Scenic area global tourist route dynamic generation method and system based on multi-modal AI

Abstract

The invention discloses a scenic spot global tourism route dynamic generation method and system based on multi-modal AI, and belongs to the technical field of smart tourism, and the method comprises the steps: setting an initial tourism route of a user; collecting user state data, user behavior data, scenic spot terrain data and scenic spot environment data; establishing an interest probability prediction model according to the user behavior data, and generating interest probability of the scenic spot type user; establishing a scenic spot value evaluation model according to the scenic spot type user interest probability, the user state data, the scenic spot terrain data and the scenic spot environment data, and generating a scenic spot value evaluation index; according to the scenic spot value evaluation index and historical data of scenic spot tourists, generating a scenic spot removal evaluation value; removing scenic spots in the initial tour route according to the scenic spot removal evaluation value, and generating a dynamic tour route; the problem that a traditional tourism route static generation system cannot meet real-time requirements of tourists is solved, and therefore the tourism experience quality and satisfaction degree of the tourists are improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of smart tourism, and particularly relates to a scenic spot global tourism route dynamic generation method and system based on multi-modal AI. BACKGROUND

[0002] With the rapid development of tourism, tourists' demand for personalized and intelligent tourism experience is increasing.

[0003] Traditional tourism route planning methods have obvious limitations: on the one hand, existing static recommendation systems only rely on historical tourist ratings or artificially designed fixed routes, and cannot real-time perceive changes in tourists' interest preferences; on the other hand, these systems lack the ability to integrate and analyze multi-dimensional dynamic factors, including tourists' real-time physiological state (such as heart rate, walking speed, etc.), behavioral characteristics (such as stay duration, photographing frequency, etc.), and environmental parameters (such as terrain slope, crowd density, etc.). This static planning mode can lead to the following problems with the recommended route: when the tourist is exhausted, high-intensity scenic spots are still recommended, when the interest shifts, the same type of scenic spots are still pushed, or when the crowd is high, popular scenic spots are recommended, thereby making the traditional tourism route planning method unable to adapt to tourists' real-time needs, reducing tourists' satisfaction and experience quality. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a scenic spot global tourism route dynamic generation method and system based on multi-modal AI, which solves the above problems.

[0005] To achieve the above purpose, the present application is implemented by the following technical scheme: a scenic spot global tourism route dynamic generation method based on multi-modal AI, comprising the following steps:

[0006] Setting an initial tourism route for a user;

[0007] Collecting user state data, user behavior data, scenic spot terrain data, and scenic spot environment data;

[0008] Establishing an interest probability prediction model according to the user behavior data to generate a scenic spot type user interest probability;

[0009] Establishing a scenic spot value evaluation model according to the scenic spot type user interest probability, the user state data, the scenic spot terrain data, and the scenic spot environment data to generate a scenic spot value evaluation index;

[0010] Obtaining historical data of scenic spot tourists, generating a scenic spot removal evaluation value in the initial tourism route according to the scenic spot value evaluation index and the historical data of scenic spot tourists;

[0011] Based on the attraction removal evaluation value in the initial tour route, attractions in the initial tour route are removed, and a dynamic tour route is generated.

[0012] Based on the above technical solutions, the present invention also provides the following optional technical solutions:

[0013] Further technical solution: The method for generating the user interest probability of the attraction type specifically includes the following steps:

[0014] Normalize user behavior data;

[0015] Construct a user behavior feature matrix based on the normalized user behavior data;

[0016] An interest probability prediction model is established based on the user behavior feature matrix to generate user interest probabilities for different attraction types.

[0017] Further technical solution: The expression of the interest probability prediction model is specifically as follows:

[0018] ;

[0019] In the expression, This represents the probability of user interest in the i-th type of attraction. This represents the i-th attraction type in the scenic area, B. i W represents the user behavior feature matrix for the i-th attraction type. i This represents the weight matrix of a user's behavioral characteristics at the i-th attraction type within a scenic area. The weight matrix W i transpose, b i B is the bias term for the i-th attraction type in the scenic area. j W represents the user behavior feature matrix for the j-th attraction type. j This represents the weight matrix of user behavior features at the j-th attraction type within the scenic area. The weight matrix W j transpose, b j Let n be the bias term for the j-th attraction type in the scenic area, and n be the number of scenic area types.

[0020] Further technical solution: The method for generating the scenic spot value assessment index specifically includes the following steps:

[0021] Based on user status data and scenic spot terrain data, a user physical load index is generated; the user status data includes user heart rate and user walking speed; the scenic spot terrain data includes scenic spot slope, scenic spot walking distance length and scenic spot road type.

[0022] Based on the scenic area's environmental data, an environmental degradation factor is generated.

[0023] A scenic spot value assessment model is established based on the probability of user interest in scenic spot type, user physical burden index and environmental degradation factor, and a scenic spot value assessment index is generated.

[0024] Further technical solution: The method for generating the user's physical load index specifically includes:

[0025] Through the formula:

[0026] ;

[0027] Generate user physical load index E k ;

[0028] In the formula, H represents the user's heart rate. This represents the increase in a user's heart rate after viewing attraction k. H represents the estimated increase in heart rate after a user completes visit k. max V represents the user's maximum safe heart rate, V represents the user's current walking speed, and V0 represents the average walking speed of tourists of the same age at the attraction. , All are weighting coefficients, and .

[0029] Further technical solution: The method for generating the heart rate growth estimate specifically includes:

[0030] Through the formula:

[0031] ;

[0032] Generate heart rate growth estimate ;

[0033] In the formula, This indicates the slope of the scenic spot, L. k This indicates the length of the tour route to the attractions. H is the indicator value for the road surface type in the scenic area. rec This indicates the user's heart rate recovery speed, H rest This represents the user's resting heart rate.

[0034] Further technical solution: The specific method for generating the environmental degradation factor includes:

[0035] Through the formula:

[0036] ;

[0037] Generate environmental degradation factor ;

[0038] In the formula, Q k t represents the normalized characteristic value of the real-time visitor density of attraction k. w,k T represents the normalized eigenvalue of the estimated waiting time for attraction k. k This represents the normalized characteristic value of the real-time temperature of scenic spot k.

[0039] Further technical solution: The specific expression of the scenic spot value assessment model is as follows:

[0040] ;

[0041] In the expression, A k This represents the scenic spot's value assessment index. R represents the environmental degradation factor. k This represents the inherent rating of attraction k, E. k This represents the user's physical burden index after browsing attraction k. denoted as , where represents the probability of user interest in the type of attraction k, and B is the user behavior feature matrix corresponding to attraction k.

