Signal system, method and application for urban street view visual sentiment prediction

Abstract

The present disclosure provides a signal system, method and application for urban street view visual emotion prediction, comprising a monitoring module, a processing module, a first construction module, a second construction module and an output module; the monitoring module collects the brain electrical signals and emotional evaluation reaction indicators when the subjects watch the street view images; the processing module obtains the relative power of the brain electrical signals in the single-image presentation segment and the multi-image sequence segment; the first construction module uses the relative power of the brain electrical signals in the multi-image sequence segment and the emotional evaluation reaction indicators to construct a brain electrical signal emotion prediction model, which is applied to the brain electrical signals in the single-image presentation segment to obtain the emotional response of any street view image; the second construction module is used for constructing a machine learning model to obtain a street view image emotion prediction model based on the brain electrical signal recognition, and is applied to the street view images in the output module to obtain a city street view visual emotion map; the present disclosure has the advantages of sensitive expression and accurate emotion capture.

Description

Technical Field

[0001] This disclosure relates to the field of brainwave processing technology, and in particular to a signal system, method and application for visual emotion prediction of urban street scenes based on brainwave signals and image recognition. Background Technology

[0002] Electroencephalogram (EEG) is a method of recording brain activity using electrophysiological indicators. It is formed by summing the postsynaptic potentials that occur synchronously among a large number of neurons during brain activity. It records the changes in electrical waves during brain activity and is a comprehensive reflection of the electrophysiological activity of brain nerve cells on the surface of the cerebral cortex or scalp.

[0003] As a comprehensive carrier of urban form and human activities, the visual environment quality of urban streetscapes is directly related to residents' physical and mental health and spatial perception efficiency. Existing research shows that scientific streetscape design can effectively improve the livability of urban spaces and have a positive impact on residents' mental health. However, with the acceleration of urbanization and the surge in building density, the proportion of original natural landscapes has significantly decreased. Streetscape environments are characterized by dense building facades, fragmented green spaces, and excessive artificial colors, leading to a decline in residents' visual comfort and frequent occurrences of related psychological problems. Urban designers need accurate resident visual-emotional evaluation models to improve urban streetscapes. Against this backdrop, constructing accurate visual environment-emotional response correlation models has become an urgent need for humanistic urban design.

[0004] Accurately capturing emotional states under the influence of the urban visual environment is crucial for model construction. Existing research assesses urban residents' emotional evaluation levels through online questionnaires or text analysis surveys. However, traditional assessment methods mainly rely on subjective evaluation systems, which suffer from technical bottlenecks such as insensitivity to emotional experience expression, inaccurate emotional capture, and low processing efficiency.

[0005] Therefore, there is an urgent need to develop an emotion prediction system based on EEG signals that is sensitive to expression and accurately captures emotions under the influence of urban visual environment. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this disclosure provides a signal system, signal processing method, and application based on electroencephalogram (EEG) signals, which aims to solve at least one technical problem in the background art.

[0007] The technical solution adopted in this disclosure is:

[0008] A signal system for predicting visual sentiment in urban street scenes, based on electroencephalogram (EEG) signals and image recognition, includes:

[0009] The acquisition module is used to acquire any street view image and obtain the street view environment indicators corresponding to the street view image; the street view environment indicators include: building density, green view rate, openness, number of colors and color entropy; wherein, building density, green view rate and openness are obtained through semantic segmentation model, and number of colors and color entropy are obtained through image color features;

[0010] The monitoring module interacts with the acquisition module to collect the EEG signals and emotional evaluation response indicators of the subjects when viewing the street view images. This includes testing the subjects under at least two conditions, with four street view image groups played in each condition. A fourteen-channel EEG acquisition device is used for EEG signal measurement, and the Discrete Affected Scale (DAS) and the SAM scale are used to record the emotional response for each street view group. The street view image groups consist of street view images of the same urban environment type.

[0011] The processing module interacts with the monitoring module to perform signal preprocessing and segmentation on the EEG signal, and obtains the power spectral density and the relative power of the full-time EEG signal. This includes: importing EEG information, filtering, removing artifact signals, and independent component analysis. The segmentation divides the full-time EEG signal into single-image presentation segments and multi-image sequence segments in the time domain. The power spectral density is obtained by fast Fourier transform to obtain the relative power of the EEG signals in the single-image presentation segments and multi-image sequence segments.

[0012] The first construction module interacts with the processing module and the monitoring module to construct an EEG signal emotion prediction model using the relative power of the EEG signals of the multi-image sequence segments and the emotion evaluation response index, thereby obtaining the emotion response of any street view image. The module includes: applying the EEG signal emotion prediction model to the EEG signals of the single-image presentation segment to obtain the emotion response of each street view image; and the EEG signal emotion prediction model is constructed using a random forest algorithm.

[0013] The second construction module interacts with the first construction module and the acquisition module to construct a machine learning model based on the emotional response of the street view image and the street view environment indicators, thereby obtaining a street view image emotion prediction model based on EEG signal recognition; the machine learning model is constructed using the gradient boosting tree algorithm.

[0014] The output module interacts with the second construction module to obtain scalar values ​​of different emotional response indicators corresponding to each street scene image based on the street scene image sentiment prediction model, and uses ArcGIS Pro to draw them on a map to obtain a visual sentiment map of the city street scene.

[0015] The step of obtaining the street view environment indicators corresponding to the street view image includes:

[0016] Based on the geographic information system platform, spatial sampling points were set at road network nodes in the study area using the equidistant sampling method, and the sampling density was set to a preset interval.

[0017] Street view images were acquired at each sampling point coordinates from four azimuth angles: due north, due east, due south, and due west. The image acquisition parameters were set to a 90° horizontal viewing angle and a 1024×800 pixel resolution. The shooting time and latitude and longitude data were automatically recorded during the image acquisition process.

