A traditional village virtual restoration style coordination evaluation method based on eye tracking

Abstract

The application discloses a kind of traditional village virtual repair style coordination evaluation method based on eye tracking, belong to traditional village protection and virtual reality technical field.The method constructs traditional village virtual repair scene three-dimensional model and divides attention area;Organization of subjects is browsed in VR environment, and eye tracking device is used to collect fixation point, gaze duration, saccade amplitude, pupil diameter and other eye movement data;Objective eye movement index of each area is extracted and combined with subjective score, and the style coordination comprehensive evaluation index is calculated by weighted fusion algorithm;With preset threshold comparison output coordination level and optimization suggestion.The application fuses objective eye movement data and subjective evaluation system, realizes the objectification and semi-quantification of style coordination evaluation, takes into account professional and public dual perspective, forms "scheme design-virtual evaluation-feedback optimization" pre-evaluation closed loop, and provides intuitive, efficient decision-making tool for scientific comparison and dynamic correction of traditional village repair scheme.

Description

Technical Field

[0001] This invention relates to the fields of traditional village preservation and virtual reality technology, and in particular to a method for assessing the harmony of the virtual restoration of traditional villages based on eye tracking. Background Technology

[0002] Traditional villages are an important legacy of agrarian civilization, carrying irreplaceable historical information, architectural techniques, and regional cultural memories. With the deepening of cultural heritage protection efforts, many traditional villages face the dual tasks of protective restoration and adaptive renewal. In this process, whether newly built, renovated, or added architectural elements can maintain harmony with the existing village landscape has become a core technical issue determining the success or failure of restoration projects. Therefore, there is an urgent need for an objective evaluation method that can scientifically measure the degree of harmony between the restoration plan and the original village landscape.

[0003] However, current practices for assessing the harmony of village landscapes have the following prominent shortcomings: First, the assessment process relies on on-site inspections and subjective judgments by a few experts. Evaluation standards are difficult to unify due to individual differences, and the assessment results lack reproducibility and horizontal comparability, failing to provide a consistent quantitative benchmark for comparing multiple schemes. Second, the assessment perspective is limited to the architectural professional paradigm, ignoring the actual visual perception and aesthetic preferences of ordinary users and local residents regarding the village landscape. Existing research shows that there is often a significant deviation between professional evaluation and public perception, and a single professional judgment cannot fully reflect the rich connotations of landscape harmony. Third, the development of virtual reality technology has provided the technical possibility for pre-simulation of restoration schemes, but currently there is a lack of an operational assessment methodology that effectively combines virtual restoration scenarios with visual perception assessment. Especially in the pre-assessment stage, how to efficiently and objectively collect and quantify the visual perception data of viewers remains an unresolved technical bottleneck.

[0004] In recent years, eye-tracking technology has shown application potential in fields such as landscape visual quality assessment and built environment perception research due to its ability to record individual visual attention allocation and cognitive processing in real time and objectively. Some scholars have already introduced eye-tracking into the visual evaluation of public spaces in traditional villages, landscape perception in historical districts, and comparative studies of architectural heritage renovation schemes. However, these exploratory studies have remained at the level of using eye-tracking data as experimental indicators for post-hoc analysis, failing to establish a standard evaluation process for practical engineering applications. Specifically, existing research lacks a dimensional framework with "virtual restoration style harmony" as its core objective, thus failing to directly guide the design optimization of restoration schemes; it lacks evaluation algorithms and grading criteria that systematically integrate objective eye-tracking indicators with subjective qualitative evaluation; and it does not address the intelligent closed loop from identifying style conflicts in the assessment to automatically outputting restoration optimization suggestions. In summary, this field urgently needs a style harmony assessment method that deeply integrates eye-tracking with virtual restoration scenarios, combining objective quantitative and subjective perception dimensions. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies in assessing the harmony of village landscape, such as reliance on expert subjective experience, difficulty in objective quantification, lack of public perception perspective, and inadequate virtual pre-assessment methods. This invention provides a virtual restoration landscape harmony assessment method based on eye-tracking. This method integrates objective eye-tracking physiological data with a subjective evaluation system through a complete technical chain of "virtual scene construction—eye-tracking data acquisition—fusion assessment modeling—result output and optimization suggestions." This achieves semi-quantitative assessment and grading of landscape harmony, providing a scientific and visualized decision-making basis for comparing, optimizing, and dynamically correcting restoration plans.

