the Nine Eyes Watchtower aims to capture its historical and architectural significance while generating a realistic representation of its original state.The combination of data acquisition, virtual interpretation, and model development approaches ensures the precision, fidelity, and scientific integrity of the virtual restoration process.This study provides a systematic framework for effectively restoring and preserving CH sites, allowing future generations to appreciate and understand the historical value of sites such as the Nine Eyes Watchtower.

Data acquisition
To establish a comprehensive data acquisition approach, this study adopts a multifaceted evidence gathering strategy, which includes physical evidence, historical data, and comparative data, inspired by the Portus Theodosiacus restoration project (Ortiz-Cordero, Pastor, and Fernández 2018).This approach involves various sources and perspectives to ensure a comprehensive understanding of the subject matter.The three different data sources employed in this research are physical evidence, historical data, and comparative data.Physical evidence refers to the existing information of the heritage site, which can be studied by mapping the form, structure, and material of the site itself.Historical data include documentary evidence such as archaeological records, sketch descriptions, and photographs.Comparative data use consistent historical backgrounds, common productivity levels, and general technological conditions to identify reference CH sites with similar functions, structures, and material characteristics.By utilising these different data sources, the reliability of the evidence can be assessed, ensuring a scientifically grounded virtual restoration process.The following section describes the process of obtaining data from each of these sources in detail.

Collection of physical evidence
Given the challenging mapping conditions posed by the watchtower's location on a steep mountain, a standard procedure was followed to obtain the necessary data.The following equipment were used: The collection of physical evidence involves four stages: (1) control measurements, (2) the use of UAVs and TLSs to capture exterior and interior information (3D point clouds) of the Nine Eyes Watchtower, (3) the fusion of the two types of point cloud data to generate a triangulated irregular network (TIN) and a 3D model with texture mapping for the existing structure, and (4) the extraction of physical evidence from the 3D model.

Control measurement
A precise control network was constructed using basic mapping instruments to meet the 5 cm precision requirement for the UAV image control points.A GPS control network was constructed according to the geographic location of the Nine Eyes Watchtower Heritage Area.Three GPS control points at the E-level of Beijing were selected for fixation and burial underground.The measurement points buried around the Great Wall near the Nine Eyes Watchtower constitute a precise control network (Fig. 5), which transforms the coordinate data of the CH area into the point cloud data, model data, and UAV data, enabling the use of one coordinate system.Figure 6 shows the distribution of the image control points deployed in the Nine Eyes Watchtower survey area.

Data acquisition using UAVs and TLSs
In this study, data related to the exterior of the watchtower were collected using a UAV camera imaging system, specifically a DJI Phantom-4 RTK, as depicted in Fig. 7a.The flight path of the UAV was divided into four loops (Fig. 7b), with a total flight duration of 12 min.To meet the requirements for 3D model reconstruction, the ground sampling distance (GSD) was maintained below 2-3 mm, and images with less than 80% overlap were removed, yielding 355 usable images.The process of generating the digital surface model (DSM) point clouds and orthophotos involved importing photos, using positioning and orientation system (POS) data, and applying camera calibration parameters in specialised software.Aerial triangulation was then performed, with point features detected and matched, and the precise camera locations for each aerial image were identified.Finally, the obtained POS data were combined in the CGCS2000 national coordinate system.A A TLS (device model: FARO Focus Plus350) survey was conducted to capture the interior masonry structure and barrel vault space of the watchtower (Fig. 8).The 360° × 270° full field of view enabled data acquisition from vaults, horizontal floors and vertical walls in the building, with a high scanning speed of up to 50,000 points/ second.Six scanning sites were set up in the interior of the Nine Eyes Watchtower, and the scanning procedure included object point cloud scanning, target scanning, and digital photo shooting to accurately represent the wall details.The scanning density was set to 5 cm/100 m.The point cloud stitching error of each site was less than 6 mm during the object point cloud scanning process.

