A method for simulating and predicting comfort in a high-intensity urban area

Abstract

The application discloses a kind of urban high-strength area comfortable feeling simulation prediction method, it is related to environmental assessment technical field, including obtaining the street view panorama of each observation point, and the space element of each street view panorama is extracted by image semantic recognition technology, and the physiological data of the space element of individual participation should, generate objective perception index and subjective perception index, build comfortable feeling simulation prediction model, establish the mapping relationship of objective perception index and subjective perception index and comfort score, the perception index of each observation point is input model, corresponding comfort score is predicted, and the comfortable feeling distribution map in the study area is generated by space interpolation algorithm.The application can directly present the comfortable feeling distribution situation of high-strength area by establishing the mapping relationship between objective perception index and subjective perception index, by inputting the perception index of observation point into model for prediction, and generating comfortable feeling distribution map in combination with space interpolation algorithm.

Description

Technical Field

[0001] This invention relates to the field of environmental assessment technology, specifically a method for simulating and predicting comfort levels in high-intensity urban areas. Background Technology

[0002] High-intensity urban areas, as important dense public spaces, serve as the primary environmental carriers for residents' daily lives, social interactions, and various commercial activities. The comfort level of these areas directly impacts residents' daily experiences and behavioral patterns, and is closely related to urban social stability, economic development, and urban vitality. Therefore, how to scientifically and comprehensively assess the environmental comfort of high-intensity urban areas has become a crucial issue in urban planning and management.

[0003] Existing research indicates that the relationship between environmental factors and comfort in high-intensity urban areas is complex and multidimensional. In other words, residents' perceived comfort level is influenced not only by visual factors but also by a combination of factors, including the area's spatial layout, traffic planning, and level of green space. However, traditional methods for assessing environmental comfort often rely on user questionnaires, which depend heavily on subjective experience and fail to comprehensively and accurately reflect residents' actual experiences in high-intensity urban areas.

[0004] In recent years, with the rapid development of street view recognition technology, environmental comfort assessment methods based on street view images have gradually attracted attention. This method analyzes various visual features in the street environment, such as street cleanliness, building appearance, green space ratio, and sky visibility, considering the impact of different proportions of visual features on residents' comfort. However, this method is limited to objective environmental evaluation and ignores the psychological and physiological reactions of residents in different spatial environments. Comfort in high-intensity areas is not only affected by objective external environmental factors but also closely related to individual residents' subjective perceptions. Summary of the Invention

[0005] 1) Technical problems to be solved

[0006] This invention provides a method for simulating and predicting comfort in high-intensity urban areas, which can comprehensively consider multi-dimensional assessments of objective environmental factors and individual subjective perceptions.

[0007] (ii) Technical Solution

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for simulating and predicting comfort levels in high-intensity urban areas, comprising:

[0009] Obtain panoramic street view images of multiple observation points in a selected high-intensity area, and extract spatial elements of each panoramic street view image using image semantic recognition technology;

[0010] Eye-tracking technology was used to record the participants' gaze trajectories as they viewed each of the street view panoramas, and the fixation duration for each spatial element was acquired. Simultaneously, the participants' physiological data was obtained. This physiological data included the average power of specific frequency bands of electroencephalograms (EEGs), the baseline level of ground-based spectral reflex (GSR), the mean GSR, and the temporal and frequency domain parameters of human velocities (HRV). The specific frequency bands included... Wave, wave and The frequency band of the wave;

[0011] Based on the gaze duration of each spatial element, calculate its gaze percentage in the street view panorama, select a preset number of spatial elements with the highest gaze duration percentage as relevant spatial elements, and obtain the physiological data of the relevant spatial elements.

[0012] The relevant spatial elements of each acquired street view panoramic image are used as objective perception indicators, and their corresponding physiological data are used as subjective perception indicators.

[0013] A comfort simulation prediction model is constructed, and a mapping relationship is established between the objective perception index, the subjective perception index and the comfort score. The objective perception index and the subjective perception index are used as input variables of the prediction model, and the prediction model is trained by using the subjective scores of participants on the street view panoramic image as the target value, so that it can predict the comfort score of the input variables.

[0014] The input variables of each observation point are obtained, input into the model to predict the corresponding comfort scores, and a comfort distribution map of the study area is generated by spatial interpolation algorithm.