[0006] To achieve the above-mentioned objectives, the present invention provides the following technical solution: This invention provides a method for assessing the harmony of traditional village virtual restoration landscape based on eye tracking, comprising the following steps: Step S1: Construct a 3D model of the virtual restoration scene of traditional villages, and divide multiple areas of interest (AOI) in the virtual environment according to the core dimensions of landscape harmony assessment. Step S2: Organize the subjects to freely browse in the virtual repair scene, and use an eye-tracking device to continuously collect multi-dimensional eye movement data at a preset sampling rate. The eye movement data includes fixation point coordinates, fixation duration, saccade amplitude, pupil diameter and fixation heatmap. Step S3: Extract objective eye movement indicators for each attention area from the eye movement data, and combine them with the subject's subjective rating of appearance coordination to calculate the comprehensive evaluation index of appearance coordination using a preset fusion algorithm; Step S4: Based on the comparison between the comprehensive evaluation index of landscape harmony and the preset level threshold, output the landscape harmony level judgment result of the repair plan and targeted optimization suggestions.

[0007] Furthermore, the virtual restoration scene construction method in step S1 is as follows: based on the real-world 3D data obtained by UAV oblique photography and / or ground laser scanning, a 3D model of the current scene is generated; then, at least one 3D model of a restoration scheme scene is constructed according to the restoration scheme to be evaluated, wherein the restoration scheme scene includes building components, decorative elements, vegetation or the overall building body to be added, renovated or replaced.

[0008] Furthermore, the division of the area of ​​interest in step S1 is based on at least one of the following: the area where the repair element is located, the transition area between the repair element and the original element, the area of ​​the core elements of the original appearance, and the overall scene background area as a visual reference.

[0009] Furthermore, the eye-tracking device mentioned in step S2 is a VR head-mounted display device with a built-in eye-tracking module or a non-invasive telemetry eye tracker, and the preset sampling rate is not less than 60 Hz; the subjects include professionals and / or non-professional members of the general public, and the number of subjects is not less than 30.

[0010] Furthermore, the objective eye movement indicators mentioned in step S3 include at least four of the following: total fixation duration, number of fixations, average fixation duration, first fixation latency, number of regressions, fixation duration percentage, and average pupil diameter.

[0011] Furthermore, the subjective rating in step S3 is collected through a semantic differential scale, and the rating dimensions include at least three of the following: coordination between old and new, consistency of style, harmony of color, matching of materials, appropriateness of scale, and overall aesthetics; the rating uses a Likert scale.

[0012] Furthermore, the fusion algorithm described in step S3 takes the following form: ; in, This is an objective eye-tracking index vector. This is an objective measure of coordination based on objective eye-tracking metrics. For subjective rating vectors, For a subjective measure of coordination based on subjective ratings, This is a measure of the consistency between objective data and subjective ratings. The weighting coefficients are and satisfy the following conditions: The weighting coefficients are determined using the analytic hierarchy process (AHP).

[0013] Furthermore, the aforementioned The calculation method is as follows: after normalizing each objective eye movement index, a weighted sum is performed based on a preset "coordination-eye movement pattern" mapping relationship; the mapping relationship includes at least the following rules: (a) If the total fixation time Exceeding the preset threshold And the number of times viewed Exceeding the preset limit This will lower the objective coordination score; (b) If the average fixation time Within the preset suitable range And the uniformity index of fixation point distribution Exceeding the preset threshold This will improve the objective coordination score; (c) If the change in pupil diameter Exceeding the preset threshold If so, the objective coordination score will be lowered.

[0014] Furthermore, the level determination results in step S4 are divided into five levels: highly coordinated, basically coordinated, slightly incongruous, moderately incongruous, and severely incongruous; the level division is based on the comprehensive evaluation index C and a preset threshold. The comparison, among which : when It was judged to be highly coordinated at that time. when It was determined to be basically coordinated at that time. when It was initially determined to be mild incoordination. when It was determined to be moderate incoordination. when It was determined to be severely incompatible.