Data fusion for creating a 3D model
The 3D model of the Nine Eyes Building was created by combining two types of 3D information from the point cloud: image-based modelling and range-based modelling results.ContextCapture software was used for imagebased modelling.In this approach, the multinomial dense matching technique is first implemented to automatically match points with the same name.Then, coarse point detection is carried out, a free mesh is constructed, the image-control point coordinates are input, and the beam method is utilised to repeatedly adjust the model according to the levelling results, including parameter settings and image-control point position adjustments, until the accuracy requirement for three-dimensional reconstruction is satisfied.3D reconstruction is a modelling process based on tile technology in which the maximum tile texture does not exceed 100 pixels.Each tile is packaged and built into a task, stereo image pairs are constructed using an aberration elimination technique to check multiview images and high-precision outer orientation elements optimised via aerial triangulation, ultra-dense 3D point clouds of the features are generated, and different 3D point clouds are constructed within the segmented area according to the set priority level.Then, according to the set priority level, the 3D point cloud in the partitioned area is constructed in different levels of the irregular triangular network (TIN) model to generate the grid of the Nine Eyes Watchtower (Fig. 9a).It took approximately 7.3 h to process the UAV data.
For range-based modelling, FAROScene ® v.6 software was used to edit the point clouds acquired with the FARO terrestrial laser scanner.Checkerboard targets are automatically detected and checked because the software may detect and interpret different patterns as targets.The point clouds from the six scans are then registered using common targets.Then, a noise reduction technique is applied to the point cloud.The point cloud is also filtered to remove unnecessary scanned points.In cases with very complex surfaces, this may result in incorrect scan points due to the offset of the scanning beam.Finally, the point cloud is successfully registered with an average error of 1 mm.
The Nine Eyes Watchtower contains both an exterior and an interior.The mesh model created from the laser scanner point cloud focuses on the interior of the Nine Eyes Watchtower, while the point cloud created using the drone-based imagery focuses on the exterior of the Nine Eyes Watchtower.The UAV photogrammetry point cloud data and TLS point cloud data are converted to the universal *.las format for point cloud data.In this case, the point clouds from the range-based and image-based models were imported into Geomagic software v. 12 for further editing.As the TLS point cloud is more accurate than the UAV photogrammetry data, the TLS point cloud is used as a benchmark during the registration and fusion process.The iterative closest point (ICP) algorithm and manual registration method are then employed to align and register the two point cloud datasets, resulting in a high-precision fused point cloud.Subsequently, a TIN of the Nine Eyes Watchtower (Fig. 9c) is generated using the point-bypoint insertion method.After texture mapping, a 3D model is generated based on the existing conditions of the remaining structure of the watchtower (Fig. 9d).

Extraction of physical evidence
After the data fusion process, the physical evidence includes the watchtower's form, structure, materials, craftsmanship, and geometric information.The architectural form refers to the style and appearance of a structure, especially within a specific geographical area, exhibiting universal characteristics.An architectural structure involves a local or an overall load-bearing structural system, ensuring structural integrity.Craftsmanship includes the technical construction processes, including material usage, structure creation, and other common techniques.Geometric information mainly refers to the dimensional data extracted from the orthophoto of the building feature parts via CAD annotation.Table 1 provides the details of the extracted physical data.For instance, the major construction materials used for the Nine Eyes Watchtower are bricks and irregular granite.Based on the orthophotos, the architectural form is determined to consist of three sections: the stone base, the first floor, and the second floor.The first floor features arches and barrel vaults, as well as arrow window holes and indoor spaces.The construction technique for arches and barrel vaults involves layer-bylayer brickwork.Furthermore, the wall exhibits 'Flemish bond' craftsmanship.The geometric information related to the materials, structure, and floor plan is extracted from the orthophotos using manual methods with the aid of Autodesk AutoCAD software.
The collection of physical evidence plays a vital role in the virtual restoration process, providing crucial insights into the original characteristics and historical context of the Nine Eyes Watchtower.By utilising advanced survey technologies and carefully analysing the extracted physical data, the accuracy and scientific integrity of the virtual restoration workflow can be guaranteed.