[0015] Furthermore, eye-tracking technology is used to track the participants' eye movements and obtain the gaze duration for each of the spatial elements. Specifically, by recording the gaze points of the participants when viewing the street view panorama, a gaze start time and end time are generated based on temporal information whenever the participants gaze at a spatial element, and the gaze start time and end time are used as the gaze period.

[0016] Furthermore, the gaze capture and the timestamp of the physiological numerical control are synchronized to align the gaze capture data and physiological data. For the gaze period of each spatial element, the average power of the specific frequency band of all EEGs within the gaze period, the baseline level of GSR, the mean of GSR, and the time and frequency domain parameters of HRV are obtained as the corresponding physiological data.

[0017] Furthermore, before constructing the comfort simulation prediction model, the input objective and subjective perception indicators are standardized, and all input data are converted to the same scale to ensure that the numerical ranges of indicators in different dimensions are consistent.

[0018] Furthermore, the comfort prediction model is constructed using an MLP model, and its model structure is as follows:

[0019]

[0020] in, For the predicted comfort score, For activation function, These are standardized objective and subjective perception indicators. As weight, For bias terms;

[0021] In the MLP model, each layer of the network calculates the output through weighted inputs, and a non-linear activation function is applied between each layer. The score of the output layer is the final comfort score prediction value.

[0022] Furthermore, the prediction model is trained using participants' subjective comfort ratings of the street view panorama as the target value, and the weights are optimized using a gradient descent algorithm. and bias terms This minimizes the error between the comfort score of the predictive model and the participants' subjective ratings; specifically,

[0023] During the training of the simulated prediction model, the weight parameters of the prediction model are adjusted by minimizing the loss function. Defined as:

[0024]

[0025] in, The comfort score predicted by the model. Subjective ratings from participants. The number of samples;

[0026] Using the gradient descent algorithm, the comfort simulation prediction model will gradually optimize the weights and biases, so that the error between the predicted comfort score and the subjective score reaches the lowest historical difference.

[0027] Furthermore, after the comfort simulation prediction model is trained, the objective and subjective perception indicators of the new observation points are standardized to form input variables, which are then input into the model for prediction. The output prediction result is the comfort score.

[0028]

[0029] in, and These are the final weights and biases obtained through training.

[0030] Furthermore, the generation of the comfort distribution within the study area using a spatial interpolation algorithm specifically involves:

[0031] Acquire the spatial location and comfort score of multiple observation points in the selected high-intensity area;

[0032] The comfort score and spatial location of each observation point are used as known data points and input into the selected interpolation algorithm to calculate the comfort score of each selected location in the entire area.

[0033] By using spatial interpolation, a continuous comfort score map is obtained, thereby showing the spatial distribution of comfort throughout the high-intensity area.

[0034] (iii) Beneficial effects:

[0035] Compared with the prior art, this invention has the following beneficial effects:

[0036] This invention calculates the percentage of time participants spend gazing at each identified spatial element, filters out the most relevant spatial elements, and combines this with physiological data such as EEG, GSR, and HRV of the participants related to the spatial elements to ensure that the model can effectively capture the impact of spatial elements on comfort and reduce reliance on irrelevant factors.

[0037] By combining objective and subjective perception indicators, the comfort simulation and prediction model can comprehensively reflect the environmental characteristics of high-intensity areas and the physiological responses of participants. By establishing a mapping relationship between objective and subjective perception indicators, it can capture the relationship between environmental characteristics and individual subjective perception within the studied high-intensity areas, avoiding the bias caused by a single indicator or a single perception dimension, thus making the prediction of comfort more comprehensive and accurate. By inputting the perception indicators of observation points into the model for prediction and combining them with spatial interpolation algorithms to generate a comfort distribution map, the comfort distribution in high-intensity areas can be presented intuitively. Attached Figure Description

[0038] Figure 1 This is a flowchart of a method for simulating and predicting comfort in high-intensity urban areas provided by an embodiment of the present invention;

[0039] Figure 2 The flowchart of the method for training the prediction model in the urban high-intensity area comfort simulation and prediction method provided in the embodiment of the present invention;

[0040] Figure 3 The flowchart illustrates the method for generating the spatial distribution of comfort within a high-intensity urban area in a comfort simulation and prediction method provided in an embodiment of the present invention.

[0041] Figure 4This is a schematic diagram illustrating the process of capturing the gaze of participants while they view a panoramic street view image, as provided in an embodiment of the present invention for a method of simulating and predicting comfort in a high-intensity urban area.