[0015] Furthermore, it also includes: overlaying and visualizing eye-tracking heatmaps with 3D models of the restoration plan scene, using color gradients to represent the degree of visual attention concentration, and marking areas of landscape perception conflict in the visualization interface.

[0016] Compared with the prior art, the present invention has the following advantages and beneficial effects: First, it achieves objective quantification and reproducibility of appearance coordination assessment. This invention uses eye-tracking data as the core basis for assessment. By extracting multiple physiological indicators such as fixation duration, number of retrospectives, and pupil diameter changes, and combining them with preset coordination-eye-tracking pattern mapping rules, it transforms the traditional subjective qualitative judgment that relies on individual expert experience into a semi-quantitative assessment based on physiological measurement data. This effectively reduces human subjective bias in the assessment process. The assessment results do not depend on any specific individual expert, have good reproducibility and horizontal comparability, and provide a unified quantitative benchmark for the objective comparison of multiple repair options.

[0017] Secondly, it integrates the dual visual perception perspectives of professionals and the general public. This invention simultaneously collects eye-tracking data from both professionals and the general public, reflecting this data in both the objective indicator extraction and subjective scoring stages. This expands the evaluation of landscape harmony from a single professional paradigm to a multi-faceted perspective of "professional + public," compensating for the shortcomings of existing assessment methods that neglect the real visual experiences of ordinary users and local residents. This allows the assessment conclusions to more comprehensively reflect the rich social connotations of the landscape harmony of traditional villages.

[0018] Third, a closed-loop pre-assessment system for virtual restoration has been established. This invention allows for a coordination pre-assessment of restoration plans in a virtual reality environment before implementation. Through an iterative process of "plan design - virtual assessment - feedback optimization," landscape conflicts can be identified and their severity quantified at the planning stage. This avoids the waste of manpower and resources caused by discovering inconsistencies only after construction is completed and then reworking and modifying the work, which is a common practice in traditional methods. This significantly reduces the risks and trial-and-error costs of traditional village preservation projects.

[0019] Fourth, it provides an intuitive and visual form of assessment. This invention overlays eye-tracking heatmaps with a three-dimensional virtual restoration scene, using color gradients to intuitively present the degree of visual attention concentration and the spatial location of areas of conflict in landscape perception. This allows planning and design personnel and decision-makers to identify problems at a glance, significantly reducing the threshold for understanding and communicating assessment results.

[0020] Fifth, an intelligent decision support chain has been constructed, from problem identification through assessment to automatic output of optimization suggestions. Based on the objective eye-tracking index characteristics of each area of ​​interest and the specific dimensions where subjective scores are low, this invention can automatically generate targeted optimization suggestions for restoration plans, realizing an intelligent closed loop from "problem diagnosis" to "improvement guidance." This provides efficient and operable technical support for the scientific comparison and dynamic correction of traditional village restoration plans. Attached Figure Description

[0021] Figure 1 This is an overall flowchart of the method of the present invention; Figure 2 This is a schematic diagram illustrating the division of areas of interest in a virtual repair scenario; Figure 3 This is a schematic diagram showing the overlay of an eye-tracking gaze heatmap and a 3D restored scene. Detailed Implementation

[0022] The present invention will be further described in detail below with reference to experimental examples and specific embodiments. However, this should not be construed as limiting the scope of the above-mentioned subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0023] This example uses a traditional Baima Tibetan village in the western Sichuan plateau as the assessment object. This village is a typical Baima Tibetan settlement, preserving a complete cluster of "cedar plank houses," a traditional village layout, and unique ethnic cultural landscape, making it a typical representative of traditional villages in western Sichuan's ethnic minority areas. In recent years, some traditional buildings in the village have developed structural safety hazards, and plans are underway for protective restoration and adaptive renovation. A new functional building (hereinafter referred to as the "new building") integrating cultural display and visitor services is proposed to be constructed at the village entrance. The design scheme will use modern techniques to abstractly interpret traditional Baima Tibetan architectural elements. An assessment is now needed regarding the harmony between this restoration plan and the original appearance of the village.