[0018] The steps for obtaining the building density, green view rate, openness, number of colors, and color entropy include:

[0019] The building density is the ratio of the number of building element pixels to the total number of image pixels, and the specific calculation formula is as follows:

[0020] ;

[0021] In the formula, The building density of the street view image. The number of pixels in the building. The total number of pixels in the image;

[0022] The green visibility rate is the ratio of the number of pixels of green plant elements to the total number of pixels in the image, and the specific calculation formula is as follows:

[0023] ;

[0024] In the formula, The green view ratio of the street view image. The number of pixels for trees. This represents the number of pixels representing vegetation. This represents the number of pixels in the grass area. The number of pixels for the palm tree. For the number of pixels in the flower, The total number of pixels in the image;

[0025] The openness is the ratio of the number of sky element pixels to the total number of image pixels, and the specific calculation formula is as follows:

[0026] ;

[0027] In the formula, OP represents the openness of the street view image. The number of pixels in the sky. The total number of pixels in the image;

[0028] The image color feature calculation includes: using OpenCV and CuPy libraries to read the RGB values ​​of the street view image and calculate the color entropy and color quantity;

[0029] The color entropy is calculated based on the information entropy theory, and the specific calculation formula is as follows:

[0030] ;

[0031] In the formula, Represents color entropy. Indicates color The probability in the image, Represents the logarithm to the base 2. Indicates the quantity of different colors;

[0032] The number of colors is calculated by determining the RGB values ​​of the colors and counting the number of unique colors. The specific calculation formula is as follows:

[0033] ;

[0034] In the formula, Indicates the number of colors. E represents the color value in the image, and E represents the color entropy.

[0035] The collection of electroencephalogram (EEG) signals and emotional evaluation indicators of subjects while viewing the street view images includes:

[0036] The EEG experiment was designed using the Latin square balance scheme, and four types of environmental stimuli were set up: building clusters, green spaces, open fields of vision, and color entropy clusters. For each type, typical and representative street scene images were selected to form a visual stimulus library.

[0037] The Latin square design is used to balance the presentation order. The images in each block are randomly ordered. The visual stimulus presentation time is preset. Before the stimulus presentation, there is a preset time period of gray blank screen to eliminate visual persistence and adjust emotions, while recording EEG signals.

[0038] Simultaneous recording of EEG signals, with recording frequency bands including the five internationally recognized frequency bands Delta, Theta, Alpha, Beta, and Gamma;

[0039] Visual stimuli were presented using an electronic display, with no other visual distractions in the participants' field of vision except for the electronic display; after each type of street scene image was completed, the participants' discrete emotion rating scale and SAM scale were collected.

[0040] The discrete emotion rating scale and SAM scale collected from the subjects include:

[0041] The discrete emotion rating scale includes twelve emotion response indicators with a 5-level rating: excitement, calmness, tranquility, worry, happiness, sadness, relaxation, depression, fatigue, tension, sadness, and joy; where 1 represents no emotion response of this type, 5 represents a strong emotion response of this type, and the numbers represent an increasing relationship of emotion response.

[0042] The SAM scale uses a 9-level facial expression chart for emotional evaluation, representing valence and the degree of arousal response;

[0043] Record all behavioral responses and synchronize and save the EEG data via wireless or wired communication.

[0044] The step of preprocessing and segmenting the EEG signal, and calculating the power spectral density to obtain the relative power of the EEG signal for a single-image presentation segment and a multi-image sequence segment includes:

[0045] A standard 10-20 electrode positioning system was used to spatially register the 14-channel signals.

[0046] FIR bandpass filters are used to filter out frequency bands below 0.5Hz and above 60Hz, and notch filters are used to eliminate AC frequency band interference in the 48-52Hz band.

[0047] Abnormal signal segments were manually marked by examining time-domain spectrograms, and artifact signal segments such as eye movement and electromyography were manually removed. Furthermore, independent component analysis was used to decompose signal components, and artifact components were manually identified and eliminated.

[0048] The segmentation process divides the EEG signal into segments in the time domain according to the image group and the visual stimulus time period of each street view image, dividing it into single-image presentation segments and multi-image sequence segments in terms of time domain features.

[0049] The single-image presentation segment captures the time window from the start of the image stimulus to the end of the image, while the multi-image sequence segment captures the continuous signal during the presentation of the entire set of images.

[0050] The power spectrum analysis transforms the time-domain features of single-image presentation segments and multi-image sequence segments of EEG signals into frequency-domain features using the Fast Fourier Transform method. Fourteen electrodes from four brain regions are used: frontal lobe: AF3, AF4, F3, F4, F7, F8, FC5, FC6; temporal lobe: T7, T8; parietal lobe: P7, P8; occipital lobe: O1, O2.

[0051] Obtain the five feature bands: Delta, Theta, Alpha, Beta, and Gamma.

[0052] The relative power value is normalized according to the formula, as shown below: ;

[0053] In the formula, This represents the absolute power for each frequency band. The absolute power is calculated across the entire frequency band from 0.5 to 60 Hz; the final generated feature matrix contains the EEG signal data structure for each segment across 5 frequency bands and 14 channels.

[0054] The step of constructing an EEG signal emotion prediction model using the relative power of the multi-image sequence segment EEG signals and the emotion evaluation response index to obtain the emotion response of any street view image includes:

[0055] Using the EEG power of multi-image sequence segments as features and the emotional response indicators recorded in the experiment as labels, the robustness of the model is enhanced by injecting Gaussian noise and feature scaling. At the same time, combined features are constructed by expanding second-order polynomial features, and PCA is used to reduce the dimensionality while retaining 95% of the variance. Then, multi-objective feature selection is performed based on mutual information to select the top 50 principal component features with the most discriminative power. Random forest algorithm is used for regression modeling.

[0056] Hyperparameter search is performed using the TPE algorithm based on Optuna, and random forest is used to adjust tree depth and number of split samples. Furthermore, the macro-average F1 score of five-fold cross-validation is used as the optimization objective, and the accuracy of the validation set is monitored simultaneously to obtain the final retained model components. The final retained model components include: feature processor, label encoder, and independent classifiers for each sentiment target.

[0057] The EEG emotion prediction model is applied to EEG signals in a single image presentation segment. During deployment, the pre-trained processing pipeline is automatically loaded, and the following process is performed on the input data: the feature processor automatically applies the same polynomial expansion and PCA transformation; the feature subset selected during training is dynamically matched; multi-emotion prediction is performed in parallel, and the results are restored to the original scale values ​​through inverse label encoding; finally, the output contains the original metadata and the predicted values ​​of each emotion dimension.

[0058] The step of constructing a machine learning model based on the emotional response of the street view image and the street view environment indicators to obtain a street view image emotion prediction model based on EEG signal recognition includes:

[0059] Street view environmental indicators from street view images are used as features, and the corresponding predicted sentiment ratings are used as labels. In the data preprocessing stage, a robust normalization process is adopted to handle one-dimensional outliers in the environmental indicators and to perform Winsorize truncation using 5%-95% quantiles. The mean and covariance matrix are calculated using the MinCovDet algorithm to achieve Mahalanobis distance-based normalization transformation. The training data is enhanced by adding 5% Gaussian noise.