[0042] Figure 5 This is a schematic diagram illustrating the acquisition of physiological data while capturing the gaze of participants in a method for simulating and predicting comfort in a high-intensity urban area provided in an embodiment of the present invention.

[0043] Figure 6 This is a schematic diagram of the heart rate variability amplitude measured in a method for simulating and predicting comfort in a high-intensity urban area provided in an embodiment of the present invention. Detailed Implementation

[0044] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0045] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0046] Furthermore, the terms "first" and "second" are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0047] It should be noted that, where there is no conflict, the features in the embodiments of the present invention can be combined with each other.

[0048] In constructing a comfort evaluation index for high-intensity urban areas, the inventors discovered that comfort in high-intensity areas is not only influenced by objective environmental factors but also closely related to individual subjective perceptions. Further research revealed that comfort assessment involves not only a response to visual factors but also more complex psychological and physiological reactions.

[0049] Therefore, to address the inaccuracy of current technical methods relying solely on visual evaluation, a combination of... Figures 1 to 5As shown, this invention provides a method for simulating and predicting comfort in high-intensity urban areas. This method comprehensively considers a multi-dimensional evaluation system that integrates objective environmental indicators and individual subjective perceptions. This method not only overcomes the subjectivity issues in traditional assessments but also, by combining it with physiological data—such as using modern physiological sensing technology to capture residents' psychological changes in different environments, including electroencephalograms (EEGs), skin conductance response (GSR), and heart rate variability (HRV)—provides a deeper understanding of people's reactions to high-intensity urban environments and more accurate subjective perception data for comfort evaluation. For example, a higher GSR value reflects an individual's anxiety or stress, while HRV reflects an individual's emotional state and level of relaxation. Combining these physiological data with objective environmental indicators constitutes a more comprehensive comfort assessment framework, enabling the comfort assessment of high-intensity urban areas to combine environmental characteristics with individual subjective perceptions, providing a more reliable basis for predicting comfort in high-intensity urban areas.

[0050] For details, please refer to Figure 1 First, data on objective environmental factors is acquired, followed by S10: acquiring multiple observation points in the selected high-intensity area, obtaining panoramic street view images of each observation point, and extracting spatial elements from each panoramic street view image using image semantic recognition technology. In this step, multiple typical observation points within the study area are selected. These observation points are typically chosen based on the geographical characteristics, urban structure, population activity density, and areas representative of comfort levels within the high-intensity area. By selecting these observation points that embody the typical environmental characteristics of the high-intensity area, the environmental features of the entire high-intensity area can be comprehensively reflected.

[0051] Regarding the acquisition of street view panoramic images of various observation points, in some embodiments, street view map API tools, such as Google Street View or Baidu Street View, can be used to obtain these panoramic images. Street view panoramic images typically provide a 360-degree panoramic view, which can display comprehensive visual information of the location, such as the presence and location of spatial elements like roads, buildings, greenery, and streets. Furthermore, these acquired street view panoramic images are usually converted into uniform-scale two-dimensional image data, thereby providing comparative information for subsequent environmental analysis, i.e., semantic recognition of the images.

[0052] In summary, it can be understood that image semantic recognition extracts different spatial element features from street view panoramic images by applying deep learning and computer vision techniques. In some embodiments of this invention, deep learning algorithms such as convolutional neural networks (CNNs) are used to train CNN classification models based on semantic segmentation, enabling them to automatically extract different spatial elements from street view panoramic images, such as buildings, sky, roads, streetlights, traffic lanes, plants, motor vehicles, and non-motor vehicles.

[0053] For example, deep learning algorithms, including but not limited to DeepLabV3+ or Mask R-CNN, can be used to perform pixel-level semantic segmentation on street view panoramic images. This identifies the feature characteristics of different regions in the image, thereby classifying spatial elements. The output is the category of each spatial element in the street view panoramic image and its location within the image. For example, the aforementioned spatial elements such as buildings, sky, roads, streetlights, traffic lanes, plants, motor vehicles, and non-motor vehicles are all individually identified and their corresponding location is labeled. Furthermore, by calculating the proportion of these spatial elements in the image, spatial indicators related to environmental comfort can be derived, such as green coverage, sky visibility, and sightline accessibility.