[0024] Example 1 like Figure 1 As shown, this embodiment performs the following four core steps: Step S1: Obtain high-precision 3D spatial data and texture information of the core area (approximately 4.5 hectares) of the traditional village using oblique photography by drones and ground laser scanning. Based on real-scene 3D reconstruction software (e.g., ContextCapture, or domestically developed Smart3D, whose core algorithm is CC-compatible and complies with national geographic information data security requirements), generate a 3D mesh model of the village as the current scene. Simultaneously, based on the architectural plans and material specifications of the visitor service center provided by the design unit, construct a detailed 3D model of the new building in Unreal Engine 5 (UE5) or Unity 3D, and place it in the predetermined position within the current scene to form a virtual scene for the restoration plan.

[0025] In the generated repair scheme scenario, the Area of ​​Interest (AOI) for landscape harmony assessment is divided into 4 AOIs.

[0026] Step S2: A total of 48 participants were recruited, including 22 individuals with backgrounds in architecture / urban planning / cultural heritage conservation and 26 members of the general public. Participants wore VR headsets (in this embodiment, an HTC VIVE ProEye with a built-in Tobii eye-tracking module and a sampling rate of 120Hz) and freely browsed the current scene and the restoration plan scene sequentially, with each scene's browsing time set to 90 seconds. During the browsing process, gaze coordinates, gaze duration, saccade amplitude, and pupil diameter data were continuously collected, and a gaze heatmap was generated in real time.

[0027] Step S3: Extract 7 objective eye-tracking indicators from each AOI; after completing the browsing, the participants subjectively rated the appearance coordination of each AOI across 6 dimensions (7-point Likert scale) in a virtual questionnaire within the VR environment. A pre-set fusion algorithm is used to calculate the comprehensive appearance coordination evaluation index C for each AOI.

[0028] Step S4: Compare the comprehensive evaluation index C of each AOI with the preset level threshold to determine the style harmony level. At the same time, overlay the eye-tracking heatmap with the 3D restoration scene to mark areas of style perception conflict, and automatically generate optimization suggestions for incompatible AOIs, outputting a formatted evaluation report.

[0029] Example 2 The specific construction process of the virtual repair scene in step S1 is as follows: (1) Real-world data acquisition: A DJI M300 RTK drone equipped with a five-lens oblique photography camera was used to plan flight routes in the core area of ​​the target village, achieving a ground resolution of 1.8cm. Simultaneously, a Leica RTC360 ground laser scanner was used to perform close-range scanning and supplementary measurements on the main alleys and key building facades of the village, obtaining detailed point cloud data of the buildings with millimeter-level precision.

[0030] (2) Current scene modeling: The oblique photogrammetry image and ground laser scanning point cloud are imported into the 3D real scene modeling software (e.g., ContextCapture, or similar software such as Smart3D developed domestically, whose core algorithm is compatible with CC and meets the national geographic information data security requirements). After aerial triangulation calculation, point cloud fusion, mesh reconstruction and texture mapping, a high-precision real scene 3D mesh model is generated as the current scene.

[0031] (3) Construction of repair scheme scene and establishment of VR interactive environment: a. Model building and import: Based on the CAD drawings and renderings provided by the design unit, perform basic geometric modeling directly in Unreal Engine 5 (UE5), or use 3ds Max / Maya to build high-precision models (such as the main structure of the building, the fir wood house components with the characteristics of Baima Tibetan ethnic group, decorative textures, etc.), and import the model into UE5 in FBX format without loss.

[0032] b. Scene Assembly and Materials: In UE5, the newly created building model is placed into the existing scene based on the real-world 3D reconstruction. Using UE5's material system, realistic PBR materials such as bricks, wood, and stone are applied to the model according to the restoration plan.

[0033] c. VR Interaction and Eye-Tracking Integration: Configure the VR Pawn in UE5 and bind it to the input of the HTC VIVE Pro Eye headset. Using UE5 blueprints or C++ programming, call the Tobii XR SDK to achieve real-time, high-precision (120Hz sampling rate) acquisition and recording of data such as gaze point, gaze duration, saccades, and pupil diameter.