[0060] The machine learning model uses the GradientBoostingRegressor framework from Scikit-learn, based on the Gradient Boosting Decision Tree algorithm. It automatically searches for key parameters, including n_estimators, learning_rate, and max_depth, through Bayesian optimization. The optimization objective is the R² score of cross-validation. Each target variable is trained independently with its own model, and the optimal parameter combination is saved. The training process is monitored through a visual learning curve.

[0061] The machine learning model evaluation phase employs K-fold cross-validation, records the fluctuations of MAE and R² indicators, and finally outputs the prediction error distribution of each target variable on the test set. During deployment, the preprocessor and model parameters are automatically loaded, the same standardized process is performed on the input data, batch predictions are made, and a result analysis report containing 95% confidence intervals is output. Furthermore, the machine learning model is saved in pkl format, and each sentiment is saved as a corresponding file.

[0062] A method for predicting visual sentiment in urban street scenes, based on the aforementioned signal system for predicting visual sentiment in urban street scenes, includes:

[0063] Acquire any street view image and obtain the corresponding street view environment indicators; the street view environment indicators include building density, green view rate, openness, number of colors and color entropy; wherein, building density, green view rate and openness are obtained through semantic segmentation model, and number of colors and color entropy are obtained through image color features;

[0064] The study collected the subjects' electroencephalogram (EEG) signals and emotional evaluation response indicators while they viewed the street view images. The test included four scenarios, with four street view image groups played in each scenario. A 14-channel EEG acquisition device was used for EEG signal measurement, and the Discrete Affected Scale (DAS) and the SAM scale were used to record the emotional response for each street view group. The street view image groups consisted of street view images of the same type of urban environment.

[0065] The EEG signal is preprocessed and segmented, and power spectral density is calculated to obtain the relative power of the EEG signal over the entire time span. This includes: importing EEG information, filtering, removing artifact signals, and independent component analysis. The segmentation divides the EEG signal into single-image presentation segments and multi-image sequence segments in the time domain. The power spectral density is obtained by fast Fourier transform to obtain the relative power of the EEG signal in the single-image presentation segment and the multi-image sequence segment.

[0066] An EEG signal emotion prediction model is constructed using the relative power of the EEG signals of the multi-image sequence segments and the emotion evaluation response index to obtain the emotion response of any street view image. This includes: applying the EEG signal emotion prediction model to the EEG signals of the single-image presentation segment to obtain the emotion response of each street view image; and the EEG signal emotion prediction model is constructed using the random forest algorithm.

[0067] A machine learning model is constructed based on the emotional response of the street view image and the street view environmental indicators to obtain a street view image emotion prediction model based on EEG signal recognition; and the machine learning model is constructed using the gradient boosting tree algorithm.

[0068] Based on the street view image to be identified, the street view image sentiment prediction model is used to obtain scalars of different sentiment response indicators corresponding to each street view image, and ArcGIS Pro is used to draw them on a map to obtain a visual sentiment map of the city street view.

[0069] An application based on the above method includes:

[0070] When initializing the EEG signal emotion prediction model, the preprocessor and the trained gradient boosting tree model are automatically loaded.

[0071] During prediction, numerical type validation and missing value handling are performed on the input data, robust standardization is based on MinCovDet, and then independent models for each target variable are called for parallel prediction.

[0072] The street scene environment indicators of all street scene images in the study area are processed in batches. The average value of the street scene environment indicators in four directions is calculated for each coordinate point as the feature data of that point, and the predicted values ​​of 14 emotional response indicators corresponding to each street scene image are obtained.

[0073] Import a CSV file containing sentiment indicators into ArcGIS Pro, set the latitude and longitude fields to generate a temporary point layer, and export it as a formal point feature class; use the spatial connection tool to set the road network data as the target feature and the sentiment point data as the connection feature, set the matching method, and retain all sentiment indicator fields in the field mapping.

[0074] After completing the data connection, the rendering method is selected in the symbol system, and the field is specified as the sentiment index; and a sentiment raster surface is generated to identify statistically significant sentiment clusters. The results are represented by gradient colors to indicate high / low value clustered road segments, resulting in a visual sentiment map of the urban street scene in the study area.

[0075] The beneficial effects of this disclosure are:

[0076] The signal system described in this disclosure collects street view images based on urban geographic information big data. It calculates visual environment indicators such as building density, green view rate, and openness using a pixel-level semantic segmentation model, and obtains color quantity and color entropy parameters by combining image color feature analysis. Next, multi-band EEG signals are collected from subjects viewing street view images through an EEG experiment, simultaneously recording subjective emotional responses from a discrete emotion rating scale and a SAM scale. Then, the EEG signals undergo preprocessing such as FIR filtering, artifact removal, and ICA component analysis, segmenting them into single-image presentation segments and multi-image sequence segments. Power spectral density characteristics of each frequency band are calculated using Fast Fourier Transform. A random forest nonlinear regression model is constructed from the multi-image sequence segments to map EEG signals to emotion scores, which is then applied to the single-image presentation segments for emotion prediction. Finally, a machine learning prediction model for street view environment indicators and emotional responses is established based on the gradient boosting tree algorithm, enabling visual emotion assessment of large-scale street view images. This disclosure provides an objective and quantitative visual emotion assessment and prediction tool for urban landscape design by integrating brainwave signals with artificial intelligence technology. It helps to create a living environment that promotes mental health and promotes the practical application of environmental psychology in smart city construction. It has the advantages of being sensitive to expression and accurately capturing emotions. Attached Figure Description

[0077] Figure 1 This is a system block diagram of the signal system described in this disclosure;

[0078] Figure 2 This is a flowchart for acquiring street view images and calculating street view environmental indicators as disclosed in this disclosure;

[0079] Figure 3 Experimental flowchart designed for this disclosure;

[0080] Figure 4 The twelve discrete emotion and two-dimensional emotion evaluation criteria designed for this disclosure;

[0081] Figure 5 This discloses the electroencephalogram (EEG) signal measurement process and frequency domain analysis method.