[0054] Assuming a commercial center area is selected as the research object for this high-intensity area, multiple observation points are first determined, such as intersections of commercial streets, plazas in front of major shopping malls, and next to street parks. These points are typical public spaces that can represent the environmental characteristics of the area. 360-degree panoramic street view images of these observation points are obtained through street view map APIs, such as Google Street View. Among them, the intersections of commercial streets provide rich information on roads, buildings, green belts, and streets, while the street view next to street parks provides more information on green elements and open spaces.

[0055] Semantic recognition of these street view panoramic images is performed using deep learning algorithms. Specifically, a pre-trained Mask R-CNN model is used to identify and segment various pre-trained spatial elements in the street view panoramic image. For example, for an image of a commercial street intersection, the Mask R-CNN model is pre-trained to segment and identify spatial elements such as main roads, side roads, pedestrian walkways, buildings, motor vehicles, non-motor vehicles, pedestrians, streetlights, and green trees in the image.

[0056] In summary, automatically extracting spatial elements from panoramic street view images using image semantic recognition technology not only significantly improves data acquisition efficiency but also enhances the accuracy of analysis results through precise identification via deep learning algorithms. Comprehensive analysis of street view images from different observation points in high-intensity areas can accurately reflect the environmental characteristics of those areas, providing a reliable data foundation for comfort simulation and prediction models.

[0057] After acquiring panoramic street view images of the observation points and identifying spatial elements within them, step S20 is performed: Eye-tracking technology is used to record the gaze trajectories of participating individuals as they view each panoramic street view image. Combined with the spatial elements of the gaze, corresponding physiological data such as electroencephalograms (EEGs), ground-level scintillation (GSR), and human gaze velocity (HRV) are obtained. The purpose of this step is to acquire subjective perception data of different individuals regarding the areas where these observation points are located. Eye-tracking technology is used to track the gaze trajectories of participating individuals when viewing each panoramic street view image, and combined with the spatial elements of the participants' gaze, their physiological responses are recorded and analyzed. Eye-tracking technology typically relies on eye-tracking devices, which capture an individual's gaze direction and gaze point by precisely measuring eye movement trajectories.

[0058] In some embodiments, eye-tracking technology utilizes eye sensors on the participant's head-mounted device, such as an infrared camera, to track the movement of the participant's eyes. The eye tracker can precisely record the focusing position of the pupil's movement and combine this with various fixation parameters to create an eye-tracking trajectory. Common parameters include fixation point position, fixation duration, saccade path, and eye movement speed. These parameters allow for the acquisition of the participant's eye-tracking trajectory while viewing a panoramic street view.

[0059] Combining the aforementioned eye-tracking device acquiring the gaze trajectory of participants while viewing the panoramic street view, it can be understood that when participants view the panoramic street view, the eye-tracking device records the spatial elements of their gaze in real time. These spatial elements are extracted using the aforementioned image semantic recognition technology. By correlating the participants' gaze trajectory with these spatial elements, a deeper understanding of the participants' level of attention to different environmental features can be gained. (See also...) Figure 4 The yellow areas in the image represent the areas that a participant gazed at for extended periods while viewing the panoramic street view images from various observation points. It can be understood that if a participant gazes at a particular spatial element for a prolonged period and their physiological data fluctuates, then that spatial element can be considered to have a significant impact on comfort.

[0060] Regarding the collection of physiological data, some embodiments utilize electroencephalography (EEG) technology. By monitoring brainwave activity through electrodes, different frequency bands of EEG signals can be extracted, such as alpha waves (8-13Hz), beta waves (13-30Hz), and theta waves (4-8Hz). These different frequency bands of EEG are closely related to an individual's psychological state. For example, alpha waves are typically associated with relaxation and focused attention, while beta waves are associated with heightened concentration and anxiety. By analyzing the changes in participants' EEG while viewing panoramic street views, the impact of the environment on an individual's psychological state can be assessed.

[0061] In addition, you can refer to Figure 5 This figure illustrates how, while capturing participants' gaze, other physiological data were obtained by measuring conductance of the skin (GSR), body temperature, heart rate, pulse amplitude, and respiration. GSR reflects an individual's physiological response by measuring changes in the skin's electrical conductivity. Emotional states, such as stress, can activate sweat glands and alter skin conductivity; therefore, GSR is an effective tool for monitoring emotional responses. In this work, GSR was used to measure participants' emotional fluctuations while viewing a street scene, such as tension, comfort, or relaxation.