Collection of historical data
In addition to physical evidence, historical data are important for the virtual restoration of the Nine Eyes Watchtower.Although the physical remnants of the watchtower cannot provide a complete understanding of its original state, historical data, including historical photographs of the Nine Eyes Watchtower, offer valuable insights.Figure 10a shows one photograph, which Table 1 Extracted physical evidence from the Nine Eyes Watchtower (Source of figures: the authors) reveals the presence of regularly distributed wooden pillar holes and spouts on the top of the outer wall.This historical evidence suggests that the Eyes Watchtower was likely a brick-wood structure, with rainwater expected to drain to the first floor through these spouts.

Collection of comparative data
In addition to the physical evidence and historical data, comparative data serve as valuable resources in the virtual restoration process.This indirect evidence complements the information obtained from other sources and provides additional insights into the design and characteristics of the Nine Eyes Watchtower.
To ensure the reliability of the virtual restoration results, the research team extensively reviewed the relevant literature and examined similar watchtowers.The literature and documents, depicted in Fig. 11, such as 'Four Towns and Three Passes' , 'The Great Wall Ji Town Map of the Ming Dynasty' , and ' Actual Record of Military Training' , serve as vital sources of comparative data.Descriptions and sketches found within these documents represent typical watchtower forms, including a base, a first floor, and a second floor, with a distinct quadrangular frustum shape and a gable and hip roof (Xieshan roof ) over the second floor (Fig. 12).
Furthermore, in the Sizhen and Xuanfu areas where the Nine Eyes Watchtower is located, as illustrated in Fig. 13, a significant characteristic of similar watchtowers is the  11 The official literature about watchtowers on the Great Wall (Source: the authors) flush gable roof design, which was prevalent during the 14th to 15th centuries (see http:// www.ncha.gov.cn/).Based on this comparative evidence, can be tentatively concluded that the roofs of these watchtowers, including the Nine Eyes Watchtower, likely featured a flush gable roof with a '人' shape.Additionally, decorative components such as Wang shou and Pao shou were commonly placed on the main ridge and vertical ridge, respectively, as depicted in Table 2.It is also plausible that the interior space of the sentry house was constructed with a barrel vault design.
By incorporating comparative data, the researchers improved their understanding of the architectural characteristics and regional styles prevalent in watchtowers near the Nine Eyes Watchtower.This comparative analysis aids in the formulation of hypotheses and assumptions for the virtual restoration process, enhancing the accuracy and authenticity of the final restoration outcome.

Virtual interpretation
To deduce the missing of the Nine Eyes Watchtower, including the floor structure, roof form, and interior space of the sentry house, a virtual interpretation workflow is proposed.The detailed analysis process is as follows:

Structure of the second floor
Considering the physical evidence and historical data, two initial structures to support the floor are proposed: a barrel vault structure and a brick-wood structure.Figure 14 presents the virtual interpretation workflow for deducing the floor structure.
In steps 1 and 2 of this workflow, a preliminary assumption of a barrel vault structure is made, and finite element analysis is conducted to verify the feasibility of this assumption.Figure 15 shows a schematic of the barrel vault layout, including the pressure (N), bending moment (M), and shear force (V) distributions.By comparing the three types of loads on the vertical wall, it is determined that the shear force poses the most critical risk.Considering various live loads and the dead weight of the brick arch, a calculation Formula (1) is derived to estimate the limit of the shear force.The results indicate that if the maximum masonry height of the arch exceeds Table 2 Comparative data from the Nine Eyes Watchtower were collected from a study of similar watchtowers in the region and from the literature (Source of figures: the authors) Fig. 14 A virtual interpretation workflow for deducing the floor structure (Source: the authors) 1.305 m, cracks may form in the weak part of the vertical wall.
Figure 15 clearly shows that when the barrel vault height reaches the limit value (H = 1.305 m), the vault does not reach the floor of the second floor (H = 1.65 m), contradicting preliminary assumption.Therefore, in steps 3 and 4, a revised assumption is made that the floor structure is likely supported by a brickwood structure.As mentioned in Sect 4.2, historical (1) photos of the Nine Eyes Watchtower show the remains of wooden pillar holes and spouts, indicating a brickwood structure with rainwater drainage to the first floor via spouts.Based on the above reasoning, a reconstruction diagrammatic drawing of the floor structure is presented (Fig. 16).The drawing includes 36 wooden pillars evenly spaced within the reserved holes of the brick walls.Each row of pillars is connected to the inner wall through beams arranged in a '回' shape to reinforce the beam and pillar lap structure.The spouts are positioned at the corners of the middle platform to facilitate rainwater flow into the drainage ditch and corridor.

Architectural form of the sentry house
Two possibilities for the architectural form of the missing sentry house are considered: a gable and hip roof or a flush gable roof (Fig. 17).The deduction process is depicted in Fig. 18.Evidence from historical documents and analyses of neighbouring in the same region as the Nine Eyes Watchtower suggest that flush gable roofs are prevalent in the region.Therefore, the architectural form of the sentry house is most likely a flush gable roof with grey round tiles, featuring Wang

Interior space of the sentry house
To verify the interior space of the missing sentry house, existing and comparative data are examined.Figure 19a shows the presence of brick barrel vault structures in neighbouring watchtowers, while Fig. 19b displays the barrel vault space on the first floor of the Nine Eyes Watchtower.Based on this evidence, it can be inferred that the missing sentry house most likely had a barrel vault space, as shown in Fig. 20.

Virtual restoration scheme of the Nine Eyes Watchtower
Based on the virtual interpretation results, a virtual restoration scheme for the Nine Eyes Watchtower was created in Autodesk AutoCAD.The scheme includes plan views, elevations, and sections of the watchtower (Fig. 21).The restored first floor plan (Fig. 21a) shows a barrel vault space in the middle, while the restored top floor plan (Fig. 21b) depicts a sentry house with a flush gable roof.Both floors are surrounded by a ring corridor.The restored sections (Fig. 21c, d) provide insights into the internal structure and spatial design of the watchtower, including the barrel vault space and the relationship between the upper and lower floors.The restored elevations (Fig. 21e, f) show the overall form and appearance of the Nine Eyes Watchtower, highlighting features such as the stone base, brick walls, arrow windows, brick eaves, stacking wall, and sentry house.
Through the virtual interpretation process, the missing parts of the Nine Eyes Watchtower were deduced, allowing for comprehensive virtual restoration of its architectural elements.The floor structure, roof form, and interior space of the sentry house were reconstructed based on a combination of existing physical evidence, historical data, and comparative analyses.The virtual restoration scheme provides a realistic representation of the original state of the Nine Eyes Watchtower, capturing its historical and architectural significance.