[0082] Figure 6 This is a flowchart of the training and prediction process for the machine learning model disclosed herein;

[0083] Figure 7 This is a flowchart of the signal processing method described in this disclosure. Detailed Implementation

[0084] To better understand the above-mentioned objectives, features, and advantages of this disclosure, the disclosure will be further described below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other. The disclosure will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of this disclosure and should not be used to limit the scope of protection of this disclosure.

[0085] As is well known, electroencephalogram (EEG) technology, with its millisecond-level temporal resolution and sensitivity in detecting cortical potentials, provides a new technical approach for the objective quantification of street scene visual stimuli. By analyzing the power spectral density of characteristic frequency bands, a mapping relationship between EEG signals and emotional states can be established, which brings a breakthrough to the scientific assessment of the street scene visual environment.

[0086] In view of this, this disclosure also discloses an embodiment: Specific Implementation Example 1:

[0088] like Figure 1 A signal system for predicting visual emotions in urban street scenes, based on electroencephalogram (EEG) signals and image recognition, includes: an acquisition module 100, a monitoring module 200, a processing module 300, a first construction module 400, a second construction module 500, and an output module 600. The acquisition module 100 acquires any street scene image and obtains corresponding street scene environmental indicators. The monitoring module 200 interacts with the acquisition module 100 to acquire EEG signals and emotional evaluation response indicators of subjects viewing the street scene image. The processing module 300 interacts with the monitoring module 200 to preprocess and segment the EEG signals, calculate power spectral density, and obtain single-image presentation segments and multi-image presentation segments. The first construction module 400 interacts with the processing module 300 and the monitoring module 200 to construct an EEG signal emotion prediction model using the relative power of the multi-image sequence EEG signals and the emotion evaluation response index, thereby obtaining the emotion response of any street view image; the second construction module 500 interacts with the first construction module 400 and the acquisition module 100 to construct a machine learning model based on the emotion response of the street view image and the street view environment index, thereby obtaining a street view image emotion prediction model based on EEG signal recognition; the output module 600 interacts with the second construction module 500 to output a city street view visual emotion map based on the street view image to be identified through the street view image emotion prediction model.

[0089] Figure 2This flowchart illustrates the process of acquiring street view images and calculating street view environmental indicators in this embodiment. Based on a geographic information system platform, spatial sampling points are set at road network nodes in the study area using an equidistant sampling method, with a sampling density of 100-meter intervals. By calling the Baidu Street View Map API service, street view images are acquired at four azimuth angles at each sampling point: 0° (due north), 90° (due east), 180° (due south), and 270° (due west). The image acquisition parameters are set to a 90° horizontal viewing angle and a 1024×800 pixel resolution to closely match the human eye's viewing angle. During image acquisition, the shooting time and latitude / longitude data are automatically recorded.

[0090] The SegNext semantic segmentation model, based on the Transformer neural network architecture, achieves an mIoU parameter of 43.45 on the ADE20K dataset. Trained for image parsing, this model accurately identifies pixel regions of elements such as buildings, trees, vegetation, grassland, palm trees, flowers, and sky, used to calculate street scene environmental indicators such as building density, green view ratio, and openness. Trees can be arborescent trees or shrubs, but do not include palm trees; flowers refer to ornamental flowering plants.

[0091] The building density mentioned above is the ratio of the number of building element pixels to the total number of image pixels. The specific calculation formula is as follows: ;

[0092] In the formula, The building density of the street view image. The number of pixels in the building. The total number of pixels in the image;

[0093] Green view rate is the ratio of the number of pixels of green plant elements to the total number of pixels in the image. The specific calculation formula is as follows:

[0094] ;

[0095] In the formula, The green view ratio of the street view image. The number of pixels for trees. This represents the number of pixels representing vegetation. This represents the number of pixels in the grass area. The number of pixels for the palm tree. For the number of pixels in the flower, The total number of pixels in the image;

[0096] Openness is the ratio of the number of sky element pixels to the total number of image pixels, and the specific calculation formula is as follows: ;

[0097] In the formula, OP represents the openness of the street view image. The number of pixels in the sky. The total number of pixels in the image;

[0098] By using OpenCV's RGB color space conversion and CuPy acceleration, image color feature parameters are extracted to calculate the number of colors and color entropy.

[0099] (1) Color entropy is the information entropy value calculated based on the RGB color histogram, which reflects the complexity of color distribution. The specific calculation formula is as follows: ;

[0100] In the formula, Represents color entropy. Indicates color The probability in the image, Represents the logarithm to the base 2. Indicates the quantity of different colors;

[0101] (2) The number of colors is the number of unique hues after discretization of the hue channel H values ​​in the RGB space, quantized in 8 bits. The specific calculation formula is as follows: ;

[0102] In the formula, Indicates the number of colors. E represents the color value in the image, and E represents the color entropy.

[0103] All indicator calculation results are stored in the spatial database along with the geographic coordinates.

[0104] In this embodiment, the electroencephalogram (EEG) experiments used to collect EEG signals and emotional response indicators from subjects while they viewed street scene images included: Figure 3 This is the experimental flowchart for this embodiment. The experimental design adopts a Latin square balance scheme, setting four types of environmental stimuli: building clusters (building density > 60%), green space (green view rate > 60%), open view (openness > 60%), and color entropy clusters (color entropy > 9.0). For each type, 72 representative street scene images were carefully selected to form the stimulus library. The Latin square design balances the presentation order, with images within each block randomly ordered. The visual stimulus presentation time is 10 seconds, followed by a 15-second gray blank screen to eliminate visual persistence and adjust emotions.

[0105] The Emotiv EPOC× wireless EEG acquisition system, a head-mounted electroencephalogram (EEG) device, was used. The system had 14 channels arranged according to the international 10-20 system, with a sampling rate of 128Hz, to simultaneously record EEG signals. The experimental environment was a temperature-controlled, unoccupied room. Visual stimuli were presented on a 120Hz refresh rate electronic display, and there were no other visual distractions in the subject's field of vision besides the electronic display.

[0106] Figure 4For the twelve discrete emotions and two-dimensional emotion ratings in this embodiment, after each street scene image type, the participants need to complete the following:

[0107] (1) Discrete affective assessment - using a 1-5 level scalar scale from 12 affective dimensions, including: excitement, calmness, tranquility, worry, happiness, sadness, relaxation, depression, fatigue, tension, sadness, and joy; select the degree of agreement, 1 represents no emotional response of this type, 5 represents a strong emotional response of this type, and the numbers represent an increasing relationship of emotional response;

[0108] (2) SAM scale - a 9-point rating system was used on two dimensions: valence and arousal. The image that best reflects the current situation was selected for the rating. All behavioral responses were recorded through the E-Prime 2.0 system, and the EEG data was strictly synchronized via Bluetooth and stored on a computer.