[0042] Further technical solutions: The method for generating the attraction removal assessment value specifically includes:

[0043] Historical visitor data is obtained, and a scenic spot popularity correction coefficient is generated based on the historical number of visitors to the scenic spot. The historical visitor data includes the historical number of visitors to the scenic spot and the historical total number of visitors to the scenic spot.

[0044] Specifically:

[0045] Through the formula:

[0046] ;

[0047] Generate attraction popularity correction coefficient ;

[0048] In the formula, M k M0 represents the historical number of visitors to attraction k, and M0 represents the historical total number of visitors to the scenic area.

[0049] Based on the attraction popularity correction coefficient and the attraction value assessment index, an attraction removal assessment value is generated;

[0050] Specifically:

[0051] Through the formula:

[0052] ;

[0053] Generate attraction removal assessment value ;

[0054] In the formula, This represents the attraction popularity correction coefficient for attraction k.

[0055] A dynamic generation system for scenic area tourism routes based on multimodal AI is provided. This system is used to execute the aforementioned dynamic generation method for scenic area tourism routes based on multimodal AI, specifically including:

[0056] The initial setup unit is used to set the user's initial travel route;

[0057] The data acquisition unit is used to collect user status data, user behavior data, scenic spot terrain data, and scenic spot environmental data.

[0058] The interest analysis unit is used to build an interest probability prediction model based on user behavior data and generate user interest probabilities for attraction types.

[0059] The scenic spot value analysis unit is used to build a scenic spot value assessment model based on the user interest probability of scenic spot type, user status data, scenic spot terrain data, and scenic spot environment data, and generate a scenic spot value assessment index.

[0060] The attraction removal analysis unit is used to generate attraction removal assessment values ​​for the initial tourist route based on the attraction value assessment index.

[0061] The initial setting unit is used to remove attractions from the initial tourist route based on the attraction removal evaluation value in the initial tourist route, and generate a dynamic tourist route.

[0062] This invention provides a method and system for dynamically generating scenic area tourism routes based on multimodal AI, which has the following advantages compared with existing technologies:

[0063] This invention establishes a dynamic evaluation model by integrating user behavior data, physiological state, and environmental parameters, and adjusts the tour route in real time. This solves the problem that traditional static tour route generation systems cannot adapt to tourists' real-time needs, thereby improving the quality of tourists' travel experience and satisfaction. Attached Figure Description

[0064] Figure 1 This is a flowchart illustrating the method for dynamically generating scenic area tourism routes based on multimodal AI provided by the present invention.

[0065] Figure 2 The flowchart of step S30 provided by the present invention.

[0066] Figure 3 The flowchart of step S40 provided by the present invention.

[0067] Figure 4The flowchart of step S50 provided by the present invention.

[0068] Figure 5 This is a schematic diagram of the structure of the dynamic generation system for scenic area tourism routes based on multimodal AI provided by the present invention.

[0069] Figure 6 This is a block diagram of the interest analysis unit provided by the present invention.

[0070] Figure 7 This is a block diagram of the scenic spot value analysis unit provided by the present invention.

[0071] Figure 8 This is a block diagram of the attraction removal analysis unit provided by the present invention. Detailed Implementation

[0072] To make the objectives, technical solutions, and advantages of this invention clearer, 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 and not intended to limit the invention.

[0073] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0074] Please see Figure 1 The present invention provides a method for dynamically generating scenic area tourism routes based on multimodal AI, comprising the following steps:

[0075] Step S10: Set the user's initial travel route;

[0076] Specifically, the initial travel route can be set by the scenic area based on factors such as the location and popularity of attractions within the area. Alternatively, the initial travel route can be set using existing navigation software (such as Gaode Maps), which involves setting the starting point and destination, and then adding waypoints to generate the route.

[0077] Step S20: Collect user status data, user behavior data, scenic spot terrain data, and scenic spot environment data;

[0078] Step S30: Build an interest probability prediction model based on user behavior data to generate user interest probabilities for attraction types;

[0079] Step S40: Establish a scenic spot value assessment model based on the user interest probability of scenic spot type, user status data, scenic spot terrain data, and scenic spot environment data, and generate a scenic spot value assessment index;

[0080] Step S50: Obtain historical data of tourists in the scenic area, and generate an assessment value for removing attractions in the initial tourist route based on the attraction value assessment index and the historical data of tourists in the scenic area.

[0081] Step S60: Based on the attraction removal evaluation value in the initial tour route, remove attractions from the initial tour route and generate a dynamic tour route;

[0082] Among them, user status data refers to information reflecting the real-time physiological status of tourists. Specifically, heart rate and step speed data can be collected through wearable devices and motion sensors to assess the tourist's current physical load capacity.

[0083] User behavior data refers to the interactive behavior characteristics of tourists in scenic spots, such as calculating the length of stay through GPS positioning, recording the frequency of taking photos by cameras, and extracting keywords through voice recognition, which are used to construct a user interest feature matrix.

[0084] The topographical data of the scenic spot includes spatial information such as slope and distance, which can be obtained through a geographic information system and used to calculate the physical exertion of the tour route;

[0085] The scenic spot's environmental data includes dynamic parameters such as real-time visitor flow and temperature, which can be collected through an Internet of Things (IoT) sensor network to assess environmental comfort.

[0086] Interest probability prediction model refers to a machine learning model that maps user behavior feature matrix to attraction type preference probability, and is used to quantify the intensity of tourists' interest in different types of attractions;

[0087] The scenic spot value assessment model refers to a comprehensive assessment function that integrates interest probability, physical burden index and environmental factors, and is used to dynamically calculate the real-time visit value of each scenic spot.

[0088] Specifically, this method first collects users' physiological indicators such as heart rate and walking speed through multi-source sensors, and combines this with geographical data such as scenic spot slope and distance length to calculate the physical load required for users to complete a visit to a scenic spot in real time. Simultaneously, it analyzes behavioral characteristics such as visitor dwell time and photo frequency, and uses a classification model to predict the probability of their preference for different types of scenic spots. It further integrates real-time data such as pedestrian flow and temperature obtained from environmental sensors to calculate the impact of environmental degradation factors on the value of scenic spots. By weighting and fusing interest probability, physical load index, and environmental factors, a dynamically updated scenic spot value assessment index is generated. When the assessment index of a scenic spot falls below a preset threshold, the system automatically generates removal suggestions, dynamically adjusting low-value scenic spots in the initial route to form an optimized route that adapts to the real-time status of visitors and environmental changes.