Virtual restoration
This case study discusses the integration of on-site point clouds and modelling methods to reconstruct one of the famous Great Wall watchtowers, the Nine Eyes Watchtower.The primary purpose of representing the original watchtower through a virtual model was to faithfully represent the architectural features and historical context before presenting the historical knowledge and cultural heritage of the original ruined building to the public.
The virtual restoration of the Nine Eyes Watchtower involves a meticulous and detailed process to recreate its missing components and accurately represent its architectural features.This study develops semantic templates for virtual reconstruction (Fig. 22), including the building blocks, relationships and hierarchies of the objects, as well as combinatorial relationships, and accurately describes the architectural features (Fig. 23).The entire process was carried out using 3ds MAX software, which allows for precise modelling and manipulation of the digital environment (Salim Ferwati and Menshawy 2021).
First, the 3D model of the Nine Eyes Watchtower created in step 4.1.2,which captures its current state, is imported into 3ds MAX.The exterior wall of the Nine Eyes Watchtower is reconstructed through a point cloud, as shown in Fig. 24.In step 1, the stone base is reconstructed according to the 'whole-part' semantic relationship.In step 2, the wainscoting on the stone masonry base is reconstructed.Then, in step 3, the masonry of the wall columns is determined.Next, in step 4, the arrow window structure on the top of the wall columns is developed based on the previous step.In step 5, the upper transverse wall on the top of the arched arrow windows is reconstructed based on Flemish bond craftsmanship.Finally, in step 6, the brick gable and the palisade wall are reconstructed.The overall process of reconstructing the stone base, brick wainscot, column, brick eaves, and stacks is based on point cloud morphology, and each part is reconstructed according to the semantic rules, leading to the gradual restoration of the exterior walls of the Nine Eyes Watchtower.
Figure 25 shows the reconstruction process of the barrel arch space inside the Nine Eyes Watchtower.In this study, the barrel arch space, arched doorways and window openings are reconstructed via a reverse approach based on the point cloud features.The wainscot part of the centre room arch and the centre room arch window are reconstructed in steps 1-2.Then, the remaining parts are reconstructed by following the same semantic rules in step 3. Finally, the arched and vertical walls of the centre room are digitally reconstructed.
Then, the floor and accessory structures are restored.Elements such as brick eaves and battlements are meticulously recreated in 3D (Fig. 26).The brick-wood structure, comprising walls, pillars, beams, and planks, is reconstructed to support the sentry house on the second floor (Fig. 26a and b).Next, the roof is restored to generate the final virtual model.The roof components, including flat tiles, round tiles, and decorative elements, are carefully designed and placed using specialised plugins in 3ds MAX software.The aim is to represent the original roof structure and capture the intricate details that are crucial to the overall appearance and character of the Nine Eyes Watchtower (Fig. 26c and d).Finally, in addition to the meticulously recreated watchtower, the surrounding environment, including mountains and trees, is reconstructed in the virtual restoration process.This holistic approach allows researchers to view the Nine Eyes Watchtower from various perspectives and to more comprehensively understand its historical context and natural setting.Figure 27 presents the virtual restoration results, demonstrating the virtual scene bringing the Nine Eyes Watchtower to life.With the detailed reconstruction of the watchtower and the accurate representation of the natural elements, a captivating virtual environment is created that closely resembles the original site.
Throughout the virtual restoration process, the accuracy, authenticity, and visual realism of the model are highly considered.By combining existing physical evidence, historical data, and comparative data, the virtual restoration captures the architectural characteristics and historical significance of the Nine Eyes Watchtower.The result is a digital representation that allows viewers to understand the watchtower's original form and appreciate its cultural and historical value.

Discussion
The virtual restoration of CH sites serves various purposes, and the approach and evaluation standards can vary depending on the specific goals of the project.In this research-oriented virtual restoration project, the focus was on achieving a high level of authenticity and integrity.By applying the Seville Charter and leveraging digital technology, a comprehensive and detailed virtual restoration workflow is developed for the Nine Eyes Watchtower on the Great Wall, emphasising the importance of virtual restorations of Great Wall sites.
Data acquisition plays a fundamental role in researchbased virtual restoration because the collected data directly influence the reliability and scientific validity of the reconstruction.By collecting evidence from multiple sources, such as existing physical data, historical data, and comparative data, this project provides a robust foundation for the reliable virtual restoration of the Nine Eyes Watchtower on the Great Wall.
Scientific extrapolation methods are essential for supporting research-based virtual restoration projects.While previous virtual restoration practices explored documentation and analyses at heritage sites, this project aims to bridge this gap by introducing a new method called the virtual deductive workflow.This workflow involves interpreting evidence sources, extracting implicit knowledge, and implementing explicit deduction steps.By incorporating this approach, this project enhances the scientific validity of the virtual restoration process and establishes a stronger connection between interpretations and hypotheses.
Importantly, the virtual restoration results are not absolute and should be continuously evaluated as new data are acquired.This project sought to formalise virtual restoration and interpretation workflows and restoration steps based on an actual case, enabling the tracking, management, and visualisation of the reconstruction process.This iterative approach allows for ongoing assessment and refinement of the virtual restoration results.

