Environmental perception evaluation method and system based on skin electric signals, terminal and storage medium

Abstract

The invention discloses an environmental perception evaluation method and system based on a skin electric signal, a terminal and a storage medium, and the method comprises the steps: obtaining panoramic image data and geographic information data, carrying out the environmental feature extraction of the panoramic image data and the geographic information data, and obtaining a target feature, inputting the target features into a structured database for storage to obtain an environment feature database; classifying the environment feature database data to obtain a target street image and a skin electric signal, and inputting the target street image and the skin electric signal into a perception prediction model for training to obtain a trained perception prediction model; and obtaining to-be-detected image data, inputting the to-be-detected image data into the trained perception prediction model for prediction, and obtaining a pressure perception prediction result. According to the method, the perception prediction model is trained according to the target street image and the skin electric signal, the influence of the psychological stress is perceived and predicted, and accurate assessment of the spatial psychological stress is realized.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an environmental perception assessment method, system, terminal, and computer-readable storage medium based on electrodermal signals. Background Technology

[0002] Existing methods for assessing street-level built environment perception vary. Some rely on expert experience to determine assessment indicators and results, while others use questionnaires to evaluate the characteristics of the built environment from citizens' perspectives. Assessments based on expert experience fail to reflect citizens' true perceptions or differentiate between the perception characteristics of different groups. Questionnaire-based methods generally include street intercept surveys and video surveys, but these still have some problems. The former's results may be affected by natural physical environmental factors such as temperature, humidity, and lighting, and may also vary depending on the respondent's purpose, process, and companions. The latter mostly presents the built environment through images, with a few studies using virtual reality. However, such studies may have reliability issues due to the influence of the respondents' state. Furthermore, the way the questionnaire questions about perception may lead to different interpretations by different individuals, thus this method may also have validity issues. The latter method also cannot accurately reflect people's perceptions of the built environment.

[0003] Existing methods for assessing perceptions of the built environment in streets mainly include: relying on expert experience to determine assessment indicators and results, and conducting surveys of citizens through street intercepts or video questionnaires. These traditional assessment methods are limited by their strong subjectivity, susceptibility to environmental interference, and difficulty in ensuring accuracy. They cannot objectively and accurately quantify people's true perception of stress in the street environment, resulting in an inability to accurately reflect people's perceptions of the built environment when using existing assessment methods. This has become a problem that urgently needs to be solved. Summary of the Invention

[0004] The main objective of this invention is to provide an environmental perception assessment method, system, terminal, and computer-readable storage medium based on electrodermal signals, aiming to solve the problem that existing assessment methods cannot accurately reflect human perception of the built environment.

[0005] To achieve the above objectives, the present invention provides an environmental perception assessment method based on electrodermal signals, the method comprising the following steps: Acquire panoramic image data and geographic information data of the target area, extract environmental features from the panoramic image data and geographic information data to obtain target features, and input the target features into a structured database for storage to obtain an environmental feature database; Multi-dimensional environmental feature data from an environmental feature database is obtained. The target street image and the skin conductance signal of the multi-dimensional environmental feature data are then input into a perception prediction model for training to obtain the trained perception prediction model. Acquire the image data to be tested, input the image data to be tested into the trained perception prediction model for prediction, and obtain the pressure perception prediction result.

[0006] Optionally, in the environmental perception assessment method based on electrodermal signals, the target features include traffic environment features, street-level built environment features, and block-level built environment features. The process of acquiring panoramic image data and geographic information data of the target area, extracting environmental features from the panoramic image data and geographic information data to obtain target features, and inputting the target features into a structured database for storage to obtain an environmental feature database specifically includes: Acquire panoramic imagery and geographic information data of the target area; Instance detection is performed on the panoramic image data to obtain traffic environment features; The panoramic image data is processed by frame extraction according to the machine learning recognition strategy to obtain street view images. The street view images are then labeled to obtain the built environment features at the street level. The geographic information data is processed using geographic information technology to obtain the built environment characteristics at the street level; The traffic environment features, the street-level built environment features, and the block-level built environment features are input into a structured database for storage, resulting in an environmental feature database.

[0007] Optionally, the environmental perception assessment method based on electrodermal signals, wherein acquiring multi-dimensional environmental feature data from an environmental feature database, and inputting the target street image and the electrodermal signals from the multi-dimensional environmental feature data into a perception prediction model for training to obtain a trained perception prediction model, specifically includes: Obtain multi-dimensional environmental feature data from the environmental feature database; The multi-dimensional environmental feature data in the environmental feature database are transformed using cluster analysis to obtain typical feature types; The typical feature types are combined and filtered according to preset data similarity to obtain target street images; Based on a head-mounted virtual reality device and a skin conductance acquisition device, the test subject's signal is measured according to the target street image to obtain a pure skin conductance signal; The pure electrodermal signal was analyzed using the NeuroKit2 function to obtain the SCL gradient signal; By constructing the typical feature types and the SCL gradient signal, a corresponding relationship data set is obtained; The corresponding data set is input into the perception prediction model for training to obtain the trained perception prediction model.

[0008] Optionally, in the environmental perception assessment method based on electrodermal signals, the step of combining and filtering the typical feature types according to preset data similarity to obtain the target street image specifically includes: If the target feature is a built environment feature at the street level, then elbow analysis is performed on all the typical feature types based on preset data similarity to obtain typical feature types at the street level. If the target feature is a street-level built environment feature, then a multi-level clustering analysis is performed on all the typical feature types according to the preset data similarity to obtain the typical feature types at the street level. The typical feature types at the block level and the typical feature types at the street level are selected and matched from the environmental feature database in this way to obtain the target street image.

[0009] Optionally, the environmental perception assessment method based on electrodermal signals, wherein the step of measuring the signal of the test subject based on the target street image using a head-mounted virtual reality device and an electrodermal signal acquisition device to obtain a pure electrodermal signal specifically includes: Obtain baseline skin conductance data from the test subjects; Based on a head-mounted virtual reality device and a skin conductance acquisition device, the test subject is subjected to signal measurement according to the target street image and a preset number of times to obtain the raw skin conductance signal; Based on the baseline data of the skin conductance and the adaptive filtering algorithm, the original skin conductance signal is subjected to baseline drift processing and noise reduction processing to obtain a pure skin conductance signal.