[0089] Compared to existing technologies, traditional methods typically use fixed weights to calculate recommended attraction values, failing to establish a dynamic correlation mechanism between physical load and environmental factors. For example, existing systems may only recommend attractions based on historical ratings, without considering whether the tourist's current heart rate allows for a visit to a steep attraction, or ignoring the impact of sudden high temperatures on the outdoor attraction experience. This invention, through multimodal data fusion, incorporates the tourist's physiological state and real-time environmental changes into the evaluation system, enabling route adjustments to simultaneously satisfy both interests and objective constraints.

[0090] Through the above technical solution, this invention effectively solves the problem of rigid route planning in traditional methods. By monitoring the user's physical load in real time, it avoids recommending attractions beyond their physical limits; by dynamically assessing environmental factors, it avoids the decline in experience caused by overcrowding or inclement weather; and by accurately predicting interests and preferences through behavioral data analysis, it improves the matching accuracy of attractions. This method achieves a technological leap from static recommendation to dynamic optimization, significantly improving the personalization and adaptability of travel route planning.

[0091] For preferred options, please refer to [link / reference]. Figure 2 The present invention also proposes a method for generating the user interest probability of the aforementioned attraction type, specifically including the following steps:

[0092] Step S31: Normalize the user behavior data; whereby the user behavior data includes the duration of stay at the attraction, the frequency of taking photos, and voice keywords;

[0093] Step S32: Construct a user behavior feature matrix based on the normalized user behavior data;

[0094] Step S33: Establish an interest probability prediction model based on the user behavior feature matrix to generate user interest probabilities for attraction types;

[0095] Normalization refers to the process of converting user behavior data of different dimensions or magnitudes into a uniform numerical range. Specifically, it can be achieved by using min-max normalization or Z-score standardization to eliminate the impact of data scale differences on model training.

[0096] The user behavior feature matrix refers to a multi-dimensional data set composed of normalized time spent at attractions, frequency of taking photos, and semantic analysis results of voice keywords. Specifically, it can be constructed by matrix splicing to comprehensively represent users' explicit behaviors and implicit interest preferences at attractions.

[0097] Interest probability prediction model refers to a mathematical model that predicts the probability of a user's preference for different types of attractions based on the user's behavior feature matrix. Specifically, it can be implemented by using the Softmax function combined with a linear transformation layer to map high-dimensional features into a probability distribution.

[0098] Specifically, the duration of visit to attractions in user behavior data is calculated using timestamp differences, the photo-taking frequency is determined by counting the number of times the image acquisition device is triggered, and voice keywords are extracted using natural language processing techniques. Normalization employs a categorized standardization strategy; for example, the duration of visit is scaled based on the maximum observed value, the photo-taking frequency is calculated based on the average number of times per unit time, and voice keywords are converted into numerical features using word frequency-inverse document frequency or semantic analysis. The user behavior feature matrix is ​​generated by combining the processed data into a fixed-dimensional array with preset weights, inputting it into a linear transformation layer containing a weight matrix and bias terms, and then normalizing it using the Softmax function to output the probability values ​​for each attraction type. Furthermore, during model training, the weight parameters can be optimized using the backpropagation algorithm to ensure that the predicted probabilities match the actual user visit behavior.

[0099] Compared to existing technologies, traditional methods typically rely on single behavioral indicators or static historical data for interest prediction, such as analyzing only dwell time or manually labeled keywords, thus ignoring the correlation and dynamic changes between multi-dimensional behaviors. This invention, by fusing multimodal data such as dwell time, photo frequency, and voice keywords, and employing normalization and feature matrix construction methods, can more comprehensively capture the explicit and implicit expressions of user interests. Simultaneously, it utilizes probabilistic models to quantify the degree of preference, overcoming the prediction bias caused by the limited data in traditional methods.

[0100] Through the above technical solution, this invention solves the accuracy problem in user interest prediction caused by the single data dimension and insufficient dynamic adaptability. By integrating multi-source behavioral data and constructing a structured feature matrix, the model can simultaneously capture users' interest expressions at the temporal, visual, and semantic levels, improving the comprehensiveness of interest type identification. Normalization ensures the balanced contribution of different behavioral indicators in model training, avoiding weight shifts caused by differences in units. Probabilistic output provides quantifiable preference basis for subsequent scenic spot value assessment, supporting accurate decision-making in dynamic route adjustments.

[0101] Preferably, the present invention further proposes the following expression for the interest probability prediction model:

[0102] ;

[0103] In the expression, This represents the probability of user interest in the i-th type of attraction. This represents the i-th attraction type in the scenic area, B. i W represents the user behavior feature matrix for the i-th attraction type. i This represents the weight matrix of a user's behavioral characteristics at the i-th attraction type within a scenic area. The weight matrix W itranspose, b i B is the bias term for the i-th attraction type in the scenic area. j W represents the user behavior feature matrix for the j-th attraction type. j This represents the weight matrix of user behavior features at the j-th attraction type within the scenic area. The weight matrix W j transpose, b j Let n be the bias term for the j-th attraction type in the scenic area, and n be the number of scenic area types.

[0104] Among them, the user behavior feature matrix refers to the multi-dimensional feature set formed after normalizing user behavior data. Specifically, it can be achieved by using linear weighting or principal component analysis to reduce the dimensionality of user behavior data such as the duration of stay at the scenic spot, the frequency of taking photos, and voice keywords, in order to characterize user behavior patterns.

[0105] W i W j It refers to a weight matrix that reflects the strength of the correlation between different types of tourist attractions and user behavior characteristics. , Wj is the transpose of the weight matrices Wi and Wj, which can be obtained by training historical user behavior data using the gradient descent algorithm. It is used to dynamically adjust the contribution of features to the probability of interest.

[0106] Bias term b i b j This refers to a constant term used to correct the position of the classification decision surface. Specifically, it can be implemented using the intercept term calculation method in the logistic regression model, and is used to optimize the probability distribution boundary of different types of scenic spots.

[0107] Specifically, the user behavior feature matrix integrates behavioral data such as user dwell time and photo frequency to form a multi-dimensional feature input reflecting behavioral preferences. After a linear combination of the weight matrix and the user behavior feature matrix, an initial score for each attraction type is generated using a bias term. This initial score is mapped to a positive number space, amplifying the probability differences of high-scoring types. The scores for all types are then normalized using a summation term in the denominator, ultimately outputting the probability of user interest in each type of attraction. For example, when a user's dwell time at a certain type of attraction increases significantly, the corresponding weight matrix will enhance the contribution of that behavioral feature to the probability of interest, thereby dynamically increasing the recommendation priority of that type of attraction. This model can adaptively capture changes in user interests, avoiding recommendation bias caused by data lag in traditional static models.