Conclusions
Through the application of digital technologies, this project enabled a successful and detailed virtual restoration of the Nine Eyes Watchtower on the Great Wall.The virtual restoration workflow includes data collection, virtual interpretation, and virtual restoration model development processes, all of which are supported by scientific research and inferential logic.By integrating three critical data sources-existing physical evidence, historical data, and comparative data-and employing computer modelling techniques with inferential logic, the rationality and scientific integrity of the virtual restoration strategy for complex sites such as the Nine Eyes Watchtower are significantly improved.The virtual restoration of the Nine Eyes Watchtower serves as a valuable example of preserving and understanding important CH sites.By integrating advanced survey technologies, architectural knowledge, historical research, and computer modelling techniques, a comprehensive virtual restoration model was achieved.The virtual model accurately represents the architectural and historical significance of the Nine Eyes Watchtower, and the interactive elements enable an immersive experience for exploration and engagement.This virtual restoration contributes to ongoing research, education, and appreciation of the Great Wall's cultural legacy.
Overall, this project demonstrates the potential of digital technologies in preserving and interpreting CH sites.By combining scientific techniques, data-driven approaches, and interpretative reasoning, more accurate and meaningful representations of CH sites can be obtained, ensuring their continued relevance and appreciation for future generations.The meticulous implementation of the proposed virtual restoration methodology demonstrates the potential of digital technologies in preserving and celebrating our CH.

Fig. 1
Fig. 1 Location of the Nine Eyes Watchtower (Source: the authors, background map approved by National Bureau of Surveying and Mapping Geographic Information, map approval code: GS[2022]4316)

Fig. 2 aFig. 3
Fig. 2 a Status of the Nine Eyes Watchtower at the top of the ridge.b Patrol view through the arrow window (Source: Zhang Yukun et al. of Tianjin University)

Fig. 4
Fig. 4 Approach for the virtual restoration of the Nine Eyes Watchtower (Source: the authors)

Fig. 5 a
Fig. 5 a Distribution of control points.b Burying of GPS observation points.c Positioning of observation points (Source: the authors)

Fig. 6
Fig. 6 Distribution and coordinates of image control points (Source: the authors)

Fig. 8
Fig. 8 TLSs were used to scan the interior of the Nine Eyes Watchtower (Source: the authors)

Fig. 9 a
Fig. 9 a The point cloud data of the exterior of the Nine Eyes Watchtower.b The point cloud data of the interior of the Nine Eyes Watchtower.c The TIN model of the Nine Eyes Watchtower.d The 3D status model of the Nine Eyes Watchtower (Source: the authors)

Fig. 10 a
Fig. 10 a Traces of the spout on the second floor.b Remains of the wooden pillar holes (Source: the authors)

Fig. 15
Fig. 15 Schematic of the layout of the barrel vault structure (Source: the authors)

Fig. 17 a
Fig. 17 a The gable and hip roof.b the flush gable roof (Source: the authors)

Fig. 20
Fig. 20 Schematic of the barrel vault of the Nine Eyes Watchtower (Source: the authors)

Fig. 26
Fig. 26 Details of the virtual restoration: a The floor structure of the brick-wood structure; b battlements; c roof; d roof tiles (Source: the authors)

Fig. 27
Fig.27The virtual restoration results of the Nine Eyes Watchtower (Source: the authors)