[0010] Optionally, the environmental perception assessment method based on electrodermal signals, wherein the step of analyzing the pure electrodermal signals based on the NeuroKit2 function to obtain the SCL gradient signal, further includes: Obtain the pressure characteristics of the SCL gradient signal; The stress characteristics are normalized using a fractional normalization algorithm to obtain a standardized cycling stress index. The standardized cycling stress index is used to detect the user's psychological and physiological stress level.

[0011] Optionally, the environmental perception assessment method based on electrodermal signals, wherein the step of inputting the corresponding relationship data set into the perception prediction model for training to obtain the trained perception prediction model specifically includes: The corresponding data set is divided according to a preset ratio to obtain a training set and a validation set; The training set is input into the perception prediction model for training to obtain the trained perception prediction model. The performance of the trained perception prediction model is then verified using the validation set.

[0012] Furthermore, to achieve the above objectives, the present invention also provides an environmental perception assessment system based on electrodermal signals, wherein the environmental perception assessment system based on electrodermal signals: An environmental feature construction module is used to acquire panoramic image data and geographic information data of the target area, extract environmental features from the panoramic image data and geographic information data to obtain target features, and input the target features into a structured database for storage to obtain an environmental feature database. The environment model training module is used to acquire multi-dimensional environmental feature data from the environmental feature database, and input the target street image and the skin conductance signal of the multi-dimensional environmental feature data into the perception prediction model for training to obtain the trained perception prediction model. The environmental perception and prediction module is used to acquire multi-dimensional environmental feature data from the environmental feature database, and input the target street image and the skin conductance signal of the multi-dimensional environmental feature data into the perception and prediction model for training to obtain the trained perception and prediction model.

[0013] In addition, to achieve the above objectives, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores an environmental perception assessment program based on electrodermal signals, and when the environmental perception assessment program based on electrodermal signals is executed by a processor, it implements the steps of the environmental perception assessment method based on electrodermal signals as described above.

[0014] In this invention, panoramic image data and geographic information data of the target area are acquired. Environmental features are extracted from the panoramic image data and geographic information data to obtain target features. These target features are then stored in a structured database to obtain an environmental feature database. Multi-dimensional environmental feature data from the environmental feature database are acquired. Target street images and skin conductance signals from the multi-dimensional environmental feature data are input into a perception prediction model for training to obtain a trained perception prediction model. Test image data is acquired and input into the trained perception prediction model for prediction to obtain a stress perception prediction result. This invention, based on cluster analysis of the environmental feature database and training of the perception prediction model, performs perception prediction of the psychological stress impact of the street environment, achieving accurate and quantitative assessment of urban spatial psychological stress. Attached Figure Description

[0015] Figure 1 This is a flowchart of a preferred embodiment of the environmental perception assessment method based on skin electrical signals of the present invention; Figure 2 This is a schematic diagram of the experimental design of a preferred embodiment of the environmental perception assessment method based on electrodermal signals of the present invention; Figure 3 This is a schematic diagram of the street-level index clustering K value of a preferred embodiment of the environmental perception assessment method based on electrodermal signals of the present invention; Figure 4 This is a schematic diagram of the K-value of the left-side index cluster at the street level, which is a preferred embodiment of the environmental perception assessment method based on electrodermal signals of the present invention. Figure 5 This is a schematic diagram of the K-value of right-side index clustering at the street level, representing a preferred embodiment of the environmental perception assessment method based on electrodermal signals of the present invention. Figure 6 This is a schematic diagram of the skin electrical signal of a preferred embodiment of the environmental perception assessment method based on skin electrical signal of the present invention; Figure 7 This is a schematic diagram of the symbolic regression model of a preferred embodiment of the environmental perception assessment method based on electrodermal signals of the present invention; Figure 8 This is a schematic diagram of a multi-source feature extractor, representing a preferred embodiment of the environmental perception assessment method based on electrodermal signals of the present invention. Figure 9 This is a schematic diagram of a preferred embodiment of the environmental perception assessment method based on electrodermal signals of the present invention for assessing cycling stress. Figure 10 This is a structural diagram of a preferred embodiment of the environmental perception and assessment system based on skin electrical signals of the present invention; Figure 11 This is a structural diagram of a preferred embodiment of the terminal of the device of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0017] Existing methods for assessing perception of the built environment mainly include: relying on expert experience to determine assessment indicators and results, and conducting surveys of citizens through street intercepts or video questionnaires. These traditional assessment methods are limited by their strong subjectivity, susceptibility to environmental interference, and difficulty in ensuring accuracy. They cannot objectively and accurately quantify people's true perception of stress from the built environment. Therefore, a method based on electrodermal signaling (EDS) is needed. This method would use cluster analysis of a database of built environment characteristics and train a perception prediction model to predict the psychological stress impact of the built environment, achieving a precise and quantitative assessment of psychological stress from the built environment and avoiding the problem of inaccurately reflecting people's perception of the built environment.

[0018] The preferred embodiment of the environmental perception assessment method based on electrodermal signals of the present invention, such as... Figure 1 As shown, the environmental perception assessment method based on electrodermal signals includes the following steps: Step S10: Obtain panoramic image data and geographic information data of the target area, extract environmental features from the panoramic image data and geographic information data to obtain target features, and input the target features into a structured database for storage to obtain an environmental feature database.

[0019] Step S10 includes: Step S11: Obtain panoramic image data and geographic information data of the target area; Step S12: Perform instance detection on the panoramic image data to obtain traffic environment features; Step S13: Perform frame extraction processing on the panoramic image data according to the machine learning recognition strategy to obtain street view images, and annotate the street view images to obtain the built environment features at the street level. Step S14: Perform geographic information technology processing on the geographic information data to obtain the built environment characteristics at the street level; Step S15: Input the traffic environment features, the street-level built environment features, and the block-level built environment features into a structured database for storage to obtain an environment feature database.