[0109] The experimental procedure included: a start screen, visual stimuli, and an end screen. The start screen provided experimental instructions and an informed consent form. The visual stimuli consisted of four sets of images designed using a Latin square balance. After each set of images was played, the electronic display screen asked participants to fill out an emotional evaluation form. The end screen indicated the end of the experimental procedure. The final dataset consisted of raw EEG signals with temporal domain features, in .edf format.

[0110] In this embodiment, the preprocessing, segmentation, and power spectral density calculation of the EEG signal include: using the MATLAB R2023b runtime environment, importing the raw EEG data in .edf format using the EEGLAB toolbox, and spatially registering the 14-channel signal using a standard 10-20 electrode positioning system. The EEG signal is filtered using filters: (1) an FIR bandpass filter (0.5-60Hz) filters low-pass and high-frequency noise; (2) a 50Hz power frequency notch filter (with a bandwidth of ±2Hz) eliminates AC frequency band interference, and a time-domain spectrum is generated through EEGLAB's Scroll channel activities, with abnormal signal segments manually marked. Independent component analysis is performed using the ICA algorithm, and artifact components are automatically detected using the ADJUST plugin. The segmentation is based on the image stimulus presentation time recorded during the experiment. The single-image presentation segment is cut from the time window from the start of the image stimulus to the end of the image, and the multi-image sequence segment is cut from the continuous signal during the presentation of the entire group of images; each group contains 18 street scene images, with a duration of 180s.

[0111] Figure 5The EEG signal measurement process and frequency domain analysis method in this embodiment are described. Frequency domain feature extraction uses Hanning window weighted FFT transform for fourteen electrodes in four brain regions, including: frontal lobe: AF3, AF4, F3, F4, F7, F8, FC5, FC6; temporal lobe: T7, T8; parietal lobe: P7, P8; occipital lobe: O1, O2. A total of five feature frequency bands are calculated: Delta (frequency: 0.5-4Hz), Theta (frequency: 4-8Hz), Alpha (frequency: 8-13Hz), Beta (frequency: 13-30Hz), and Gamma (frequency: 30-60Hz).

[0112] The relative power value is normalized according to the formula, as shown below: ;

[0113] In the formula, This represents the absolute power for each frequency band. The absolute power is calculated across the entire frequency band from 0.5 to 60 Hz. The final generated feature matrix contains the EEG signal data structure for each segment across 5 frequency bands and 14 channels.

[0114] In this embodiment, building a model to predict the emotional response to a single street view image includes: Figure 6 This is a flowchart illustrating the machine learning model training and prediction process in this embodiment. The flowchart shows the process of constructing an EEG emotion prediction model using EEG data and emotion evaluation, and then applying this model to street view images and street view environmental indicators for machine learning training to construct a street view image emotion prediction model. The following are the steps for constructing the EEG emotion prediction model:

[0115] 1. Data Preprocessing and Feature Engineering: EEG power of multi-image sequence segments was used as a feature, and corresponding emotional response indicators recorded in the experiment were used as labels. Gaussian noise injection (μ=0, σ=0.05) and feature scaling with random coefficients of 0.95-1.05 were employed to enhance model robustness. Combined features were constructed through second-order polynomial feature expansion, and PCA dimensionality reduction was used to retain 95% of the variance. Multi-objective feature selection was performed based on mutual information (MI), selecting the top 50 principal component features with the highest discriminative power for each emotional label. For discrete emotional labels, a random forest algorithm was used.

[0116] 2. Multi-objective model optimization: Hyperparameter search is performed using the TPE algorithm from Optuna, a model parameter optimization tool, with parameters such as 40 iterations, a random forest adjusting tree depth of 3-10, and a split sample number of 2-10. The macro-average F1 score from cross-validation is used as the optimization objective, while validation set accuracy is monitored simultaneously. The final retained model components include: a feature processor, a label encoder, and independent classifiers for each sentiment objective. The feature processor incorporates normalization, multinomial expansion, and PCA; the label encoder supports floating-point-safe encoding.

[0117] 3. Prediction System Implementation: Upon deployment, the pre-trained processing pipeline is automatically loaded. For input data, such as 70-dimensional features, the following process is executed:

[0118] (1) The feature processor automatically applies the same polynomial extension and PCA transformation;

[0119] (2) Dynamically match the feature subset selected during training, and skip invalid features to allow for error correction;

[0120] (3) Perform multi-emotion prediction in parallel, and restore the original scale values ​​through inverse label encoding.

[0121] The system output includes raw metadata and predicted values ​​for each sentiment dimension, and supports batch prediction in CSV format.

[0122] The following are the steps to build a street view image sentiment prediction model:

[0123] Street view environmental indicators from street view images are used as features, and the corresponding predicted sentiment ratings are used as labels. In the data preprocessing stage, a robust normalization process is employed to handle one-dimensional outliers in five environmental indicators, including building density, and Winsorize truncation is performed using 5%-95% quantiles. The mean and covariance matrices are calculated using the MinCovDet algorithm to achieve Mahalanobis distance-based normalization. Gaussian noise with a standard deviation of 0.05 is injected during the data augmentation stage.

[0124] Model training utilizes the GradientBoostingRegressor framework from Scikit-learn, based on the Gradient Boosting Decision Tree algorithm. Key parameters are automatically searched using Bayesian optimization, including n_estimators (50-300), learning_rate (0.01-0.3), and max_depth (3-10). The optimization objective is the R² score of cross-validation. A dedicated model is trained for each target variable, and the optimal parameter combination is saved. The training process is monitored through visual learning curves.

[0125] During the model evaluation phase, K-fold cross-validation was employed to record fluctuations in MAE and R² indices. Finally, the prediction error distribution for each label was output on the test set, and the optimal model was saved as the street view image sentiment prediction model. The model was saved in PKL format, with each sentiment index saved as a separate file. During deployment, the preprocessor and model parameters were automatically loaded, and the same standardization process was applied to the input data to batch predict the sentiment ratings of street view images in the study area.