[0034] d. Scene Deployment and Execution: The completed UE5 project is packaged into a Windows 64-bit executable program. Participants simply need to launch the program and wear a VR headset to freely browse the virtual restoration scene in real-time and immersively. The system automatically records all interactions and eye-tracking data in the background. Compared to traditional offline rendering models, the real-time rendering and interactive environment provided by UE5 greatly enhances the realism and ecological validity of the evaluation.

[0035] Example 3 Figure 2 This embodiment illustrates the division of the area of ​​interest (AOI) in the virtual restoration scene. Based on the core dimensions of landscape harmony assessment, this embodiment divides the scene into four areas of interest (AOI). Each AOI is defined and stored using a set of vertex coordinates of a closed 3D bounding box: AOI-1 (Area where the repair element is located): The area extending 1m outward from the main building of the new visitor service center and its outer outline. This area is the core object of the assessment and directly corresponds to the newly added architectural elements in the repair plan.

[0036] AOI-2 (Transition Area): The space between the new building and the nearest traditional Baima Tibetan wooden house, extending 5m along the opposite sides of the buildings, including the alleyway, paved road surface, and green belt between them. This area is used to assess the quality of the transition between the new and old buildings in terms of spatial scale and visual perception.

[0037] AOI-3 (Existing Feature Core Elements Area): Located along the main line of sight, this area comprises three Ming and Qing dynasty residential courtyards (including gable walls or eaves decorations of the cedar-plank houses, traditional gate towers, and courtyard entrances) that best represent the traditional Baima Tibetan style, approximately 15m to 40m from the new building. This area serves as a reference background for assessing whether the new building interferes with the visual integrity of the existing feature core elements.

[0038] AOI-4 (Visual Reference Background Area): The outline of the mountains behind the village and the ancient camphor tree at the village entrance (approximately 300 years old, with a canopy covering an area of ​​approximately 180m²) serve as the natural background reference for the overall scene.

[0039] The four AOIs mentioned above correspond to four categories of classification criteria: repair elements, transitions, landscape references, and scene backgrounds.

[0040] Example 4 Device Configuration: This embodiment uses the HTC VIVE Pro Eye VR headset, which has a built-in Tobii eye-tracking module and a binocular sampling rate of 120Hz, meeting the requirement of not less than 60Hz in the claims. The gaze direction accuracy is approximately... (In the central region of the field of view) to meet the needs of high-precision eye-tracking data acquisition.

[0041] Participant Recruitment and Grouping: A total of 48 participants were recruited (meeting the requirement of at least 30 participants), including 22 professionals (8 architecture professors and graduate students, 7 urban and rural planning practitioners, and 7 cultural heritage protection practitioners) and 26 non-professional members of the general public (9 local community residents, 17 students from non-related majors, and other members of the public). The participants ranged in age from 18 to 62 years old, with a male-to-female ratio of approximately 1:1.1. All participants had corrected visual acuity of at least 1.0 in both eyes and no color blindness, color weakness, or other visual impairments. Participants did not consume alcohol or take any medications that could affect visual attention within 48 hours prior to the experiment, and maintained normal rest on the day of the experiment.

[0042] Eye-tracking calibration: After wearing the device, the subject first completes a 5-point calibration procedure in the VR environment. The calibration target points are located at the center, upper left, upper right, lower left, and lower right of the virtual field of vision. The system records the calibration accuracy, requiring a root mean square error ≤ 0.5° of visual angle to pass. Those who do not meet the standard are recalibrated until they do.

[0043] Example 5 In step S3, this embodiment extracts all seven objective eye-tracking indicators (meeting the requirement of "at least four") for each AOI to form an objective eye-tracking indicator vector. :

[0044] Taking AOI-1 (newly constructed building area) as an example, the average objective eye movement indexes for the professional group and the non-professional public group are shown in the table below:

[0045] Example 6 In step S3, subjective ratings are collected through a semantic differential scale questionnaire interface built into the VR environment. After the participants have completed free browsing of each scene, the system automatically presents the evaluation questionnaire on the virtual panel corresponding to each AOI.