[0020] Specifically, panoramic image data and geographic information data (panoramic video + GPS trajectory) of the target area are acquired. The panoramic image data is processed by frame extraction according to a machine learning recognition strategy to obtain street view images. Instance detection is performed on the panoramic image data to obtain traffic environment features (instance detection is performed on the panoramic video to statistically analyze indicators such as motor vehicle traffic flow, non-motorized vehicle lane traffic flow, non-motorized vehicle wrong-way rate, and pedestrian lane occupancy rate). The street view images are labeled to obtain street-level built environment features. The geographic information data is processed using geographic information technology to obtain block-level built environment features (based on GPS coordinates, regional features such as land use mixing, residential population density, non-motorized vehicle lane density, and distance to subway stations / schools / parks are processed using open-source GIS (Geographic Information System) geographic information technology, and finally integrated into a regional database for subsequent inference and feature extraction to obtain block-level built environment features). The traffic environment features, street-level built environment features, and block-level built environment features are matched with GPS data and then input into a structured database for storage to obtain an environmental feature database (a cycling environment feature database, with specific indicators and value ranges for each module, stored using a structured database).

[0021] In this embodiment, Area A, characterized by its urgent need for cycling environments and rich features, was selected, and a "panoramic imagery + GPS trajectory" data acquisition scheme was adopted. Staff rode bicycles equipped with a panoramic camera and a GPS system, selecting sections of the area containing non-motorized vehicle lanes, and acquiring panoramic images containing the entire view of these sections. During the acquisition process, the geographic coordinates of each panoramic image were recorded via GPS, laying the foundation for subsequent association with street-level data.

[0022] Step S20: Obtain multi-dimensional environmental feature data from the environmental feature database, input the target street image and the skin conductance signal of the multi-dimensional environmental feature data into the perception prediction model for training, and obtain the trained perception prediction model.

[0023] Step S20 includes: Step S21: Obtain multi-dimensional environmental feature data from the environmental feature database; Step S22: Transform the multi-dimensional environmental feature data in the environmental feature database using cluster analysis to obtain typical feature types; Step S23: Combine and filter the typical feature types according to preset data similarity to obtain target street images; Step S24: Based on the head-mounted virtual reality device and the skin conductance acquisition device, the test subject is subjected to signal measurement according to the target street image to obtain a pure skin conductance signal; Step S25: Analyze the pure electrodermal signal based on the NeuroKit2 function to obtain the SCL gradient signal; Step S26: Construct the typical feature types and the SCL gradient signal to obtain a corresponding relationship data set; Step S27: Input the corresponding relationship data set into the perception prediction model for training to obtain the trained perception prediction model.

[0024] Specifically, after step S25, the method further includes acquiring the pressure characteristics of the SCL (Skin Conductance Level) gradient signal, and normalizing the pressure characteristics (which represent the derivative of the SCL gradient signal corresponding to the cycling image) using a fractional normalization algorithm to obtain a standardized cycling pressure index (the standardized cycling pressure index is obtained by normalizing the original trial pressure level characteristics using the Z-fractional normalization method). The standardized cycling stress index is used to detect the user's psychophysiological stress level (the higher the index value, the higher the psychophysiological stress level experienced by the subject under the video stimulus of the typical road segment. This index constitutes a high-quality, objective physiological response-based supervision label for the subsequent training of the "street environment features → cycling stress" prediction model). Multi-dimensional environmental feature data from the environmental feature database is obtained, and the multi-dimensional environmental feature data in the environmental feature database is transformed using cluster analysis to obtain typical feature types. These typical feature types are then combined and filtered according to preset data similarity to obtain the target street image (for the ring...). Cluster analysis is performed on multi-dimensional environmental feature data from the environmental feature database to obtain typical feature types for streets and blocks. The clustering results are combined and divided according to preset similarity, and then matched from the original street image database to obtain target street images (i.e., the typical feature types are combined and divided according to preset data similarity, and target street images are obtained sequentially). Based on a head-mounted virtual reality device and a skin conductance acquisition device, signal measurements are performed on the test subject based on the target street images to obtain pure skin conductance signals. The pure skin conductance signals are analyzed based on the NeuroKit2 function to obtain SCL (Skin Conductance Level) gradient signals (SCL gradient signals are often used to measure relatively smooth changes in emotion or stress). The typical feature types and the standardized cycling stress index (processed SCL gradient signals) are constructed to obtain a corresponding relationship data set. The corresponding relationship data set is input into the perception prediction model for training to obtain the trained perception prediction model.

[0025] In this embodiment, SCL (Self-Care Scale) can effectively capture cyclists' emotional fluctuations in different environments, making it suitable for assessing cyclists' perceptual experiences during long-distance cycling. Secondly, compared to subjective questionnaires, SCL data has higher objectivity and reliability. It measures physiological signals rather than relying on the cyclist's subjective descriptions, avoiding data distortion caused by individual misunderstandings or memory errors. Therefore, SCL can accurately reflect cyclists' emotional fluctuations and stress levels in different cycling environments in real time, making it ideal physiological perceptual data for measuring cycling perception.

[0026] For example, such as Figure 2 As shown, firstly, based on 76 typical road sections, peak-hour and off-peak-hour images were extracted to generate a total of 152 first-person cycling perspective videos. Then, all videos were uniformly edited into fixed-length 20-second segments. To facilitate the experiment and control for order effects, these 152 video segments were randomly and evenly divided into 8 groups (labeled A, B, C, ..., H), with each group containing 19 videos. Throughout the experimental period, a double-blind randomization method was used to select and assign stimulus sequences to each participant from these 8 groups. Specifically, neither the experimenter nor the participants knew the specific content grouping of the videos used in each test, thereby minimizing expectation effects and selection bias. Figure 2 As shown, the single-unit experimental procedure is as follows: After entering the laboratory and putting on the head-mounted virtual reality device and skin conductance sensor, the device is first debugged and the signal is calibrated. Then, data collection officially begins: Baseline acquisition: The subject sits and relaxes for 3 minutes, during which their resting skin conductance signal is recorded as the baseline of the individual's physiological response. Stimulus-recovery cycle: The system begins playing a set of 19 randomly assigned videos. Each stimulus-recovery cycle includes: a 20-second continuous cycling scene stimulus presentation, followed by a 60-second resting recovery period (the screen displays a neutral image). This "20-second stimulus + 60-second recovery" cycle is repeated 19 times until all videos in the set have been played. Rest between sets: After completing a set of 19 cycles, the subject leaves the device and rests for at least 5 minutes to relieve visual and mental fatigue. Progress and interval control across experimental units: To balance the experimental load and ensure data quality, strict progress control rules were designed: Each time the subject visits the laboratory, they must complete two different sets of video stimuli consecutively (i.e., a total of 38 stimulus-recovery cycles). After completing two sets of tests, the system (or experimenter) must determine whether the subject has completed all eight sets of videos. If not, the next test for the same subject must be scheduled at least 48 hours later to strictly control the potential confounding effects of fatigue on physiological data. This process is repeated until the subject completes all eight sets of tests. Once a subject has completed all eight sets of tests, their individual experimental portion is complete.