[0126] The signal system described in this embodiment constructs a scientific street scene image emotion prediction model by combining the objective mapping relationship between electroencephalogram (EEG) signals and emotional responses with street scene environmental feature analysis. This overcomes the subjective limitations of traditional questionnaire surveys and achieves cross-scale analysis from microscopic neural signals to macroscopic urban emotion maps. The generated emotion map can intuitively display the emotional distribution characteristics of urban space, providing a quantitative basis for urban planning and landscape design. Specific Implementation Example 2:

[0128] This disclosure provides an embodiment:

[0129] like Figure 7 A method for predicting visual emotion in urban street scenes, based on the signal system described in Specific Embodiment 1, is used to predict visual emotion in urban street scenes by collecting electroencephalogram (EEG) signals, including:

[0130] S100: Acquire street view images of the target area and calculate street view environmental indicators;

[0131] S200 collects brainwave signals and emotional response indicators of subjects when they view street scene images through electroencephalography (EEG) experiments.

[0132] S300 performs preprocessing, segmentation, and power spectral density calculation on EEG signals;

[0133] S400, constructing a nonlinear regression model to predict the emotional response to a single street view image;

[0134] S500, establishes a machine learning prediction model for street view environmental indicators and emotional responses;

[0135] S600 uses predictive models to create urban street view sentiment maps.

[0136] The specific steps are as follows: First, acquire any street view image and obtain the corresponding street view environment indicators. Second, acquire the electroencephalogram (EEG) signals and emotional evaluation response indicators of the subject when viewing the street view image. Third, perform signal preprocessing and segmentation on the EEG signals, and calculate the power spectral density to obtain the relative power of EEG signals in single-image presentation segments and multi-image sequence segments. Fourth, construct an EEG signal emotion prediction model using the relative power of the multi-image sequence segment EEG signals and the emotional evaluation response indicators to obtain the emotional response to any street view image. Fifth, construct a machine learning model based on the emotional response to the street view image and the street view environment indicators to obtain a street view image emotion prediction model based on EEG signal recognition. Sixth, output a city street view visual emotion map based on the street view image to be identified using the street view image emotion prediction model. Specific Implementation Example 3:

[0138] This disclosure also provides an embodiment:

[0139] The method of using the prediction model described in Specific Implementation 1 to generate a visual sentiment map of urban street scenes includes:

[0140] During initialization, the preprocessor and the trained gradient boosting tree model are automatically loaded. During prediction, the input data is first subjected to strict numerical type validation and missing value handling. Robust normalization based on the MinCovDet anomaly detection algorithm, consistent with the training phase, is applied. The saved mean and covariance matrices are used. Then, the independent models for each sentiment classification are called for parallel prediction.

[0141] This system batch processes street view environmental indicators from all street view images within the study area. For each coordinate point, it calculates the mean of the street view environmental indicators in four directions as the feature data for that point. After correcting outliers in the feature data, it generates prediction results. The output file retains the original data and adds a column of predicted values. It also provides descriptive statistics and 95% confidence interval analysis. All outliers are accompanied by detailed error logs. The system obtains predicted values ​​for 14 sentiment response indicators for each street view image.

[0142] Import the CSV file containing sentiment indicators into ArcGIS Pro 3.1.5, set the latitude and longitude fields to generate a temporary point layer, and export it as a formal point feature class. Then use the spatial connectivity tool to set the road network data as the target feature and the sentiment point data as the connectivity feature, with the matching method set to "nearest". In the field mapping, retain all sentiment indicator fields.

[0143] After completing the data connection, the "gradient color" rendering method was selected in the symbology system. The value field was specified as the sentiment index, using a blue-red gradient color band, where red represents a high arousal value and green represents a low arousal value. The "polyline to raster" tool was used to generate a sentiment raster surface. Finally, the "hotspot analysis" tool was used to identify statistically significant sentiment clusters (p < 0.05). The results were represented by a red-blue gradient to indicate high / low value clustered road segments, resulting in a street view image sentiment map of the study area.

[0144] The application described in this embodiment innovatively introduces electroencephalogram (EEG) signal analysis technology into the field of urban environmental emotional response evaluation, breaking through the limitations of traditional subjective questionnaire survey methods. It establishes an EEG-based emotional prediction model and a street view image-based emotional prediction model, significantly improving the scientific rigor and reliability of the evaluation results. Furthermore, through advanced machine learning algorithms, it achieves automated and batch processing of urban street view emotional evaluation, changing the inefficient traditional manual point-by-point assessment mode and providing a scientific method for emotional mapping. This embodiment, through its standardized analysis process, exhibits strong repeatability. EEG experimental measurements enable rapid processing and analysis of urban street view data at different scales, helping to reduce errors in subjective human perception and achieving seamless expansion from individual streets to entire urban areas. This provides innovative technical support and decision-making basis for creating human-centered urban spaces.

[0145] Although embodiments of the present disclosure have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the present disclosure, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A signal system for predicting visual emotion in urban street scenes, based on electroencephalogram (EEG) signals and image recognition, characterized in that, include: The acquisition module is used to acquire any street view image and obtain the street view environment indicators corresponding to the street view image; The street view environmental indicators include: building density, green view rate, openness, number of colors, and color entropy; among which, building density, green view rate, and openness are obtained through semantic segmentation model, and number of colors and color entropy are obtained through image color features. The monitoring module interacts with the acquisition module to collect the EEG signals and emotional evaluation response indicators of the subjects when viewing the street view images. This includes: testing the subjects under at least two conditions, playing four street view image groups in each condition, measuring EEG signals using a 14-channel EEG acquisition device, and using the Discrete Affect Scale (DAS) and the SAM scale to record the emotional response for each street view group; the street view image groups consist of street view images of the same urban environment type. The processing module interacts with the monitoring module to perform signal preprocessing and segmentation on the EEG signal, and obtains the power spectral density and the relative power of the full-time EEG signal. This includes: importing EEG information, filtering, removing artifact signals, and independent component analysis. The segmentation divides the full-time EEG signal into single-image presentation segments and multi-image sequence segments in the time domain. The power spectral density is obtained by fast Fourier transform to obtain the relative power of the EEG signals in the single-image presentation segments and multi-image sequence segments. The first construction module interacts with the processing module and the monitoring module to construct an EEG signal emotion prediction model using the relative power of the EEG signals of the multi-image sequence segments and the emotion evaluation response index, thereby obtaining the emotion response of any street view image. The module includes: applying the EEG signal emotion prediction model to the EEG signals of the single-image presentation segment to obtain the emotion response of each street view image; and the EEG signal emotion prediction model is constructed using a random forest algorithm. The second construction module interacts with the first construction module and the acquisition module to construct a machine learning model based on the emotional response of the street view image and the street view environment indicators, thereby obtaining a street view image emotion prediction model based on EEG signal recognition; the machine learning model is constructed using the gradient boosting tree algorithm. The output module interacts with the second construction module to obtain scalar values ​​of different emotional response indicators corresponding to each street scene image based on the street scene image sentiment prediction model, and uses ArcGIS Pro to draw them on a map to obtain a visual sentiment map of the city street scene.