[0046] The scoring dimensions in this embodiment include all six items below (meeting the requirement of "at least three"): Harmony between New and Old: The degree to which the newly built elements harmonize with the surrounding existing buildings in terms of overall appearance; Stylistic consistency: The degree to which newly constructed elements are consistent with the traditional style of the village in terms of architectural style and design language; Color harmony: The degree of harmony between the color scheme of the newly added elements and the original color scheme of the village; Material matching degree: The degree to which the materials used in the new elements visually match the traditional materials of the village; Appropriateness of scale: The degree to which the size and height of the new element are appropriate in relation to the scale of the surrounding traditional buildings and street spaces; Overall aesthetic appeal: The subject's subjective feeling about the overall visual aesthetics of the area.

[0047] The ratings used a 7-point Likert scale, with 1 point representing "very incongruous / very inappropriate" and 7 points representing "very harmonious / very appropriate". Taking AOI-1 as an example, the mean ratings for each dimension are as follows (standard deviations are in parentheses):

[0048] Subjective evaluation vector The score is composed of the above six dimensions.

[0049] Example 7 In step S3, the comprehensive evaluation index C of landscape harmony is calculated using the following fusion algorithm: ; Meaning of each item and specific values ​​in this embodiment: (1) Objective coordination measurement function : First, the objective eye-tracking indicators are normalized using Min-Max and mapped to the [0, 1] interval.

[0050] (2) Subjective coordination measurement function : The arithmetic mean of the six subjective rating dimensions is taken and normalized to the interval [0, 1]. In this embodiment:

[0051] Taking the AOI-1 non-professional public group as an example, After normalization .

[0052] (3) Consistency measurement function : definition ,in and They are respectively and The value after normalization to the [0,1] interval. This function reflects the degree of consistency between the coordination level indicated by objective eye-tracking data and subjective scores; the smaller the deviation between the two, the closer the h value is to 1.

[0053] (4) Determination of weighting coefficients: This embodiment uses the Analytic Hierarchy Process (AHP) to determine the weights. Five experts with over five years of experience in the field of traditional village preservation and architectural heritage assessment were invited to conduct pairwise comparisons of α, β, and γ using a 1-9 scale to construct a judgment matrix, which was then tested for consistency. Then, the arithmetic mean weights are calculated, and the final result is determined: , , ; Verified to meet the requirements .

[0054] (5) Example of fused computing (AOI-1, non-professional public group): Assuming that it is calculated , ,but:

[0055]

[0056] Example 8 In this embodiment The calculation steps are as follows: Step 1: Indicator Normalization. The seven objective eye-tracking indicators were normalized using Min-Max normalization, with the following formula: ; in, and These represent the minimum and maximum values ​​of this indicator for all subjects across all AOIs, respectively.

[0057] For directional indicators (larger values ​​reflect poorer coordination), reverse the process:

[0058] Step 2: Weighted summation based on the "coordination-eye movement pattern" mapping relationship.

[0059]

[0060] Weights Similarly determined using the AHP method, the values ​​in this embodiment are shown in the table below:

[0061] Step 3: Rule-based adjustments.

[0062] right The preliminary calculation results are then adjusted according to rules: Rule (a): Total fixation duration Exceeding the preset threshold And the number of times viewed Exceeding the preset limit If so, the objective coordination score will be lowered.

[0063] In this embodiment, based on the pre-experiment calibration: Take 1.5 times the average total fixation time of all AOIs (i.e., approximately 24,500 ms). Take 1.5 times the average number of AOI regressions (approximately 15 times). Testing showed that AOI-1 was performed by the non-professional public group. (not exceeding) ), but R = 18 times (exceeding) Meanwhile, the total fixation time has also approached the threshold. Based on the merging judgment rule, for... Apply a penalty coefficient of 0.85.

[0064] Rule (b): Average fixation duration Within the preset suitable range Furthermore, the fixation point distribution uniformity index U exceeds the preset threshold. This will improve the objective coordination score.

[0065] In this embodiment, , (Based on empirical intervals of comfortable gaze duration for architectural scenes in visual cognition research literature). (Uniformity threshold). The average fixation duration of AOI-1 is 381.3ms, which is within the range of [250, 400], but the fixation point distribution uniformity index U is calculated to be 0.48 (below 0.6), which does not meet the reward conditions.