[0027] In this embodiment, the K-means clustering method from the Scikit-learn machine learning library is used to classify and study roads with non-motorized vehicle lanes in the main and secondary roads of area A in a Python environment. Specifically, built environment indicators at the block level and street level are selected as clustering factors, and cluster analysis is performed on non-motorized vehicle lanes that are shared by pedestrians and non-motorized vehicles and those that are shared by motorized vehicles, respectively, to reveal the potential impact of different built environment characteristics on cycling perception. The K-means algorithm updates the cluster centers iteratively by minimizing the sum of squared errors.

[0028] In this embodiment, the built environment characteristics at the street level are determined by manually auditing the built environment characteristic information contained in the street view images, judging indicators such as the width (narrow or wide) of the non-motorized vehicle lane, the form of separation between motorized and non-motorized vehicles, the type of shared road, whether it includes a bus stop, whether it includes obstructions, whether the greenery on both sides is discontinuous, whether the greenery on both sides is continuous, whether the greenery on both sides is a motorized vehicle lane, whether the greenery on both sides is a sidewalk, whether the greenery on both sides is a non-motorized vehicle parking space, and whether the greenery on both sides is a construction road (with the non-motorized vehicle lane as the central axis, and the first section type on its left side, hereinafter named Left 1, and similarly, Left 1, Left 2, Right 1, and Right 2 form "both sides"). The manual review ensures an accuracy rate of over 90%. At the same time, the green view rate and sky visibility are calculated through semantic segmentation algorithms and reclassified into three levels: high, medium, and low.

[0029] As an example, the built environment characteristics at the street level are analyzed using open-source GIS geographic information technology, based on GPS coordinates. These characteristics include land use mix, population density, intersection density, non-motorized vehicle lane density, and distance to subway stations / schools / parks. The data is then integrated into a regional database for subsequent feature extraction and inference, employing a "panoramic video + GPS trajectory" data collection scheme. Staff members rode bicycles equipped with a panoramic camera and Garmin GPS system, controlling the weather, and collected 76 30-second panoramic videos of classic cycling routes during peak and off-peak hours. During the collection process, the geographic coordinates of each panoramic image were recorded via GPS, laying the foundation for subsequent correlation with street-level data. These 152 video clips were then edited into 152 20-second panoramic videos to serve as stimuli for cycling perception experiments.

[0030] For example, such as Figure 3As shown, the elbow method was used to analyze the intersection density, floor area ratio, population density, density of five types of POIs (Points of Interest), density of bus stops and subway stations, and functional mixing at the street level. The figure shows that the rate of decrease in SEE slows down when K=2. Therefore, the built environment at the street level is divided into two categories. Based on the spatial visualization of the K-means clustering results, the built environment of non-motorized lanes in area A shows a "core-periphery" differentiation pattern. The central urban area is clustered as a street with convenient access, characterized by high density, high functional mixing, and complete slow-traffic facilities. The surrounding areas are classified as streets with potential, which are limited by low development intensity, single function, and insufficient facilities. The central area has formed a mature slow-traffic network based on historical accumulation, while the peripheral area faces a dynamic imbalance between facility supply and population growth due to planning lag.

[0031] Furthermore, such as Figure 4 and Figure 5 As shown, the optimal number of classifications was determined using the elbow method. It was found that non-motorized vehicle lanes and sidewalks sharing a single road surface slab (shared slab for pedestrians and non-motorized vehicles) can be divided into three categories based on the isolation strength: strong, medium, and weak. Motorized vehicle lanes and non-motorized vehicle lanes on the same road surface slab at the same elevation (shared slab for motorized and non-motorized vehicles) are divided into two categories: strong and medium. Furthermore, the strong and medium isolation categories exhibit differentiated functional combinations. In the weak isolation type, the right side of the cross-section connects the sidewalk and continuous green belt sequentially, while the left side is a discontinuous green belt and motorized vehicle lane, forming a limited buffer space. In the medium isolation type, the right side maintains the combination of sidewalk and continuous green belt, while the left side strengthens the continuous green belt isolation, improving the separation efficiency of the motorized vehicle lane. In the strong isolation type, the right side is adjusted to a staggered layout of continuous green belt and sidewalk, while the left side continues the connection mode of continuous green belt and motorized vehicle lane, constructing multiple physical buffer layers.

[0032] In this embodiment, in the shared-slab cross-section of motorized and non-motorized lanes, the strong isolation type has two subcategories: the right side is connected to the sidewalk with continuous or discontinuous greenery, while the left side is a combination of continuous greenery and motorized lanes. The medium isolation type is further differentiated, with continuous or discontinuous greenery on the right side combined with the sidewalk, and the left side having a bidirectional extension of the motorized lanes, weakening the protection of non-motorized vehicle right-of-way. The clustering results of the two types of cross-sections show that the difference in isolation intensity is not only reflected in the continuity of greenery, but also in the spatial sequence recombination of sidewalks, green belts, and motorized lanes, ultimately resulting in three types of non-motorized lane cross-sections: strong isolation, medium isolation, and weak isolation, and two types of built environment at the street level: potential type and convenient type.

[0033] In this embodiment, the built environment characteristics of the pedestrian and non-motorized vehicle shared-slab streets in Area A can be summarized into two categories: "potential street type" and "convenient street type." The differences lie in the road grade, isolation form, and spatial function combination. Potential street types are mostly distributed on branch roads and some main and secondary roads, dominated by non-physical isolation, with asphalt as the main paving material and weak to medium isolation strength. The road surface function on both sides is often a weak separation combination of green belt and motor vehicle lane, and the right-of-way division is relatively vague. Convenient street types are concentrated on main and secondary roads, with a significantly increased proportion of physical isolation. The paving material covers asphalt and stone bricks, and the isolation strength is mainly medium to strong. The priority of slow traffic is enhanced by continuous greening and nested sidewalks on both sides. Both types of streets show the same performance in terms of flatness and control of visual obstructions, and there are no significant bumps or obstacles.