2. The signal system for visual emotion prediction in urban street scenes according to claim 1, characterized in that, The step of obtaining the street view environment indicators corresponding to the street view image includes: Based on the geographic information system platform, spatial sampling points were set at road network nodes in the study area using the equidistant sampling method, and the sampling density was set to a preset interval. Street view images were acquired at each sampling point coordinates from four azimuth angles: due north, due east, due south, and due west. The image acquisition parameters were set to a 90° horizontal viewing angle and a 1024×800 pixel resolution. The shooting time and latitude and longitude data were automatically recorded during the image acquisition process.

3. The signal system for visual emotion prediction in urban street scenes according to claim 1, characterized in that, The steps for obtaining the building density, green view rate, openness, number of colors, and color entropy include: The building density is the ratio of the number of building element pixels to the total number of image pixels, and the specific calculation formula is as follows: ； In the formula, BD represents the building density of the street view image. The number of pixels in the building. The total number of pixels in the image; The green visibility rate is the ratio of the number of pixels of green plant elements to the total number of pixels in the image, and the specific calculation formula is as follows: ； In the formula, The green view ratio of the street view image. The number of pixels for trees. This represents the number of pixels representing vegetation. This represents the number of pixels in the grass area. The number of pixels for the palm tree. For the number of pixels in the flower, The total number of pixels in the image; The openness is the ratio of the number of sky element pixels to the total number of image pixels, and the specific calculation formula is as follows: ； In the formula, OP represents the openness of the street view image. The number of pixels in the sky. The total number of pixels in the image; The image color feature calculation includes: using OpenCV and CuPy libraries to read the RGB values ​​of the street view image and calculate the color entropy and color quantity; The color entropy is calculated based on the information entropy theory, and the specific calculation formula is as follows: ； In the formula, Represents color entropy. Indicates color The probability in the image, Represents the logarithm to the base 2. Indicates the quantity of different colors; The number of colors is calculated by determining the RGB values ​​of the colors and counting the number of unique colors. The specific calculation formula is as follows: ； In the formula, Indicates the number of colors. E represents the color value in the image, and E represents the color entropy.

4. The signal system for visual emotion prediction in urban street scenes according to claim 1, characterized in that, The collected EEG signals and emotional evaluation indicators of the subjects while viewing the street view images include: The EEG experiment was designed using the Latin square balance scheme, and four types of environmental stimuli were set up: building clusters, green spaces, open fields of vision, and color entropy clusters. For each type, typical and representative street scene images were selected to form a visual stimulus library. The Latin square design is used to balance the presentation order. The images in each block are randomly ordered. The visual stimulus presentation time is preset. Before the stimulus presentation, there is a preset time period of gray blank screen to eliminate visual persistence and adjust emotions, while recording EEG signals. Simultaneous recording of EEG signals, with recording frequency bands including the five internationally recognized frequency bands Delta, Theta, Alpha, Beta, and Gamma; Visual stimuli were presented using an electronic display, with no other visual distractions in the participants' field of vision except for the electronic display; after each type of street scene image was completed, the participants' discrete emotion rating scale and SAM scale were collected.

5. The signal system for visual emotion prediction in urban street scenes according to claim 4, characterized in that, The discrete emotion rating scale and SAM scale collected from the subjects include: The discrete emotion rating scale includes twelve emotion response indicators with a 5-level rating: excitement, calmness, tranquility, worry, happiness, sadness, relaxation, depression, fatigue, tension, sadness, and joy; where 1 represents no emotion response of this type, 5 represents a strong emotion response of this type, and the numbers represent an increasing relationship of emotion response. The SAM scale uses a 9-level facial expression chart for emotional evaluation, representing valence and the degree of arousal response; Record all behavioral responses and synchronize and save the EEG data via wireless or wired communication.

6. The signal system for visual emotion prediction in urban street scenes according to claim 1, characterized in that, The process of preprocessing and segmenting the EEG signal, obtaining the power spectral density, and acquiring the relative power of the full-time EEG signal includes: importing EEG information, filtering, removing artifact signals, and independent component analysis; the segmentation divides the full-time EEG signal into single-image presentation segments and multi-image sequence segments in the time domain; the power spectral density is obtained by using Fast Fourier Transform to acquire the relative power of the EEG signals in the single-image presentation segments and multi-image sequence segments, including: A standard 10-20 electrode positioning system was used to spatially register the 14-channel signals. FIR bandpass filters are used to filter out frequency bands below 0.5Hz and above 60Hz, and notch filters are used to eliminate AC frequency band interference in the 48-52Hz band. Abnormal signal segments were manually marked by examining time-domain spectrograms, and ocular movement and electromyography artifact signal segments were manually removed. Furthermore, independent component analysis was used to decompose signal components, and artifact components were manually identified and eliminated. The segmentation process divides the EEG signal into segments in the time domain according to the image group and the visual stimulus time period of each street scene image, and divides the EEG signal into single image presentation segments and multi-image sequence segments in the time domain. The single-image presentation segment captures the time window from the start of the image stimulus to the end of the image, while the multi-image sequence segment captures the continuous signal during the presentation of the entire set of images. Power spectrum analysis transforms the time-domain features of EEG signals from single-image presentation segments and multi-image sequence segments into frequency-domain features using the Fast Fourier Transform method. Fourteen electrodes from four brain regions are used: frontal lobe: AF3, AF4, F3, F4, F7, F8, FC5, FC6; temporal lobe: T7, T8; parietal lobe: P7, P8; occipital lobe: O1, O2. Obtain the five feature bands: Delta, Theta, Alpha, Beta, and Gamma. The relative power value is normalized according to the formula, as shown below: ; In the formula, This represents the absolute power for each frequency band. The absolute power is calculated across the entire frequency band from 0.5 to 60 Hz; the final generated feature matrix contains the EEG signal data structure for each segment across 5 frequency bands and 14 channels.