[0108] Compared to existing technologies, traditional methods often rely on fixed rules or single ratings to predict interest probabilities, such as statistical models based on historical visit frequencies, which fail to capture the non-linear correlation between real-time behavioral data and attraction types. This invention, by introducing trainable weight matrices and bias terms, combined with a normalized exponential function, achieves dynamic modeling of multi-dimensional behavioral features, solving the problem of insufficient recommendation accuracy caused by traditional models neglecting feature interactions and real-time updates.

[0109] Through the above technical solution, this invention can accurately predict the probability distribution of interest in different types of attractions based on real-time user behavior data, overcoming the shortcomings of traditional recommendation systems that rely on static data and cannot reflect dynamic preferences. By linearly combining the feature matrix and weight matrix, the model's ability to analyze the relationship between user behavior patterns and attraction types is enhanced, thereby improving the accuracy and real-time performance of personalized travel route recommendations.

[0110] For preferred options, please refer to [link / reference]. Figure 3 The present invention further proposes a method for generating the scenic spot value assessment index, specifically including the following steps:

[0111] Step S41: Generate the user's physical load index based on user status data and scenic spot terrain data; wherein, user status data includes user heart rate and user walking speed; scenic spot terrain data includes scenic spot terrain slope, scenic spot walking distance length and scenic spot road surface type;

[0112] Step S42: Generate an environmental degradation factor based on the scenic area's environmental data;

[0113] Step S43: Establish a scenic spot value assessment model based on the user interest probability of scenic spot type, user physical load index and environmental degradation factor, and generate scenic spot value assessment index;

[0114] Among them, the user physical load index is a quantitative indicator of physical exertion established by the correlation between user heart rate, walking speed and the slope, distance and road type of the scenic spot. Specifically, it can be realized by calculating the comprehensive impact of user heart rate changes and terrain parameters using a formula, and is used to assess the user's physical endurance when visiting scenic spots.

[0115] The environmental degradation factor is a dynamic adjustment coefficient that reflects the impact of real-time visitor flow, waiting time and temperature on the visitor experience. Specifically, it can be achieved by weighting the normalized environmental parameters using an exponential function, and is used to dynamically analyze the state of the scenic spot's environment.

[0116] The scenic spot value assessment model refers to a multimodal data calculation framework that integrates interest probability, physical burden, and environmental degradation to comprehensively quantify the visit value of a scenic spot at a specific point in time.

[0117] Specifically, in the process of generating the user's physical load index, user heart rate data is linearly correlated with terrain slope and distance. For example, when the slope increases or the distance lengthens, the increase in heart rate is dynamically calculated and incorporated into the load assessment. The environmental decay factor is calculated by collecting real-time data on pedestrian density, estimated waiting time, and temperature. For example, when the number of visitors to a scenic spot exceeds a threshold or the temperature is abnormal, the decay factor will exponentially reduce the scenic spot's value score. The scenic spot value assessment model uses interest probability as the basic weight and dynamically adjusts it in conjunction with the user's physical load index and the environmental decay factor. For example, when a user has a high interest in a certain type of scenic spot but is currently experiencing excessive physical exertion, the model will automatically balance the relationship between the two to generate an optimized assessment index.

[0118] Compared to existing technologies, traditional attraction recommendation methods are typically based solely on static ratings or historical data, failing to reflect real-time user physical condition and environmental changes. For example, existing technologies do not consider the impact of terrain slope on heart rate, nor do they incorporate real-time visitor flow and temperature as dynamic adjustment factors. This invention, through multimodal data fusion, achieves accurate prediction of user physical load and real-time response to environmental factors, enabling the attraction value assessment results to be dynamically updated according to changes in user status and external conditions.

[0119] Through the above technical solution, this invention solves the problem that traditional methods cannot dynamically adjust routes in real time by combining user physical exertion, environmental changes, and interests. The introduction of the user physical load index avoids tour interruptions caused by physical exhaustion, the dynamic calculation of the environmental attenuation factor reduces the impact of crowds or bad weather on the experience, and the multimodal fusion model ensures a balanced optimization of attraction recommendations in terms of interests, physical endurance, and environmental adaptation.

[0120] Preferably, the present invention further proposes a method for generating the user's physical load index, specifically including:

[0121] Through the formula:

[0122] ;

[0123] Generate user physical load index E k ;

[0124] In the formula, H represents the user's heart rate. This represents the increase in a user's heart rate after viewing attraction k. H represents the estimated increase in heart rate after a user completes visit k. max V represents the user's maximum safe heart rate, V represents the user's current walking speed, and V0 represents the average walking speed of tourists of the same age at the attraction. , All are weighting coefficients, and ;

[0125] Among them, the user's heart rate H refers to a physiological indicator monitored in real time through wearable devices, specifically smart bracelets or heart rate belts, which can be used to collect data to reflect the user's current physical state.

[0126] Heart rate increase value It refers to the change in heart rate after a user completes a visit to a scenic spot. Specifically, it can be predicted using historical exercise data and terrain parameters to quantify the impact of visiting scenic spots on the user's physical load.

[0127] Maximum safe heart rate H max This refers to a heart rate threshold set based on the user's age and health condition to prevent the user from overexerting themselves.

[0128] The user's current walking speed V refers to the movement speed obtained through an accelerometer or GPS positioning data. Specifically, it can be measured in real time using the built-in sensor of the mobile terminal to assess the compatibility between the user's movement rhythm and the sightseeing route.

[0129] The average walking speed V0 of tourists of the same age refers to the average movement speed of a group of tourists matching the user's age group within the same scenic spot. This can be obtained through historical data statistics from the scenic area management system and used to establish a dynamic reference benchmark.

[0130] Weighting coefficient , This refers to the adjustment parameter used to balance the influence of heart rate and walking speed indicators. Specifically, it can be determined using a dynamic allocation algorithm based on the user's physical fitness test results. For example, a higher heart rate weight coefficient can be set for users with weaker physical fitness.

[0131] Specifically, the calculation of the user's physical load index is broken down into a weighted sum of heart rate-related terms and walking speed-related terms. Heart rate-related terms The influence of individual differences on the assessment results was eliminated by normalizing the ratio of the sum of real-time heart rate and predicted heart rate increases to the maximum safe heart rate. (Step speed related items) The ratio of a user's current walking speed to the average walking speed of tourists of the same age reflects the degree of match between the user's physical ability and the terrain requirements of the scenic spot. Weighting coefficient , The system dynamically adjusts its parameters based on the user's physical characteristics. For example, for users with cardiovascular health issues, the system automatically increases the weighting of heart rate-related parameters. By collecting user physiological data in real time and combining it with scenic spot terrain parameters, the system can dynamically assess the user's physical exertion trends during the tour, providing a quantitative basis for subsequent route optimization.