[0034] Furthermore, in Area A, typical streets with shared lanes for motor vehicles and non-motor vehicles are mainly main and secondary roads, generally using physical separation and asphalt paving materials. The separation strength is concentrated at the strong and medium levels. The functions of the road surface on both sides are clearly separated from the motor vehicle lanes by continuous green belts. The cycling space is flat and there is no visual obstruction or interference. The street classification is dominated by "convenience type". Its core characteristics are high-density physical separation facilities and green buffer layers, which enhance the collaborative safety of non-motor vehicle lanes and motor vehicle lanes. Although a few potential streets maintain physical separation and flat paving, they still face shortcomings in spatial vitality and connection efficiency due to low road grade or insufficient functional mixing.

[0035] Step S23 includes: Step S231: If the target feature is a built environment feature at the street level, then perform elbow analysis on all the typical feature types according to the preset data similarity to obtain typical feature types at the street level. Step S232: If the target feature is a street-level built environment feature, then perform multi-level cluster analysis on all the typical feature types according to the preset data similarity to obtain the typical feature types at the street level. Step S233: Select and match the typical feature types at the block level and the typical feature types at the street level from the environmental feature database to obtain the target street image.

[0036] Specifically, if the target feature is a built environment feature at the street level, then elbow analysis (combination division) is performed on all the typical feature types according to preset data similarity (multi-dimensional cross-combination is performed according to preset data similarity judgment rules to form several groups of multi-dimensional feature combination types covering both street and street scales) to obtain typical feature types at the street level (street level indicator clustering distribution map). If the target feature is a built environment feature at the street level, then multi-level clustering analysis is performed on all the typical feature types according to preset data similarity to obtain typical feature types at the street level (street level indicator clustering distribution map). The typical feature types at the street level and the typical feature types at the street level are then selected and matched from the environmental feature database to obtain the target street image. Further, using the formed multi-dimensional feature combination type as the matching standard, the original street image database is traversed, and the environmental features of each street image in the database are compared with each multi-dimensional feature combination type to select street images whose environmental features match the corresponding combination type and meet the preset matching conditions, which are then used as the target street images required for subsequent processing.

[0037] Step S24 includes: Step S241: Obtain baseline skin conductance data of the test subject; Step S242: Based on the head-mounted virtual reality device and the skin conductance acquisition device, the test subject is subjected to signal measurement according to the target street image and a preset number of times to obtain the raw skin conductance signal; Step S243: Based on the baseline data of the skin conductance and the adaptive filtering algorithm, the original skin conductance signal is subjected to baseline drift processing and noise reduction processing to obtain a pure skin conductance signal.

[0038] Specifically, baseline skin conductance data of the test subjects was obtained (the subjects sat and relaxed for 3 minutes, during which their skin conductance signals in a resting state were recorded as the baseline of individual physiological response). Based on a head-mounted virtual reality device and a skin conductance acquisition device, the test subjects were subjected to signal measurement according to the target street image and a preset number of times to obtain raw skin conductance signals. Based on the skin conductance baseline data and an adaptive filtering algorithm, the raw skin conductance signals were processed for baseline drift and noise reduction to obtain a clean skin conductance signal (first, the raw skin conductance signal was preprocessed, and then adaptive filtering was performed by the eda_clean function of the NeuroKit2 Python toolkit for psychophysiological signal processing to remove baseline drift and noise, and finally a clean skin conductance signal was obtained).

[0039] Furthermore, such as Figure 6 As shown, the eda_phasic function of the Python toolkit NeuroKit2 is used to extract pure electrodermal signals (such as...). Figure 6The original signal in (a) is decomposed into two physiologically significant components: such as Figure 6 (b) Basic activities and such Figure 6 (c) Phase activity. Phase activity reflects short-term fluctuations closely related to discrete external or internal stimuli, such as sudden loud noises, visual stimuli, or cognitive tasks that elicit arousal responses. These rapid changes, also known as the Skin Conductance Response (SCR), are responses to specific events, characterized by a sharp, transient increase in skin conductance followed by a gradual return to baseline; baseline activity, on the other hand, reflects the overall level of arousal over a longer period. It is often referred to as skin conductance level (SCL), reflecting an individual's overall physiological state or emotional baseline. The tonic component changes gradually, influenced by factors such as stress, attention, or persistent emotional state, rather than direct stimuli. Baseline activity provides crucial context for interpreting an individual's broader arousal state. This decomposition effectively separates background arousal from transient stress responses, providing an ideal data foundation for subsequent feature extraction.

[0040] In this embodiment, based on the decomposed tonic activity, baseline correction is performed for each 20-second stimulus trial window using the median tonic activity within a 5-second time window prior to the stimulus event. This correction eliminates the bias in tonic activity before the stimulus, allowing the Skin Conductance Level (SCL) feature to more accurately reflect the immediate stress response of that trial. For each corrected tonic activity within a stimulus trial window, the derivative of the tonic activity within each trial is calculated. This represents the stress level of the subject for each stimulus.

[0041] Step S27 includes: Step S271: Divide the corresponding relationship data set according to a preset ratio to obtain a training set and a validation set; Step S272: Input the training set into the perception prediction model for training to obtain the trained perception prediction model, and verify the performance of the trained perception prediction model based on the validation set.

[0042] Specifically, the corresponding data set is divided according to a preset ratio to obtain a training set and a validation set (the dataset is divided into a training set (for model training) and a validation set (for performance verification) at an 8:2 ratio). The training set is input into the perception prediction model for training to obtain a trained perception prediction model (using key features of traffic environment, street-level built environment, and block-level built environment as input variables, i.e., the training set, and cumulative skin conductance change E (or cycling stress level: low / medium / high) as the output variable, training a symbolic regression model, and optimizing the model by adjusting parameters such as learning rate and tree depth, with a target prediction accuracy ≥85%). The performance of the trained perception prediction model is tested based on the validation set (the model performance is tested using prediction set data; if the accuracy does not meet the standard, experimental data is supplemented or feature variables (such as street green coverage) are added, and iterative optimization is performed until the requirements are met). Furthermore, a street environment feature recognition model based on YOLOv8 (the trained perception prediction model) ensures that the required input features of the model can be quickly extracted from newly acquired street scene images.