[0062] Heart rate variability (HRV) is obtained by measuring heart rate in individuals. HRV reflects the activity of the autonomic nervous system of the heart. By analyzing the frequency and time-domain indicators of heart rate variability, the physiological stress response and relaxation level of participants can be revealed. (See reference below.) Figure 6 The figure shows the heart rate variability (HRV) of a participant, with HRV values ​​obtained through the RR pulse interval. Lower HRV values ​​are generally associated with higher stress levels or discomfort, while higher HRV indicates a more relaxed state.

[0063] In summary, eye-tracking technology and physiological data collection can provide crucial subjective physiological data for comfort analysis in high-intensity areas. By combining eye-tracking technology with physiological indicators such as EEG, GSR, and HRV, the impact of environmental characteristics on individual physiological and psychological states can be more comprehensively assessed, providing strong data support for subsequent comfort simulation and prediction models.

[0064] After acquiring the eye-tracking and physiological data of the participating individuals, the following steps are performed: S30: Based on the fixation duration of each spatial element, calculate its fixation percentage in the street view panorama, select a specific number of spatial elements with the highest fixation duration percentage as relevant spatial elements, and acquire the physiological data of these relevant spatial elements. And S40: Use the relevant spatial elements of each acquired street view panorama as objective perception indicators, and their corresponding physiological data as subjective perception indicators.

[0065] Regarding the acquisition of fixation duration, in some embodiments, gaze-capture technology, such as the aforementioned eye tracker, is used to track the participants' eye movements and record their fixation points. Whenever a participant fixates on a spatial element, the system generates a fixation start time and end time, typically accurate to the millisecond level, and the same millisecond unit as heart rate. For example, if a participant fixates on a building in a street view panorama for 550 milliseconds, the system records the fixation start and end times and labels this spatial element as "Building A". In this way, the fixation duration for each participant on each spatial element is precisely labeled as a time period.

[0066] By synchronizing timestamps, gaze capture data and physiological data are aligned. It is important to ensure that physiological signals at each moment, such as EEG, GSR, and HRV data, match the participant's gaze activity.

[0067] For example, during EEG acquisition, suppose a time series of data is obtained, timestamped from T1 to T2. Simultaneously, during this time period, the participant is fixating on a spatial element called "Building A." Therefore, the EEG data during this time period is correlated with the fixation data for "Building A." This can be understood as follows: for each fixation period of a spatial element, all data within that time period are retrieved from the physiological data acquisition system. The EEG data, GSR data, and HRV data within that time period are labeled as data belonging to that spatial element. This can be directly matched using timestamps, ensuring that each data point belongs to a specific spatial element.

[0068] Finally, each relevant spatial element and its corresponding physiological data are combined to prepare for subsequent analysis. In some embodiments, a data table is generated after data collection to facilitate the subsequent integration of physiological data from multiple participating individuals. The table columns include:

[0069] Related spatial elements: Spatial element name;

[0070] Timestamp: The start and end times of a period of time for observing related spatial elements;

[0071] Gaze duration: The length of time a participant gazes at the relevant spatial element;

[0072] Physiological data: Physiological data recorded during this time period, namely the average power of the EEG band, the baseline and mean of GSR, and the time and frequency domain parameters of HRV;

[0073] Fixation percentage: The proportion of fixation time for this relevant spatial element to the total fixation time.

[0074] Finally, step S50 is performed: Based on objective and subjective perception indicators, a comfort simulation and prediction model is constructed, and a mapping relationship between objective and subjective perception indicators and comfort scores is established. Specifically, objective and subjective perception indicators are used as input variables of the model, and the subjective comfort scores of participating individuals on the street view panorama are used as target values ​​to establish a mapping relationship between input indicators and comfort scores.

[0075] It is important to note that before constructing a comfort simulation prediction model, the input objective and subjective perception indicators must first be standardized. This is to ensure that indicators of different dimensions are calculated on the same scale, thus preventing certain indicators with larger dimensions from dominating the model training process. In some embodiments, all input data are transformed into a dataset with zero mean and unit variance, ensuring that each input feature contributes to the model training within the same range.