[0066] Rule (c): Change in pupil diameter Exceeding the preset threshold If so, the objective coordination score will be lowered.

[0067] = Current AOI average pupil diameter − Global scene average pupil diameter. In this embodiment... Take 0.15mm. AOI-1 non-professional public group: average pupil diameter in the global scene is approximately 3.76mm, and the average pupil diameter in AOI-1 is 3.94mm. Triggering a penalty, for Apply a penalty coefficient of 0.90.

[0068] Based on the above rules, the initial values ​​after normalization and weighted summation are... After the punishment: After normalization to [0,1] .

[0069] Example 9 In step S4, the landscape harmony level is divided into five levels. This embodiment selects 20 historical case scenarios independently evaluated by experts as the training set, and uses ROC curve analysis to determine the optimal segmentation threshold. The calibration results are as follows: , , ,

[0070] satisfy The constraint relationship.

[0071] The rules for determining the level are as follows:

[0072] The comprehensive evaluation index and grade determination results of the four AOIs in this embodiment are as follows (the weighted average of the professional group and the non-professional public group):

[0073] The assessment results indicate that the new buildings themselves and the transitional areas connecting them with the existing residences have mild issues of style inconsistency, requiring targeted optimization; while the core elements of the original style and the natural background have not been significantly visually disturbed.

[0074] Example 10 (1) Visual display of eye-tracking heatmap overlay.

[0075] Figure 3 This diagram illustrates the overlay of an eye-tracking gaze heatmap and a 3D model of the restoration plan scene. The specific implementation method is as follows: The gaze coordinates of all subjects collected in step S2 are aggregated into a unified world coordinate system, and a kernel density estimation algorithm (with a bandwidth parameter set to 0.5m) is used to generate the gaze density distribution on the scene surface. The generated thermal texture is then mapped onto the surface of the 3D model of the restoration scheme scene for overlay display.

[0076] exist Figure 3 The visualization interface clearly shows that the newly constructed building area in AOI-1 exhibits a significant concentration of warm-colored (red-orange) gaze, particularly at the junction of the roofline of the new building and the skyline of the traditional cedar houses behind it, and at the junction of the second-floor facade and the gable walls of the surrounding residences, forming high-density gaze hotspots (highlighted in the image). The transition area in AOI-2 also shows concentrated gaze characteristics. The system automatically marks areas with gaze density exceeding 1.5 times the global average and a subjective score below 4.0 (out of 7) as areas of conflict in landscape perception (marked with dashed boxes and annotations in the image).

[0077] (2) Targeted optimization suggestions are automatically generated.

[0078] Based on the correspondence between the objective eye-tracking metrics of each AOI and the dimensions with lower subjective scores, the system automatically generates optimization suggestions: AOI-1 optimization recommendations: The subjective rating for "scale appropriateness" was the lowest (mean 2.4 / 7.0). At the same time, the viewing time in this area accounted for as much as 31% (significantly higher than the global reference value of about 20%), the number of times the view was revisited was 18 (exceeding the threshold), and the change in pupil diameter was 0.18mm (exceeding the threshold), indicating that the visual cognitive load was too high. Recommendation: Adjust the new building from 2+1 floors (partial viewing platform) to 2 floors, reduce the building eaves height from 11.5m to 8.5m, which is similar to the surrounding traditional houses, and reduce the volume ratio from about 2.5:1 to about 1.6:1.

[0079] The subjective score for "style unity" was low (mean 3.5 / 7.0), with attention focused on the junction of the new building and the gable or eave decoration of the existing cedar house. Recommendation: Incorporate elements of the gable or eave decoration of the traditional cedar house into the roof outline design of the new building, and adopt a tiered gable design to enhance the connection with the traditional skyline.

[0080] AOI-2 optimization recommendations: The subjective score for "coordination between old and new" was low, and the number of times the view was repeated was high. Recommendation: Add landscape buffer design with traditional elements of Baima Tibetan culture to the transition area to enrich the visual transition of the intermediate level.