[0043] In this embodiment, key features of the traffic environment, the built environment of the neighborhood, and the built environment of the street are used as input variables, and a standardized cycling stress index is used. For the output variable, a symbolic regression model is trained. The model is optimized by adjusting parameters such as learning rate and tree depth, aiming for a prediction accuracy of ≥85%. A syntax set G={T, F} is defined, where the set of terminators (T) contains all standardized environmental feature variables {x1, x2, ..., xd} and constant terms {c1, c2, ..., ck}. Represents the first standardized environmental characteristic variable. This represents the second standardized environmental characteristic variable. Denotes the d-th standardized environmental characteristic variable. Represents the first constant term. This represents the second constant term. Let represent the k-th constant term, with the constant randomly generated within the range [-10, 10]. The function set (F) includes basic mathematical operators such as addition, subtraction, multiplication, protective division, exponential function, protective logarithmic function, and square root. Model training is performed using the SymbolicRegressor function from the Python library gplearn. To optimize algorithm performance, key hyperparameters (including population size, crossover probability, mutation probability, maximum tree depth, and complexity penalty coefficient λ) are automatically tuned using a Bayesian optimization framework. This framework aims to minimize the prediction error on the validation set, efficiently searching within a predefined hyperparameter space to form a closed loop of "parameter configuration → model training → performance evaluation → feedback optimization," thereby ensuring a stable and highly generalizable model. Finally, the algorithm selects the mathematical expression with the lowest mean squared error and relatively simple structure on the validation set from all evolutionary generations as the prediction model. Model Validation: Finally, we conducted a comprehensive model diagnostic on the trained model: (1) Residual Analysis: Check whether the prediction residuals satisfy the independent and identically distributed assumption, and whether there are obvious patterns or heteroscedasticity; (2) Feature Importance Analysis: Quantify the contribution of each environmental feature to the model prediction through permutation-based feature importance assessment; (3) Partial Dependency Analysis: Visualize the marginal relationship between key environmental features and prediction pressure, revealing potential nonlinear effects; (4) Stability Test: Check the stability of the model expression through repeated training to ensure that the discovered relationship is a reliable pattern in the data rather than random noise. Auxiliary Model Support: Simultaneously improve the street environment feature recognition model based on YOLOv8 to ensure that the required input features of the model can be quickly extracted from newly acquired street scene images.

[0044] For example, such as Figure 7 As shown, symbolic regression was used as the modeling method. Symbolic regression is a programmed machine learning method that automatically searches for the mathematical expression that best describes the data relationship in a solution space composed of basic mathematical elements by simulating the natural evolutionary process. The algorithm starts with a randomly generated initial population (a set of formulas) and optimizes these formulas through an iterative evolutionary process. In each generation, the fitness of each individual formula is evaluated by the MSE error between its predicted value and the actual physiological stress index. Individuals with high fitness are retained and used as parents to generate offspring. This process is repeated, so that the overall fitting ability of the population increases generation by generation, eventually converging to one or more explicit mathematical formulas that achieve a balance between prediction accuracy and complexity. Its final output is not an uninterpretable "black box," but a clearly structured, such as... Figure 7 The model framework shown.

[0045] Step S30: Obtain the image data to be tested, input the image data to be tested into the trained perception prediction model for prediction, and obtain the pressure perception prediction result.

[0046] Specifically, the image data to be tested is acquired, and the image data to be tested is input into the trained perception prediction model for prediction to obtain the stress perception prediction result (cycling stress level).

[0047] In this embodiment, for streets in the target area that have not been measured, environmental features are extracted from street view images and input into a trained prediction model to quickly obtain the cycling pressure level (low / medium / high, i.e., pressure perception prediction results) of each road segment. This achieves an automated process from "image input to feature extraction, and then to perception prediction," efficiently completing the perception measurement of large-area streets. Based on the pressure perception prediction results, combined with GIS network analysis, priority update road segments with "high pressure and high centrality" (i.e., road segments with high-frequency traffic and poor cycling experience) are identified. For the key problem characteristics of different road segments, renovation strategies are proposed, such as adding physical barriers to road segments with "high interference from motor vehicles" and widening non-motorized vehicle lanes to road segments with "narrow lanes." Workshops are organized with planners, experts, and citizens to transform the prediction results into implementable update plans, guiding the refined renovation of the urban cycling environment.

[0048] For example, such as Figure 8 As shown, firstly, a pre-trained semantic segmentation model (such as DeepLabv3+) is used to perform pixel-level analysis of the input image, identifying static elements such as motor vehicle lanes, non-motor vehicle lanes, sidewalks, greenery, buildings, and the sky. Then, street-scale built environment features are extracted, including the proportion of non-motor vehicle lanes, the ratio of motor vehicle to non-motor vehicle area, green space ratio, sky visibility, and spatial continuity indicators. Secondly, a target detection model (such as YOLOv8) is used to identify dynamic targets such as motor vehicles, non-motor vehicles, and pedestrians in the image, statistically analyzing their quantity, spatial density, and overlap with non-motor vehicle lanes to quantify dynamic traffic characteristics. Simultaneously, an instance detection model is used to process the input panoramic image, obtaining traffic environment feature indicators including motor vehicle traffic flow, non-motor vehicle traffic flow, non-motor vehicle wrong-way rate, and pedestrian lane encroachment rate. Finally, based on GPS coordinate matching with geographic information data, street-scale built environment features are extracted, such as road grade, intersection density, building density, functional mixing, and public transportation facility density, achieving the fusion of visual data and macro-geographic information. Finally, the extracted static street features, dynamic traffic features, and block features are standardized and matched with GPS data to form a unified multi-source cycling environment feature vector. This ensures consistency with the data structure during the model training phase and is then input into the prediction model for evaluation.

[0049] Furthermore, such as Figure 9As shown, the environmental feature vector obtained in the previous step is input into the trained cycling stress prediction model (a mathematical expression based on symbolic regression), which automatically calculates the predicted value of continuous cycling stress. The inference process requires no manual intervention. The results can be directly output as a stress index or discretized into "low, medium, and high" levels according to thresholds. The system binds and stores stress values ​​with images and geographic locations, supports batch inference, and can automatically filter high-stress road sections, providing a priority list for planning interventions. This mechanism provides planning decision-makers with an efficient analytical tool, enabling full-network cycling stress diagnosis and stress heatmap generation. It allows for simulation prediction and quantitative comparison of planning schemes (such as adding barriers or increasing greenery), accurately identifying high-risk road sections requiring priority improvement, thereby achieving low-cost, large-scale application of small-sample physiological calibration results.