7. The signal system for visual emotion prediction in urban street scenes according to claim 1, characterized in that, The step of constructing an EEG signal emotion prediction model using the relative power of the multi-image sequence segment EEG signals and the emotion evaluation response index to obtain the emotion response of any street view image includes: Using the EEG power of multi-image sequence segments as features and the emotional response indicators recorded in the experiment as labels, the robustness of the model is enhanced by injecting Gaussian noise and feature scaling. At the same time, combined features are constructed by expanding second-order polynomial features, and PCA is used to reduce the dimensionality while retaining 95% of the variance. Then, multi-objective feature selection is performed based on mutual information to select the top 50 principal component features with the most discriminative power. Random forest algorithm is used for regression modeling. Hyperparameter search is performed based on Optuna's TPE algorithm, and random forest is used to adjust tree depth and number of split samples. Furthermore, the macro-average F1 score of five-fold cross-validation is used as the optimization objective, and the validation set accuracy is monitored simultaneously to obtain the final retained model components. The final retained model components include: feature processor, label encoder, and independent classifiers for each emotion target. The EEG signal emotion prediction model is applied to EEG signals in a single image presentation segment. During deployment, the pre-trained processing pipeline is automatically loaded, and the following process is performed on the input data: the feature processor automatically applies the same polynomial expansion and PCA transformation; the feature subset selected during training is dynamically matched; multi-emotion prediction is performed in parallel, and the results are restored to the original scale values ​​through inverse label encoding; finally, the output contains the original metadata and the predicted values ​​of each emotion dimension.

8. The signal system for visual emotion prediction in urban street scenes according to claim 1, characterized in that, The step of constructing a machine learning model based on the emotional response of the street view image and the street view environment indicators to obtain a street view image emotion prediction model based on EEG signal recognition includes: Street view environmental indicators from street view images are used as features, and the corresponding predicted sentiment ratings are used as labels. In the data preprocessing stage, a robust normalization process is adopted to handle one-dimensional outliers in the environmental indicators and to perform Winsorize truncation using 5%-95% quantiles. The mean and covariance matrix are calculated using the MinCovDet algorithm to achieve Mahalanobis distance-based normalization transformation. The training data is enhanced by adding 5% Gaussian noise. The machine learning model uses the GradientBoostingRegressor framework of Scikit-learn, which is based on the gradient boosting tree algorithm. It automatically searches for key parameters, including n_estimators, learning_rate, and max_depth, through Bayesian optimization. The optimization objective is the R² score of cross-validation. Each target variable is trained independently with its own model and the optimal parameter combination is saved. The training process is monitored through a visual learning curve. The machine learning model evaluation phase employs K-fold cross-validation, records the fluctuations of MAE and R² indicators, and finally outputs the prediction error distribution of each target variable on the test set. During deployment, the preprocessor and model parameters are automatically loaded, the same standardized process is performed on the input data, batch predictions are made, and a result analysis report containing 95% confidence intervals is output. Furthermore, the machine learning model is saved in pkl format, and each sentiment is saved as a corresponding file.

9. A method for predicting visual sentiment in urban street scenes, based on the signal system for predicting visual sentiment in urban street scenes as described in any one of claims 1-8, characterized in that, include: Acquire any street view image and obtain the corresponding street view environment indicators; the street view environment indicators include building density, green view rate, openness, number of colors and color entropy; wherein, building density, green view rate and openness are obtained through semantic segmentation model, and number of colors and color entropy are obtained through image color features; The study collected the subjects' electroencephalogram (EEG) signals and emotional evaluation response indicators while they viewed the street view images. The test included four scenarios, with four street view image groups played in each scenario. A 14-channel EEG acquisition device was used for EEG signal measurement, and the Discrete Affected Scale (DAS) and the SAM scale were used to record the emotional response for each street view group. The street view image groups consisted of street view images of the same type of urban environment. The EEG signal is preprocessed and segmented, and power spectral density is calculated to obtain the relative power of the EEG signal over the entire time span. This includes: importing EEG information, filtering, removing artifact signals, and independent component analysis. The segmentation divides the EEG signal into single-image presentation segments and multi-image sequence segments in the time domain. The power spectral density is obtained by fast Fourier transform to obtain the relative power of the EEG signal in the single-image presentation segment and the multi-image sequence segment. An EEG signal emotion prediction model is constructed using the relative power of the EEG signals of the multi-image sequence segments and the emotion evaluation response index to obtain the emotion response of any street view image. This includes: applying the EEG signal emotion prediction model to the EEG signals of the single-image presentation segment to obtain the emotion response of each street view image; and the EEG signal emotion prediction model is constructed using the random forest algorithm. A machine learning model is constructed based on the emotional response of the street view image and the street view environmental indicators to obtain a street view image emotion prediction model based on EEG signal recognition; and the machine learning model is constructed using the gradient boosting tree algorithm. Based on the street view image to be identified, the street view image sentiment prediction model is used to obtain scalars of different sentiment response indicators corresponding to each street view image, and ArcGIS Pro is used to draw them on a map to obtain a visual sentiment map of the city street view.

10. An application based on the method as described in claim 9, characterized in that, include: When initializing the EEG signal emotion prediction model, the preprocessor and the trained gradient boosting tree model are automatically loaded. During prediction, numerical type validation and missing value handling are performed on the input data, robust standardization is based on MinCovDet, and then independent models for each target variable are called for parallel prediction. The street scene environment indicators of all street scene images in the study area are processed in batches. The average value of the street scene environment indicators in four directions is calculated for each coordinate point as the feature data of that coordinate point, and the predicted values ​​of 14 emotional response indicators corresponding to each street scene image are obtained. Import a CSV file containing sentiment indicators into ArcGIS Pro, set the latitude and longitude fields to generate a temporary point layer, and export it as a formal point feature class; Use the spatial connection tool to set the road network data as the target element and the sentiment point data as the connection element, set the matching method, and retain all sentiment indicator fields in the field mapping. After completing the data connection, the rendering method is selected in the symbol system, and the field is specified as the sentiment index; and a sentiment raster surface is generated to identify statistically significant sentiment clusters. The results are represented by gradient colors to indicate high / low value clustered road segments, resulting in a visual sentiment map of the urban street scene in the study area.

Images