[0132] Compared to existing technologies, traditional methods typically rely solely on static step counts or fixed heart rate thresholds to assess user fatigue. They fail to incorporate dynamic parameters such as terrain slope and path length to predict heart rate changes and lack adaptation analysis to the user's exercise rhythm and sightseeing needs. This invention, by integrating real-time physiological data, terrain parameters, and group behavior data, constructs a dynamic weighting mechanism. This ensures that the physical load assessment results reflect the user's immediate state while adapting to the terrain characteristics of different scenic spots.

[0133] Through the above technical solution, this invention achieves accurate prediction of users' physical exertion, avoiding tour interruptions or health risks caused by route planning exceeding users' capabilities. By dynamically adjusting the weighting of heart rate and walking speed, the system can adapt to the personalized needs of users with different fitness levels, improving the matching degree between route planning and users' actual exercise capabilities. Simultaneously, based on the reference benchmark of the average walking speed of tourists of the same age, it solves the problem of potential random bias in individual user data, enhancing the objectivity of the physical load assessment results.

[0134] Preferably, the present invention further proposes a method for generating the heart rate growth estimate, specifically including:

[0135] Through the formula:

[0136] ;

[0137] Generate heart rate growth estimate ;

[0138] In the formula, This indicates the slope of the scenic spot, L. k This indicates the length of the tour route to the attractions. H is the indicator value for the road surface type in the scenic area. rec This indicates the user's heart rate recovery speed, H rest This represents the user's resting heart rate;

[0139] Among them, the terrain slope of the scenic spot It refers to the ground tilt angle of the area where the scenic spot is located. It can be measured using a geographic information system or slope sensor to quantify the impact of terrain on the user's physical exertion.

[0140] Scenic spot tour route length L k This refers to the total distance a user needs to walk within the scenic area. It can be determined by combining scenic area map data with GPS positioning, reflecting the spatial cumulative effect of the user's physical exertion.

[0141] The sign value of scenic road surface type It refers to the influence coefficient of different road surface materials on walking resistance. For example, stone slab road can be set to 1.2, and asphalt road can be set to 1.0. It is achieved by preset a database of different road surface marker values ​​and is used to correct the calculation of physical exertion under different road conditions.

[0142] User's heart rate recovery speed H rec It refers to the user's ability to recover their heart rate from an active state to a resting state per unit of time. Specifically, it can be calculated by monitoring the rate of heart rate decline after exercise through wearable devices and is used to assess the individual physiological recovery characteristics of users.

[0143] User's resting heart rate H rest It refers to the user's baseline heart rate at rest, which can be achieved by continuously monitoring and averaging the heart rate while the user is at rest using wearable devices, and serves as a benchmark parameter for personalized heart rate calculation.

[0144] Specifically, this technical solution calculates the estimated heart rate increase by fusing multi-dimensional parameters. Among them, This system combines terrain slope, distance, road surface type, and walking speed to calculate the time required for a user to traverse a scenic spot. The terrain slope is treated with a sine function, reflecting the actual phenomenon that physical exertion increases non-linearly with increasing slope. Road surface type indicator value... Weighting is applied to different road conditions, such as gravel roads. The value is greater than that of a flat road surface. The cumulative effect of decreased walking speed leading to prolonged traverse time on heart rate is demonstrated by the reciprocal of the user's current walking speed V. This study introduces the ratio of heart rate recovery speed to heart rate reserve to dynamically adjust for individual user differences: for users with strong recovery capabilities, this value approaches 1, reducing the impact of terrain time on heart rate; for users with weak recovery capabilities, this value approaches 0, amplifying the impact of terrain time. The final (H) max -H rest This option constrains the calculation results within the user's individual safe heart rate range, preventing the estimated value from exceeding physiological limits.

[0145] Compared to existing technologies, traditional methods typically estimate heart rate growth based solely on fixed formulas, failing to consider individual differences in users' resting heart rate and recovery ability, leading to calculation results that deviate from actual physiological states. For example, existing technologies may only use a linear combination of slope and distance to estimate load, ignoring the influence of road surface type on walking resistance. This invention introduces a road surface type indicator value. Establish a physical exertion correction mechanism for different road conditions; by integrating H rec and H rest Parameters are used to construct a personalized heart rate response model, so that the estimated value reflects both environmental factors and adapts to the user's real-time physiological characteristics.

[0146] Through the above technical solution, this invention effectively solves the problem of heart rate prediction bias caused by neglecting individual physiological differences in traditional methods. By integrating terrain parameters, road condition data, and real-time physiological indicators of users, it achieves accurate prediction of heart rate changes after a user visits a scenic spot. This prediction value can provide accurate physical load data for subsequent scenic spot value assessment, ensuring that dynamically generated travel routes not only conform to user interests and preferences but also match their physiological endurance, avoiding unreasonable route planning problems caused by load estimation bias.

[0147] Preferably, the present invention further proposes a method for generating the environmental degradation factor, specifically including:

[0148] Through the formula:

[0149] ;

[0150] Generate environmental degradation factor ;

[0151] In the formula, Q k t represents the normalized characteristic value of the real-time visitor density of attraction k. w,k T represents the normalized eigenvalue of the estimated waiting time for attraction k. k This represents the normalized characteristic value of the real-time temperature of scenic spot k;

[0152] Among them, real-time visitor density refers to the dynamic monitoring value of the number of tourists per unit area. Specifically, it can be achieved by combining camera image recognition with infrared sensor data fusion, and is used to quantify the negative impact of overcrowding on the visitor experience.

[0153] Estimated waiting time refers to the predicted length of time tourists need to queue before entering an attraction. It can be achieved by using regression analysis of historical queuing data and real-time tourist flow to reflect the depreciation effect of time costs on the value of the attraction.

[0154] Real-time temperature refers to the current temperature measurement value of the area where the scenic spot is located. Specifically, it can be collected using an Internet of Things temperature sensor network to capture the dynamic impact of climate conditions on the comfort of the visitor.

[0155] The normalized characteristic values ​​of real-time pedestrian flow density, estimated waiting time, and real-time temperature refer to the standardization process of mapping the original data to the [0,1] interval through linear transformation. Specifically, the maximum and minimum value normalization method can be used to eliminate the dimensional differences of different environmental parameters and achieve weighted fusion.