[0050] Furthermore, such as Figure 10 As shown, based on the above-described environmental perception assessment method based on electrodermal signals, the present invention also provides an environmental perception assessment system based on electrodermal signals, wherein the environmental perception assessment system based on electrodermal signals includes: The environmental feature construction module 51 is used to acquire panoramic image data and geographic information data of the target area, extract environmental features from the panoramic image data and geographic information data to obtain target features, and input the target features into a structured database for storage to obtain an environmental feature database. The environment model training module 52 is used to acquire multi-dimensional environmental feature data from the environmental feature database, and input the target street image and the skin conductance signal of the multi-dimensional environmental feature data into the perception prediction model for training to obtain the trained perception prediction model. The environmental perception prediction module 53 is used to acquire the image data to be tested, input the image data to be tested into the trained perception prediction model for prediction, and obtain the pressure perception prediction result.

[0051] Furthermore, such as Figure 11 As shown, based on the above-mentioned environmental perception assessment method and system based on skin electrical signals, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 11 Only some of the terminal components are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0052] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory. In other embodiments, the memory 20 may be an external storage device of the terminal, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc. Further, the memory 20 may include both internal and external storage devices. The memory 20 is used to store application software and various types of data installed on the terminal, such as program code installed on the terminal. The memory 20 can also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 20 stores an environmental perception assessment program 40 based on electrodermal signals, which can be executed by the processor 10 to implement the environmental perception assessment method based on electrodermal signals in this application.

[0053] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in the memory 20 or process data, such as executing the environmental perception assessment method based on electrodermal signals.

[0054] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. The terminals communicate with each other via a system bus.

[0055] In one embodiment, when the processor 10 executes the environment perception assessment program 40 based on electrodermal signals in the memory 20, the following steps are performed: Acquire panoramic image data and geographic information data of the target area, extract environmental features from the panoramic image data and geographic information data to obtain target features, and input the target features into a structured database for storage to obtain an environmental feature database; Multi-dimensional environmental feature data from an environmental feature database is obtained. The target street image and the skin conductance signal of the multi-dimensional environmental feature data are then input into a perception prediction model for training to obtain the trained perception prediction model. Acquire the image data to be tested, input the image data to be tested into the trained perception prediction model for prediction, and obtain the pressure perception prediction result.

[0056] The target features include traffic environment features, street-level built environment features, and block-level built environment features; The process of acquiring panoramic image data and geographic information data of the target area, extracting environmental features from the panoramic image data and geographic information data to obtain target features, and inputting the target features into a structured database for storage to obtain an environmental feature database specifically includes: Acquire panoramic imagery and geographic information data of the target area; Instance detection is performed on the panoramic image data to obtain traffic environment features; The panoramic image data is processed by frame extraction according to the machine learning recognition strategy to obtain street view images. The street view images are then labeled to obtain the built environment features at the street level. The geographic information data is processed using geographic information technology to obtain the built environment characteristics at the street level; The traffic environment features, the street-level built environment features, and the block-level built environment features are input into a structured database for storage, resulting in an environmental feature database.

[0057] Specifically, the step of acquiring multi-dimensional environmental feature data from an environmental feature database involves inputting the target street image and the electrodermal signal from the multi-dimensional environmental feature data into a perception prediction model for training, resulting in a trained perception prediction model. Obtain multi-dimensional environmental feature data from the environmental feature database; The multi-dimensional environmental feature data in the environmental feature database are transformed using cluster analysis to obtain typical feature types; The typical feature types are combined and filtered according to preset data similarity to obtain target street images; Based on a head-mounted virtual reality device and a skin conductance acquisition device, the test subject's signal is measured according to the target street image to obtain a pure skin conductance signal; The pure electrodermal signal was analyzed using the NeuroKit2 function to obtain the SCL gradient signal; By constructing the typical feature types and the SCL gradient signal, a corresponding relationship data set is obtained; The corresponding data set is input into the perception prediction model for training to obtain the trained perception prediction model.

[0058] Specifically, the step of combining and filtering the typical feature types based on preset data similarity to obtain the target street image includes: If the target feature is a built environment feature at the street level, then elbow analysis is performed on all the typical feature types based on preset data similarity to obtain typical feature types at the street level. If the target feature is a street-level built environment feature, then a multi-level clustering analysis is performed on all the typical feature types according to the preset data similarity to obtain the typical feature types at the street level. The typical feature types at the block level and the typical feature types at the street level are selected and matched from the environmental feature database in this way to obtain the target street image.

[0059] Specifically, the method of using a head-mounted virtual reality device and a skin conductance acquisition device to measure signals from the test subject based on the target street image to obtain a pure skin conductance signal includes: Obtain baseline skin conductance data from the test subjects; Based on a head-mounted virtual reality device and a skin conductance acquisition device, the test subject is subjected to signal measurement according to the target street image and a preset number of times to obtain the raw skin conductance signal; Based on the baseline data of the skin conductance and the adaptive filtering algorithm, the original skin conductance signal is subjected to baseline drift processing and noise reduction processing to obtain a pure skin conductance signal.

[0060] The method of analyzing the pure electrodermal signal based on the NeuroKit2 function to obtain the SCL gradient signal further includes: Obtain the pressure characteristics of the SCL gradient signal; The stress characteristics are normalized using a fractional normalization algorithm to obtain a standardized cycling stress index. The standardized cycling stress index is used to detect the user's psychological and physiological stress level.

[0061] Specifically, the step of inputting the corresponding relationship data set into the perception prediction model for training to obtain the trained perception prediction model includes: The corresponding data set is divided according to a preset ratio to obtain a training set and a validation set; The training set is input into the perception prediction model for training to obtain the trained perception prediction model. The performance of the trained perception prediction model is then verified using the validation set.

[0062] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores an environmental perception assessment program based on electrodermal signals, and the environmental perception assessment program based on electrodermal signals, when executed by a processor, implements the steps of the environmental perception assessment method based on electrodermal signals as described above.