[0076] After standardizing the data, an MLP model is used to construct a comfort prediction model. An MLP is a feedforward neural network with at least one hidden layer, where each layer computes its output through weighted inputs, and a non-linear activation function is applied between each layer. The model structure can be represented as follows:

[0077]

[0078] in, The predicted comfort score, i.e., the model's output, The activation function is a non-linear function; common activation functions include ReLU, Sigmoid, and Tanh. The role of the activation function is to add non-linearity to the output of each neuron in the neural network, increasing the model's expressive power and enabling each layer's activation function to learn complex non-linear mappings.

[0079] This refers to the overall characteristics of standardized objective and subjective perception indicators. As weight, The summation is a weighted input, meaning that for each input feature, that is, for both objective and subjective perception metrics, there will be a corresponding weight. This weight represents the influence of the feature on the prediction result, and can be expressed as a weighted sum of all input features:

[0080]

[0081] in, Indicates the first Each input feature is a standardized objective or subjective perception indicator. It is the first The weights of each input feature, weights It also represents input features. It's important to note that the contribution of comfort prediction to comfort level is optimized during training.

[0082] Bias terms are additional parameters in each layer of a neural network that help the model better fit the data when all input features are used. When the weighted sum is zero, the bias term allows a non-zero output result. The specific data is learned through training, and no value is specified here.

[0083] Overall, It represents the output of a neuron after the activation function has been applied.

[0084] Suppose the input of a certain layer is It is linearly correlated, activation function This linear combination can be transformed into a non-linear result. Non-linear activation functions can help the model learn complex mapping relationships. For example, the ReLU function passes the original value when the input feature value is greater than 0, and outputs 0 when the input feature value is less than 0. This helps the neural network learn the non-linear relationship between different features.

[0085] During the training of an MLP model, it is necessary to adjust the network weights. and bias terms To minimize the difference between predicted and actual values, in some embodiments, a loss function can be used to measure the error between predicted and actual values. The loss function is commonly represented by mean squared error (MSE), and its formula is as follows:

[0086]

[0087] in, The comfort score predicted by the model is the comfort score calculated by the MLP model based on the input features. Rate the actual comfort level. Given the number of samples, during training, the loss function calculates the average error across all samples. It is the total number of samples participating in the training.

[0088] The goal of training a model is to adjust its parameters, i.e., the weights. and bias terms This makes the loss function value To minimize this, gradient descent is employed in some implementations.

[0089] First, input data using the current weights and biases. Calculate the predicted value Then calculate the loss function. Loss function The gradient is the derivative of the loss function with respect to the weights and biases; it is also called the partial derivative. The gradient represents how sensitive the loss function is to changes in the parameters. Specifically, each parameter, i.e., the weight, needs to be calculated. and bias terms The contribution to the loss function, i.e., the gradient of the loss function with respect to the parameters:

[0090]

[0091]

[0092] The gradient is calculated, and the weights and biases are updated using the gradient descent algorithm. The gradient descent algorithm adjusts the parameters according to the gradient value, gradually reducing the loss function value, so that the predicted value is as close as possible to the actual value.

[0093] After training, a final set of weights and biases is obtained, which are the optimal parameters obtained through continuous optimization during training. For new observation points, namely new street view panoramas and their spatial elements' objective and subjective perception data, this data can be input into the trained MLP model for prediction, outputting the corresponding comfort score:

[0094]

[0095] in, The final weights obtained through training, This is the final bias term obtained through training.

[0096] The process described above involves using a trained MLP model to predict the comfort level corresponding to a new street view panorama and its spatial elements. This process includes standardizing the input data, calculating a weighted sum, generating a prediction result through an activation function, and outputting a new comfort score. The trained weights and biases help the model map different perceptual indicators to the final comfort score, thus providing effective comfort predictions for new observation points.

[0097] In summary, the overall methodology for simulating and predicting comfort levels in high-intensity urban areas can be understood as follows: First, multiple observation points are selected within the chosen high-intensity urban area, and a panoramic street view image is acquired for each point. This panoramic image is a comprehensive view from a specific observation point, realistically reflecting the visual characteristics of the street environment. Using image semantic recognition technology, spatial elements are extracted from each panoramic street view image. Different spatial elements have varying impacts on participants' emotions and comfort levels. Eye-tracking technology is used to record participants' gaze trajectories while viewing each panoramic street view image, calculating the gaze duration for each spatial element. Simultaneously, physiological data is collected. This physiological data reflects the participants' psychological and physiological states when viewing different spatial elements, such as relaxation, tension, or pleasure.