[0156] Specifically, the environmental degradation factor achieves nonlinear adjustment by fusing three types of dynamic environmental parameters. The normalized characteristic value of real-time pedestrian density is collected and standardized in real-time by monitoring equipment, reflecting the congestion level of the attraction; the normalized characteristic value of estimated waiting time is dynamically calculated based on a queuing model, characterizing the time consumption cost; and the normalized characteristic value of real-time temperature is continuously updated through a sensor network, indicating environmental comfort. The three parameters are multiplied by their corresponding weighting coefficients and then summed, with the result mapped to the (0,1] interval using the exponential function exp(-x). When environmental conditions deteriorate, the weighted sum increases, leading to… When the value approaches zero, it has a diminishing effect on subsequent assessments of the scenic spot's value; when environmental conditions are excellent, The value approaches one, preserving the inherent value of the attraction. This calculation method makes the adjustment of the attraction's value by environmental factors exhibit non-linear characteristics, which is more in line with the increasing marginal effect of environmental degradation in actual experience.

[0157] Compared to existing technologies, traditional tourism route planning systems often employ fixed environmental parameters or single-dimensional evaluation, such as setting static weights based solely on historical visitor flow, which cannot respond to sudden surges in visitor numbers during holidays or unpredictable weather changes. This invention, through real-time acquisition of multi-source data, normalization processing, and nonlinear function fusion, achieves dynamic and precise adjustment of environmental factors in the assessment of scenic spot value.

[0158] Through the above technical solution, this invention solves the problems of delayed updates to environmental factors and inaccurate quantification of impact in traditional methods. For example, in a theme park scenario, when the temperature in the queuing area of ​​a ride drops due to sudden rainfall, the system automatically reduces the environmental degradation factor of that ride to avoid incorrectly increasing its recommendation priority due to unsuitable temperature. When a popular exhibition hall experiences a surge in visitor density due to instantaneous overcrowding, the degradation factor rapidly decreases, triggering route adjustments and guiding visitors to alternative attractions with shorter waiting times. This technology achieves real-time coupling between changes in environmental parameters and dynamic route optimization, ensuring that recommended routes always match the actual experience needs of visitors.

[0159] Preferably, the present invention further proposes the following expression for the scenic spot value assessment model:

[0160] ;

[0161] In the expression, A k This represents the scenic spot's value assessment index. R represents the environmental degradation factor. k This represents the inherent rating of attraction k, E. k This represents the user's physical burden index after browsing attraction k. B represents the probability of user interest in the type of attraction to which attraction k belongs, and B is the user behavior feature matrix corresponding to attraction k.

[0162] Among them, environmental degradation factor It refers to dynamic parameters that reflect the impact of the current environment of a scenic spot on its attractiveness to tourists. Specifically, it can be achieved by collecting real-time data on the density of visitors, estimated waiting time, and temperature, and then normalizing the data using an exponential function. This is used to quantify the decay effect of environmental factors on the value of a scenic spot.

[0163] Inherent rating R k It refers to the static value indicators of a scenic spot based on historical evaluation and expert assessment. Specifically, it can be achieved by integrating the scenic spot's popularity, cultural value, and visitor rating data, in order to preserve the basic quality attributes of the scenic spot.

[0164] User physical load index E k It refers to a quantitative indicator of the physical exertion of users when visiting attractions. Specifically, it can be achieved by building a calculation model that combines parameters such as user heart rate, walking speed and the terrain slope and distance of the attractions, in order to assess the physical cost of the user's visit.

[0165] Tourist Attraction Type User Interest Probability This refers to the probability of a user's preference for the type of attraction k. Specifically, it can be achieved by constructing a feature matrix by analyzing behavioral data such as user dwell time and photo frequency, and then using a classification probability model for prediction, in order to reflect the dynamic changes in user interests.

[0166] Specifically, this technical solution uses environmental degradation factors With inherent rating R k The product of these factors is used as the numerator, representing the user's physical load index E. k As the denominator, we construct the ratio between the value of the attraction and the physical exertion. Environmental degradation factor. By monitoring environmental parameters such as visitor flow and temperature in real time, the attractiveness assessment of tourist attractions is dynamically adjusted. For example, when the visitor flow of a tourist attraction exceeds a threshold, its environmental degradation factor is adjusted. It will decrease significantly. User physical burden index E k Combining real-time heart rate and walking speed data with scenic spot terrain parameters, such as the increase in heart rate at scenic spots with steep slopes, the data can be used to determine the increase in heart rate. The parameters are dynamically calculated and incorporated into the load assessment. Interest probability is based on a user behavior feature matrix; for example, when a user stays longer than the average time at a certain type of attraction, the interest probability for that type is increased. By multiplying these parameters as an attraction value assessment index, a logic for selecting attractions with high interest matching, low physical exertion, and suitable environmental conditions is achieved.

[0167] Compared to existing technologies, traditional methods typically rely solely on static attraction ratings or single user preference data for route recommendations. This invention, however, constructs a multi-dimensional evaluation model by integrating real-time environmental data, user physiological status, and dynamic interest preferences. Existing technologies cannot address the issue of inappropriate routes caused by sudden environmental changes or insufficient user physical strength. For example, high temperatures may reduce the attractiveness of attractions without being detected by the system, or users may be recommended high-intensity attractions despite being physically exhausted. This invention, through the dynamic calculation of environmental degradation factors and physical load indices, can identify such problems in real time and adjust the evaluation results accordingly.

[0168] Through the above technical solution, this invention can automatically optimize routes based on the user's real-time status, avoiding tour interruptions caused by environmental degradation or physical overexertion. For example, when a rainstorm causes a surge in visitors to a scenic spot, the system automatically lowers its evaluation index by reducing the environmental degradation factor and prioritizes recommending indoor alternative attractions. When a user's heart rate rises abnormally, the system lowers the evaluation value of high-intensity attractions by increasing the physical load index, thereby generating a dynamic route that matches the user's actual tolerance. Simultaneously, dynamic prediction of interest probability ensures that the recommended route always matches the user's preferences. For example, for users with a high probability of interest in historical and cultural attractions, the system will prioritize retaining such attractions under the same conditions.

[0169] For preferred options, please refer to [link / reference]. Figure 4 The present invention further proposes a method for generating the attraction removal evaluation value, specifically including:

[0170] Step S51: Obtain historical tourist data and generate a scenic spot popularity correction coefficient based on the historical number of tourists visiting the scenic spot; wherein, the historical tourist data of the scenic spot includes the historical number of tourists visiting the scenic spot and the historical total number of tourists visiting the scenic spot.