[0063] In summary, this invention provides an environmental perception assessment method, system, terminal, and storage medium based on electrodermal signals. The method includes: acquiring panoramic image data and geographic information data of a target area; extracting environmental features from the panoramic image data and geographic information data to obtain target features; inputting the target features into a structured database for storage to obtain an environmental feature database; acquiring multi-dimensional environmental feature data from the environmental feature database; inputting target street images and electrodermal signals from the multi-dimensional environmental feature data into a perception prediction model for training to obtain a trained perception prediction model; acquiring image data to be tested; inputting the image data to be tested into the trained perception prediction model for prediction to obtain a stress perception prediction result; and performing cluster analysis on the environmental feature database and training the perception prediction model to perceive and predict the psychological stress impact of the street environment, thereby achieving a precise and quantitative assessment of urban spatial psychological stress.

[0064] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal system that includes that element.

[0065] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The computer-readable storage medium can be a memory, magnetic disk, optical disk, etc.

[0066] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. An environmental perception assessment method based on electrodermal signals, characterized in that, The environmental perception assessment method based on electrodermal signals includes: Acquire panoramic image data and geographic information data of the target area, extract environmental features from the panoramic image data and geographic information data to obtain target features, and input the target features into a structured database for storage to obtain an environmental feature database; Multi-dimensional environmental feature data from an environmental feature database is obtained. The target street image and the skin conductance signal of the multi-dimensional environmental feature data are then input into a perception prediction model for training to obtain the trained perception prediction model. Acquire the image data to be tested, input the image data to be tested into the trained perception prediction model for prediction, and obtain the pressure perception prediction result.

2. The environmental perception assessment method based on electrodermal signals according to claim 1, characterized in that, The target features include traffic environment features, street-level built environment features, and block-level built environment features; The process of acquiring panoramic image data and geographic information data of the target area, extracting environmental features from the panoramic image data and geographic information data to obtain target features, and inputting the target features into a structured database for storage to obtain an environmental feature database specifically includes: Acquire panoramic imagery and geographic information data of the target area; Instance detection is performed on the panoramic image data to obtain traffic environment features; The panoramic image data is processed by frame extraction according to the machine learning recognition strategy to obtain street view images. The street view images are then labeled to obtain the built environment features at the street level. The geographic information data is processed using geographic information technology to obtain the built environment characteristics at the street level; The traffic environment features, the street-level built environment features, and the block-level built environment features are input into a structured database for storage, resulting in an environmental feature database.

3. The environmental perception assessment method based on electrodermal signals according to claim 1, characterized in that, The process of acquiring multi-dimensional environmental feature data from an environmental feature database involves inputting the target street image and the electrodermal signal from the multi-dimensional environmental feature data into a perception prediction model for training, resulting in a trained perception prediction model. Specifically, this includes: Obtain multi-dimensional environmental feature data from the environmental feature database; The multi-dimensional environmental feature data in the environmental feature database are transformed using cluster analysis to obtain typical feature types; The typical feature types are combined and filtered according to preset data similarity to obtain target street images; Based on a head-mounted virtual reality device and a skin conductance acquisition device, the test subject's signal is measured according to the target street image to obtain a pure skin conductance signal; The pure electrodermal signal was analyzed using the NeuroKit2 function to obtain the SCL gradient signal; By constructing the typical feature types and the SCL gradient signal, a corresponding relationship data set is obtained; The corresponding data set is input into the perception prediction model for training to obtain the trained perception prediction model.

4. The environmental perception assessment method based on electrodermal signals according to claim 3, characterized in that... ； The step of combining and filtering the typical feature types based on preset data similarity to obtain the target street image specifically includes: If the target feature is a built environment feature at the street level, then elbow analysis is performed on all the typical feature types based on preset data similarity to obtain typical feature types at the street level. If the target feature is a street-level built environment feature, then a multi-level cluster analysis is performed on all the typical feature types according to the preset data similarity to obtain the typical feature types at the street level. The typical feature types at the block level and the typical feature types at the street level are selected and matched from the environmental feature database in this way to obtain the target street image.

5. The environmental perception assessment method based on electrodermal signals according to claim 3, characterized in that, The method, based on a head-mounted virtual reality device and a skin conductance acquisition device, measures the signal of the test subject according to the target street image to obtain a pure skin conductance signal, specifically including: Obtain baseline skin conductance data from the test subjects; Based on a head-mounted virtual reality device and a skin conductance acquisition device, the test subject is subjected to signal measurement according to the target street image and a preset number of times to obtain the raw skin conductance signal; Based on the baseline data of the skin conductance and the adaptive filtering algorithm, the original skin conductance signal is subjected to baseline drift processing and noise reduction processing to obtain a pure skin conductance signal.

6. The environmental perception assessment method based on electrodermal signals according to claim 5, characterized in that, The method involves analyzing the pure electrodermal signal using the NeuroKit2 function to obtain the SCL gradient signal, followed by: Obtain the pressure characteristics of the SCL gradient signal; The stress characteristics are normalized using a fractional normalization algorithm to obtain a standardized cycling stress index. The standardized cycling stress index is used to detect the user's psychological and physiological stress level.

7. The environmental perception assessment method based on electrodermal signals according to claim 3, characterized in that, The step of inputting the corresponding relationship data set into the perception prediction model for training to obtain the trained perception prediction model specifically includes: The corresponding data set is divided according to a preset ratio to obtain a training set and a validation set; The training set is input into the perception prediction model for training to obtain the trained perception prediction model. The performance of the trained perception prediction model is then verified using the validation set.

8. An environmental perception and assessment system based on electrodermal signals, characterized in that, The environmental perception and assessment system based on electrodermal signals includes: An environmental feature construction module is used to acquire panoramic image data and geographic information data of the target area, extract environmental features from the panoramic image data and geographic information data to obtain target features, and input the target features into a structured database for storage to obtain an environmental feature database. The environment model training module is used to acquire multi-dimensional environmental feature data from the environmental feature database, and input the target street image and the skin conductance signal of the multi-dimensional environmental feature data into the perception prediction model for training to obtain the trained perception prediction model. The environmental perception and prediction module is used to acquire the image data to be tested, input the image data to be tested into the trained perception and prediction model for prediction, and obtain the pressure perception prediction result.

9. A terminal, characterized in that, The terminal includes: a memory, a processor, and an environmental perception assessment program based on electrodermal signals stored in the memory and executable on the processor. When the environmental perception assessment program based on electrodermal signals is executed by the processor, it implements the steps of the environmental perception assessment method based on electrodermal signals as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an environmental perception assessment program based on electrodermal signals, which, when executed by a processor, implements the steps of the environmental perception assessment method based on electrodermal signals as described in any one of claims 1-7.

Images