[0098] Because high-intensity streetscape panoramas contain various spatial elements, and not all spatial elements significantly affect participants' emotions and comfort, analyzing the gaze duration of each spatial element allows us to identify the elements with the greatest impact on participants as relevant spatial elements. Specifically, based on the gaze duration of each spatial element, its gaze percentage in the streetscape panorama is calculated, and the spatial elements with the highest gaze duration percentages are selected as relevant spatial elements. These spatial elements, due to their longer gaze duration, are more likely to have a significant impact on participants' emotions and comfort. Understandably, the reason for using gaze duration as a selection criterion for relevant spatial elements is that the length of gaze time is usually related to a person's interest in or emotional impact on an object. For example, participants might pay more attention to a green space than tall buildings because green spaces are more relaxing. Furthermore, selecting relevant spatial elements avoids processing too much redundant information, highlighting the spatial elements most important for comfort prediction, and improving the model's efficiency and accuracy.

[0099] Based on the collected spatial elements and corresponding physiological data, objective and subjective perception indicators were constructed. The objective perception indicators consist of the attribute values ​​of the relevant spatial elements, while the subjective perception indicators consist of the physiological data of participants when viewing the relevant spatial elements. The objective and subjective perception indicators describe comfort from the perspectives of environmental characteristics and human physiological responses, respectively.

[0100] We employ an MLP model as the prediction tool, using objective and subjective perception indicators as input variables, and participants' subjective comfort ratings of the street view panorama as the target value. MLP is a non-linear model, capable of handling complex non-linear relationships between input features and target values. For example, the impact of green space ratio on comfort may not be a simple linear relationship, but rather exhibits a threshold effect. This threshold effect can be understood as the perception of comfort only occurring when the green space in an area exceeds a certain area's coverage.

[0101] By establishing a mapping relationship between input indicators and comfort scores during the training process, as described above, the following can be summarized: using objective and subjective perception indicators as input variables, and the participants' comfort scores for the street view panorama as the target value, a mapping relationship between input variables and comfort scores is constructed. The model parameters are then adjusted using the backpropagation algorithm to make the predicted values ​​as close as possible to the true values.

[0102] The trained MLP model can take into account the perception indicators of new observation points and predict their corresponding comfort scores. Since the number of observation points in a target high-intensity area is limited and cannot completely cover the entire study area, spatial interpolation is used in some embodiments. Common spatial interpolation methods, such as Kriging interpolation or IDW interpolation, are used. This involves extrapolating the scores of other locations in the study area based on the known comfort scores of the observation points. This can be understood as generating a comfort distribution map of the entire study area through spatial interpolation algorithms, providing a reference for urban planning.

[0103] Finally, S60 is performed: the perception indicators of each observation point are input into the model to predict the corresponding comfort scores, and a comfort distribution map of the study area is generated using a spatial interpolation algorithm. The spatial interpolation algorithm is used to estimate and generate the comfort distribution of other areas based on known comfort score data points.

[0104] In some embodiments of the present invention, comfort scores at known observation points are extrapolated to unknown points using an inverse distance weighting method. The formula for the inverse distance weighting method is:

[0105]

[0106] in: It is the point to be predicted The comfort score, It is an observation point The comfort score, It is an observation point To the point to be predicted distance, It is the distance attenuation parameter, which controls the degree to which distance affects the prediction results.

[0107] After predicting comfort scores for multiple observation points, a spatial interpolation algorithm is used to predict the comfort score for each location within the study area. Then, a comfort distribution map of the entire area is generated based on these predicted scores. This map can be displayed as an image or map, and different color codes can be used to represent the comfort levels of different areas. Assuming 10 observation points are selected within the study area, and the objective and subjective perception indices for each point are recorded, these data are standardized and input into a trained MLP model to predict the comfort score for each observation point.

[0108] Interpolation is performed using the inverse distance weighted method. Assume that for a certain point... There are multiple observation points around it ( , , ..., Each observation point has a corresponding comfort score. The predicted comfort value for that point is calculated using a formula. .

[0109] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. The scope of patent protection of the present invention shall be determined by the claims. Similarly, any equivalent structural changes made based on the description and drawings of the present invention shall also be included within the scope of protection of the present invention.