[0171] Specifically:

[0172] Through the formula:

[0173] ;

[0174] Generate attraction popularity correction coefficient ;

[0175] In the formula, M k M0 represents the historical number of visitors to attraction k, and M0 represents the historical total number of visitors to the scenic area.

[0176] Step S52: Generate the attraction removal assessment value based on the attraction popularity correction coefficient and the attraction value assessment index;

[0177] Specifically:

[0178] Through the formula:

[0179] ;

[0180] Generate attraction removal assessment value ;

[0181] In the formula, This represents the attraction popularity correction coefficient for attraction k;

[0182] Among them, the attraction removal assessment value U k It refers to a quantitative indicator used to assess whether to remove a specific attraction from the initial route. Specifically, it can be achieved by multiplying the attraction value assessment index and the attraction popularity correction coefficient, and by combining the real-time value of the attraction with historical popularity data through mathematical modeling.

[0183] Tourist attraction popularity correction factor This refers to the parameter used to adjust the impact of the distribution of historical visitors to a scenic spot on the removal decision. Specifically, it can be calculated using the normalized proportion of historical visitors. When the proportion of historical visitors to a scenic spot is high, the correction coefficient is reduced, and when the proportion is low, the correction coefficient is increased.

[0184] Historical number of visitors to attraction k M k This refers to the total number of visitors received by a single scenic spot within a statistical period. It can be specifically obtained through the scenic spot's ticketing system or visitor location data collection, and is used to reflect the long-term popularity distribution characteristics of the scenic spot.

[0185] The total number of historical visitors to a scenic area, M0, refers to the total number of visitors received by the target scenic area within the statistical period. This number can be obtained from the aggregated data of the scenic area management system and serves as the benchmark for calculating the popularity ratio of a single attraction.

[0186] Specifically, in the calculation of the attraction removal assessment value, the attraction popularity correction coefficient is dynamically adjusted based on the ratio of historical visitor numbers to the total number of visitors to the scenic area. When the historical visitor number ratio of a certain attraction approaches 0 (i.e., fewer visitors visit this attraction compared to other attractions), the correction coefficient approaches 1. At this point, the removal assessment value mainly relies on the attraction value assessment index A. k When the proportion approaches 1, the correction coefficient approaches 2, significantly improving the scenic spot value assessment index A of popular attractions. k This design enhances the likelihood of retaining the data. It transforms long-term statistical data on popularity distribution into quantifiable correction parameters, which, together with real-time assessed attraction value, contribute to dynamic route optimization, avoiding route homogenization caused by over-reliance on real-time data.

[0187] In some specific implementations, the statistical period for historical visitor numbers can be data from the past year, while the total number of visitors to the scenic area corresponds to the sum of visitors to all attractions during the same period. Furthermore, the calculation of the attraction popularity correction coefficient can be further processed using a logarithmic function to reduce the impact of extreme values ​​on the coefficient.

[0188] Compared to existing technologies, traditional route planning methods typically rely solely on real-time visitor flow or static ratings to recommend attractions, neglecting the long-term impact of historical popularity distribution on the visitor experience. For example, existing systems may continuously recommend historically high-visit attractions, leading to overcrowding and a decline in the visitor experience. This invention, however, uses a dynamic correction mechanism to overlay historical popularity weights on top of real-time evaluations, automatically balancing the recommendation priorities of popular and less popular attractions during route generation. This avoids overcrowding while enhancing route diversity.

[0189] Through the above technical solution, this invention solves the problem of route homogenization caused by ignoring historical popularity data in traditional route recommendations, and realizes comprehensive decision-making based on long-term statistics and real-time evaluation. By transforming historical tourist distribution into quantifiable parameters through mathematical modeling, the system can dynamically adjust attraction retention strategies, optimize the allocation of scenic area resources while ensuring tourist experience, and improve the rationality and sustainability of personalized route recommendations.

[0190] Please see Figure 5 This invention further proposes a dynamic generation system for scenic area tourism routes based on multimodal AI. This system is used to execute the aforementioned dynamic generation method for scenic area tourism routes based on multimodal AI, specifically including:

[0191] The initial setting unit 10 is used to set the user's initial travel route;

[0192] Data acquisition unit 20 is used to collect user status data, user behavior data, scenic spot terrain data, and scenic spot environment data;

[0193] Interest analysis unit 30 is used to build an interest probability prediction model based on user behavior data and generate user interest probability for attraction types.

[0194] The scenic spot value analysis unit 40 is used to establish a scenic spot value assessment model based on the user interest probability of scenic spot type, user status data, scenic spot terrain data and scenic spot environment data, and generate a scenic spot value assessment index.

[0195] The attraction removal analysis unit 50 is used to generate attraction removal assessment values ​​in the initial tourist route based on the attraction value assessment index.

[0196] The initial setting unit 60 is used to remove attractions from the initial tourist route based on the attraction removal evaluation value in the initial tourist route, and generate a dynamic tourist route.

[0197] For preferred options, please refer to [link / reference]. Figure 6 The present invention further proposes that the interest analysis unit 30 specifically includes:

[0198] Data processing module 31 is used to normalize user behavior data;

[0199] The matrix construction module 32 is used to construct a user behavior feature matrix based on the normalized user behavior data.

[0200] The interest probability generation module 33 is used to establish an interest probability prediction model based on the user behavior feature matrix and generate the user interest probability of scenic spot type.

[0201] For preferred options, please refer to [link / reference]. Figure 7 The present invention further proposes that the scenic spot value analysis unit 40 specifically includes:

[0202] The load analysis module 41 is used to generate a user physical load index based on user status data and scenic spot terrain data; wherein, user status data includes user heart rate and user walking speed; scenic spot terrain data includes scenic spot terrain slope, scenic spot walking distance length and scenic spot road surface type;

[0203] The environmental analysis module 42 is used to generate environmental degradation factors based on the scenic area's environmental data.

[0204] The value analysis module 43 is used to establish a scenic spot value assessment model based on the probability of user interest in scenic spot type, user physical load index and environmental decay factor, and generate scenic spot value assessment index.

[0205] For preferred options, please refer to [link / reference]. Figure 8 The present invention further proposes that the attraction removal analysis unit 50 specifically includes:

[0206] The correction coefficient generation module 51 is used to acquire historical tourist data and generate a scenic spot popularity correction coefficient based on the historical number of tourists visiting the scenic spot; wherein, the historical tourist data of the scenic spot includes the historical number of tourists visiting the scenic spot and the historical total number of tourists visiting the scenic spot.