[0081] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for evaluating the harmony of traditional village virtual restoration landscape based on eye tracking, characterized in that, Includes the following steps: Step S1: Construct a 3D model of the virtual restoration scene of traditional villages, and divide multiple areas of interest (AOI) in the virtual environment according to the core dimensions of landscape harmony assessment. Step S2: Organize the subjects to freely browse in the virtual repair scene, and use an eye-tracking device to continuously collect multi-dimensional eye movement data at a preset sampling rate. The eye movement data includes fixation point coordinates, fixation duration, saccade amplitude, pupil diameter and fixation heatmap. Step S3: Extract objective eye movement indicators for each attention area from the eye movement data, and combine them with the subject's subjective rating of appearance coordination to calculate the comprehensive evaluation index of appearance coordination using a preset fusion algorithm; Step S4: Based on the comparison between the comprehensive evaluation index of landscape harmony and the preset level threshold, output the landscape harmony level judgment result of the repair plan and targeted optimization suggestions.

2. The method according to claim 1, characterized in that, The virtual restoration scene in step S1 is constructed as follows: based on the real-world 3D data obtained by UAV oblique photography and / or ground laser scanning, a 3D model of the current scene is generated; then, at least one 3D model of a restoration scheme scene is constructed according to the restoration scheme to be evaluated. The restoration scheme scene includes building components, decorative elements, vegetation or the overall building body to be added, renovated or replaced.

3. The method according to claim 1, characterized in that, The division of the area of ​​interest in step S1 is based on at least one of the following: the area where the repaired element is located, the transition area between the repaired element and the original element, the area of ​​the core elements of the original appearance, and the overall scene background area as a visual reference.

4. The method according to claim 1, characterized in that, The eye-tracking device mentioned in step S2 is a VR head-mounted display device with a built-in eye-tracking module or a non-invasive telemetry eye tracker, and the preset sampling rate is not less than 60 Hz; the subjects include professionals and / or non-professional members of the general public, and the number of subjects is not less than 30.

5. The method according to claim 1, characterized in that, The objective eye movement indicators mentioned in step S3 include at least four of the following: total fixation duration, number of fixations, average fixation duration, first fixation latency, number of regressions, fixation duration percentage, and average pupil diameter.

6. The method according to claim 1, characterized in that, The subjective rating in step S3 is collected through a semantic differential scale. The rating dimensions include at least three of the following: coordination between old and new, style unity, color harmony, material matching, scale appropriateness, and overall aesthetics. The rating uses a Likert scale.

7. The method according to claim 1, characterized in that, The fusion algorithm described in step S3 is in the following form: ; in, This is an objective eye-tracking index vector. This is an objective measure of coordination based on objective eye-tracking metrics. For subjective rating vectors, For a subjective measure of coherence based on subjective ratings, This is a measure of the consistency between objective data and subjective ratings. The weighting coefficients are satisfied. The weighting coefficients are determined using the analytic hierarchy process (AHP).

8. The method according to claim 7, characterized in that, The The calculation method is as follows: after normalizing each objective eye movement index, a weighted sum is performed based on a preset "coordination-eye movement pattern" mapping relationship; the mapping relationship includes at least the following rules: (a) If the total fixation time Exceeding the preset threshold And the number of times viewed Exceeding the preset limit This will lower the objective coordination score; (b) If the average fixation time Within the preset suitable range And the uniformity index of fixation point distribution Exceeding the preset threshold This will improve the objective coordination score; (c) If the change in pupil diameter Exceeding the preset threshold If so, the objective coordination score will be lowered.

9. The method according to claim 1, characterized in that, The level determination results in step S4 are divided into five levels: highly coordinated, basically coordinated, slightly incongruous, moderately incongruous, and severely incongruous. The grading is based on the comprehensive evaluation index C and a preset threshold. The comparison, among which : when It was judged to be highly coordinated at that time. when It was determined to be basically coordinated at that time. when It was initially determined to be mild incoordination. when It was determined to be moderate incoordination. when It was determined to be severely incompatible.

10. The method according to claim 1, characterized in that, Also includes: The eye-tracking heatmap is overlaid with a 3D model of the restoration plan scene for visualization. The degree of visual attention concentration is represented by color gradients, and areas of landscape perception conflict are marked in the visualization interface.

Images