Claims

1. A method for simulating and predicting the comfort of a high-density urban area, characterized by, The application comprises the following steps: Obtain the street view panorama of multiple observation points in the selected high-intensity area, and extract the spatial elements of each street view panorama through image semantic recognition technology; The line-of-sight capture technology is used to record the line-of-sight trajectory of the participants watching each street view panorama and obtain the gaze duration of each spatial element, while physiological data of the participants are obtained; the physiological data includes the average power of specific frequency bands of brain waves, the baseline level of GSR, the mean value of GSR, and the time domain and frequency domain parameters of HRV, wherein the specific frequency bands include alpha wave, beta wave and gamma wave frequency bands. According to the gaze duration of each spatial element, calculate its gaze proportion in the street view panorama, and select the top pre-set number of spatial elements with the highest gaze duration proportion as relevant spatial elements, and obtain the physiological data of the relevant spatial elements; Take the relevant spatial elements of each street view panorama as objective perception indicators, and take the corresponding physiological data as subjective perception indicators; The comfort feeling simulation prediction model is constructed to establish a mapping relationship between the objective perception index, the subjective perception index and the comfort feeling score; wherein the objective perception index and the subjective perception index are taken as input variables of the prediction model, and the MLP model is used to construct the prediction model, and the model structure is as follows: ; wherein, is a predicted comfort feeling score, is an activation function, is a standardized input variable, is a weight, is a bias term; in the MLP model, each layer of the network calculates the output by weighting the input, and a nonlinear activation function is applied between each layer, and the subjective score of the participant on the street view panorama is taken as a target value to train the prediction model, so that the comfort feeling score of the input variable is predicted; the gradient descent algorithm is used to optimize the weight and the bias term , so that the error between the comfort feeling score of the prediction model and the subjective score of the participant is minimized; Obtain the input variables of each observation point, input the model to predict the corresponding comfort score, and generate a comfort distribution map in the study area through spatial interpolation algorithm.

2. The urban high intensity zone comfort simulation and prediction method according to claim 1, characterized in that, The gaze tracking technology is used to track the eye movement trajectory of the participants, and the gaze duration of each spatial element is obtained. Specifically, by recording the gaze points of the participants watching the street view panorama, a gaze start time and end time are generated based on the time sequence information every time the participant gazes at a spatial element, and the gaze start time and end time are taken as the gaze period.

3. The urban high intensity zone comfort simulation prediction method according to claim 2, characterized in that, The time stamps of the gaze capture and the physiological data are synchronized, so that the gaze capture data and the physiological data are aligned. For the gaze period of each spatial element, the average power of all brain waves in the specific frequency range, the baseline level of GSR, the mean value of GSR, and the time domain and frequency domain parameters of HRV are obtained as the corresponding physiological data.

4. The urban high-intensity zone comfort simulation and prediction method according to claim 1, characterized in that, Before constructing the comfort simulation prediction model, the objective perception indicators and subjective perception indicators are standardized, and all input data are converted to the same scale to ensure that the numerical ranges of different dimension indicators are consistent.

5. The urban high intensity zone comfort simulation and prediction method according to claim 1, wherein, The comfort subjective score of the participants on the street view panorama is used as the target value to train the prediction model. Specifically, In the process of training the simulation prediction model, the weight parameters of the prediction model are adjusted by using the method of minimizing the loss function, and the loss function is defined as: wherein, is the model predicted comfort score, is the participant subjective score, is the sample number; Through the gradient descent algorithm, the comfort simulation prediction model will gradually optimize the weight and bias term, so that the error between the predicted comfort score and the subjective score reaches the historical minimum difference.

6. The urban high intensity zone comfort simulation prediction method according to claim 5, wherein, After the comfort simulation prediction model is trained, the objective perception indicators and subjective perception indicators of the new observation points are standardized to form input variables, which are input into the model for prediction. The output prediction result is the comfort score: wherein, and are the resulting weights and bias terms from training.

7. The urban high intensity zone comfort simulation prediction method according to claim 1, wherein, The comfort distribution in the study area is generated through the spatial interpolation algorithm. Specifically, Obtain the spatial positions and comfort scores of multiple observation points in the selected high-intensity area; Take the comfort score and spatial position of each observation point as known data points, input them into the selected interpolation algorithm, and calculate the comfort score of each selected position in the entire area; Through spatial interpolation, a continuous comfort score map is obtained, showing the spatial distribution of comfort in the entire high-intensity area.

Images