[0207] The removal assessment value generation module 52 is used to generate a removal assessment value for a scenic spot based on the scenic spot popularity correction coefficient and the scenic spot value assessment index.

[0208] Although embodiments of the invention 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 invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for dynamically generating scenic area tourism routes based on multimodal AI, characterized in that, Includes the following steps: Set the user's initial travel route; Collect user status data, user behavior data, scenic spot terrain data, and scenic spot environment data; An interest probability prediction model is built based on user behavior data to generate user interest probabilities for different attraction types. A scenic spot value assessment model is established based on the probability of user interest in scenic spot type, user status data, scenic spot terrain data, and scenic spot environment data, and a scenic spot value assessment index is generated. Obtain historical data on tourists in the scenic area, and generate an initial tourist route assessment value for removing attractions based on the attraction value assessment index and the historical data on tourists in the scenic area. Based on the attraction removal evaluation value in the initial tour route, attractions in the initial tour route are removed, and a dynamic tour route is generated.

2. The method for dynamically generating scenic area tourism routes based on multimodal AI according to claim 1, characterized in that, The method for generating the user interest probability of the attraction type specifically includes the following steps: Normalize user behavior data; Construct a user behavior feature matrix based on the normalized user behavior data; An interest probability prediction model is established based on the user behavior feature matrix to generate user interest probabilities for different attraction types.

3. The method for dynamically generating scenic area tourism routes based on multimodal AI according to claim 2, characterized in that, The specific expression for the interest probability prediction model is as follows: ； In the expression, This represents the probability of user interest in the i-th type of attraction. This represents the i-th attraction type in the scenic area, B. i W represents the user behavior feature matrix for the i-th attraction type. i This represents the weight matrix of a user's behavioral characteristics at the i-th attraction type within a scenic area. The weight matrix W i transpose, b i B is the bias term for the i-th attraction type in the scenic area. j W represents the user behavior feature matrix for the j-th attraction type. j This represents the weight matrix of user behavior features at the j-th attraction type within the scenic area. The weight matrix W j transpose, b j Let n be the bias term for the j-th attraction type in the scenic area, and n be the number of scenic area types.

4. The method for dynamically generating scenic area tourism routes based on multimodal AI according to claim 3, characterized in that, The method for generating the scenic spot value assessment index specifically includes the following steps: Based on user status data and scenic spot terrain data, a user physical load index is generated; the user status data includes user heart rate and user walking speed; the scenic spot terrain data includes scenic spot slope, scenic spot walking distance length and scenic spot road type. Based on the scenic area's environmental data, an environmental degradation factor is generated. A scenic spot value assessment model is established based on the probability of user interest in scenic spot type, user physical burden index and environmental degradation factor, and a scenic spot value assessment index is generated.

5. The method for dynamically generating scenic area tourism routes based on multimodal AI according to claim 4, characterized in that, The specific methods for generating the user's physical load index include: Through the formula: ； Generate user physical load index E k ; In the formula, H represents the user's heart rate. This represents the increase in a user's heart rate after viewing attraction k. H represents the estimated increase in heart rate after a user completes visit k. max V represents the user's maximum safe heart rate, V represents the user's current walking speed, and V0 represents the average walking speed of tourists of the same age at the attraction. , All are weighting coefficients, and .

6. The method for dynamically generating scenic area tourism routes based on multimodal AI according to claim 5, characterized in that, The specific methods for generating the heart rate growth estimate include: Through the formula: ； Generate heart rate growth estimate ; In the formula, This indicates the slope of the scenic spot, L. k This indicates the length of the tour route to the attractions. H is the indicator value for the road surface type in the scenic area. rec This indicates the user's heart rate recovery speed, H rest This represents the user's resting heart rate.

7. The method for dynamically generating scenic area tourism routes based on multimodal AI according to claim 4, characterized in that, The specific methods for generating the environmental degradation factor include: Through the formula: ； Generate environmental degradation factor ; In the formula, Q k t represents the normalized characteristic value of the real-time visitor density of attraction k. w,k T represents the normalized eigenvalue of the estimated waiting time for attraction k. k This represents the normalized characteristic value of the real-time temperature of scenic spot k.

8. The method for dynamically generating scenic area tourism routes based on multimodal AI according to claim 4, characterized in that, The specific expression of the scenic spot value assessment model is as follows: ； In the expression, A k This represents the scenic spot's value assessment index. R represents the environmental degradation factor. k This represents the inherent rating of attraction k, E. k This represents the user's physical burden index after browsing attraction k. denoted as , where represents the probability of user interest in the type of attraction k, and B is the user behavior feature matrix corresponding to attraction k.

9. The method for dynamically generating scenic area tourism routes based on multimodal AI according to claim 8, characterized in that, The specific methods for generating the attraction removal assessment value include: Historical visitor data is obtained, and a scenic spot popularity correction coefficient is generated based on the historical number of visitors to the scenic spot. The historical visitor data includes the historical number of visitors to the scenic spot and the historical total number of visitors to the scenic spot. Specifically: Through the formula: ； Generate attraction popularity correction coefficient ; In the formula, M k M0 represents the historical number of visitors to attraction k, and M0 represents the historical total number of visitors to the scenic area. Based on the attraction popularity correction coefficient and the attraction value assessment index, an attraction removal assessment value is generated; Specifically: Through the formula: ； Generate attraction removal assessment value ; In the formula, This represents the attraction popularity correction coefficient for attraction k.

10. A dynamic generation system for scenic area tourism routes based on multimodal AI, characterized in that: This system is used to execute the method for dynamically generating scenic area tourism routes based on multimodal AI as described in any one of claims 1-9, specifically including: The initial setup unit is used to set the user's initial travel route; The data acquisition unit is used to collect user status data, user behavior data, scenic spot terrain data, and scenic spot environmental data. The interest analysis unit is used to build an interest probability prediction model based on user behavior data and generate user interest probabilities for attraction types. The scenic spot value analysis unit is used to build a scenic spot value assessment model based on the user interest probability of scenic spot type, user status data, scenic spot terrain data, and scenic spot environment data, and generate a scenic spot value assessment index. The attraction removal analysis unit is used to generate attraction removal assessment values ​​for the initial tourist route based on the attraction value assessment index. The initial setting unit is used to remove attractions from the initial tourist route based on the attraction removal evaluation value in the initial tourist route, and generate a dynamic tourist route.

Images