Scenic spot security management system and method based on artificial intelligence

Abstract

The invention discloses a scenic spot security and protection management system and method based on artificial intelligence, and relates to the technical field of scenic spot security and protection, and the method comprises the steps: obtaining scenic spot multi-source data from multiple channels, dynamically adjusting the weights of Prophet and XGBoost, and outputting a scenic spot flow final prediction value # imgabs0 #; a binocular camera is used in the scenic spot to collect images in each area of the scenic spot in real time, the tourist density and the tourist average speed in the areas are analyzed, the congestion risk of each area in the scenic spot is analyzed in combination with the area terrain complexity and the final predicted value of the scenic spot flow, and a scenic spot gate regulation and control analysis instruction is sent outwards; and after a scenic region gate regulation and control analysis instruction is received, the scenic region gate opening frequency is determined based on the congestion risk Fx of each region in the scenic region and the scenic region flow final prediction value # imgabs1 #. The congestion risk of each area can be found in time, and the opening frequency of the gate is adjusted, so that the risk of safety accidents such as congestion and stampede is reduced.

Description

Technical Field

[0001] The present invention relates to the field of scenic area security technology, and in particular to an artificial intelligence-based scenic area security management system and method. Background Art

[0002] Current scenic area security management faces the following common challenges: Inadequate surveillance coverage and isolated equipment: Traditional security systems rely on fixed cameras and manual inspections, making it difficult to cover complex terrain (such as cliffs, water bodies, and dense forests). Furthermore, video surveillance, access control, and fire protection systems operate independently, with data interoperability, making it difficult to detect safety hazards in a timely manner. For example, when tourists encounter danger in blind spots, rescue efforts are often delayed due to information lags. Inefficient risk response: Abnormal events (such as tourist gatherings and out-of-bounds behavior) rely on manual judgment, resulting in a cumbersome emergency response process. The time from detection to response typically takes 5-10 minutes, making it difficult to cope with the rapid spread of emergencies such as stampedes and flash floods. Lack of dynamic quantification of tourist behavior and environmental risks: Existing systems often trigger alerts based on static rules (such as density thresholds) and fail to incorporate real-time environmental parameters (such as crowd speed, meteorological hazards, and terrain complexity). This results in a false alarm rate as high as 30%-40%. For example, the carrying capacity of the same area on sunny and rainy days can differ significantly, but traditional systems cannot dynamically adjust risk assessment weights. Insufficient privacy protection and data compliance: A large number of scenic spots rely on facial recognition technology, and the storage and use of biometric information do not strictly comply with regulations such as GDPR, posing a risk of privacy leakage.

[0003] A Chinese invention application, published as CN114567823A, discloses a scenic area security system based on video surveillance. The system includes an image acquisition module, a behavior recognition module, and an alarm module. The system triggers alarms by analyzing visitor behavior (such as climbing or falling). This invention utilizes deep learning algorithms to improve the accuracy of behavior recognition, but its limitations include: A single data dimension: Relying solely on video data, it lacks integration of environmental sensors (such as temperature, humidity, and geological displacement) and scheduled traffic flow forecasts, making it impossible to construct a global risk model. For example, landslide risk requires integration with geological sensor data, but this system struggles to provide early warnings through visual analysis alone. The response mechanism is rigid: Alarms require manual confirmation and resource dispatch, and lack automated emergency response mechanisms (such as automatic opening of evacuation routes or drone cruise guidance). Dynamic adaptability is insufficient: The system lacks time series prediction models (such as LSTM), making it unable to predict risk trends over the next 15-30 minutes and enabling only reactive responses to existing events.

[0004] In addition, some scenic spots have tried to use electronic fences and broadcasting systems for management, but they rely on fixed rules and have a low level of intelligence. For example, electronic fences have a high false alarm rate (such as animals triggering alarms) and cannot distinguish between tourists' intentions to cross the boundary (such as taking photos close to the area versus actually climbing); broadcast prompts are global broadcasts, which interfere with the tourist experience and lack targeted intervention capabilities (such as sending mobile phone warnings to tourists in specific areas).

[0005] These issues lead to frequent safety incidents in scenic areas (such as stampedes and falls) and high management costs. Therefore, there is an urgent need for a scenic area security management system that integrates multimodal perception, dynamic risk assessment, and intelligent linkage to achieve an upgrade from "passive response" to "active prevention and control."

[0006] To this end, the present invention provides a scenic area security management system and method based on artificial intelligence. Summary of the Invention

[0007] (1) Technical problems solved In view of the shortcomings of the existing technology, the present invention provides a scenic area security management system and method based on artificial intelligence. After receiving the scenic area gate control analysis instruction, the present invention is based on the congestion risk Fx of each area in the scenic area and the final predicted value of the scenic area flow. By determining the frequency of gate opening in scenic spots, congestion risks in various areas can be discovered in a timely manner and the frequency of gate opening can be adjusted. Security personnel can be allocated more reasonably to ensure that there are sufficient security forces to maintain order during peak hours, thereby reducing the risk of safety accidents such as congestion and stampedes, thereby solving the technical problems recorded in the background technology.

[0008] (2) Technical solution To achieve the above objectives, the present invention is implemented through the following technical solutions: a scenic area security management method based on artificial intelligence, comprising the following steps: Obtain multi-source data from scenic spots from multiple channels and generate basic traffic forecasts for the next 2 hours based on the Prophet model Then, the traffic characteristic dataset of the scenic spot at the current moment and the Prophet prediction residual are imported into the XGBoost correction model to obtain the corrected prediction value Based on the prediction error of the sliding window, the weights of Prophet and XGBoost are dynamically adjusted to output the final prediction value of the scenic area flow. ; Use binocular cameras inside the scenic area to collect real-time images of each area in the scenic area, analyze the tourist density and average speed of the area, and analyze the congestion risk of each area in the scenic area based on the complexity of the regional terrain and the final predicted value of the scenic area flow. Then send out the scenic area gate control analysis instructions; After receiving the gate control analysis instruction of the scenic area, based on the congestion risk Fx of each area in the scenic area and the final predicted value of the scenic area flow , determine the frequency of opening the scenic area gates.

[0009] Furthermore, through WeChat official accounts, mini-programs and other channels, the number of reservations and the entry verification rate are counted in real time. APIs such as Hefeng Weather and Caiyun Weather are called to obtain real-time weather data such as temperature, precipitation probability, and wind speed. Through the open platforms of Gaode Map and Baidu Map, the real-time congestion level and travel time of roads around the scenic area are obtained. It is connected with parking platforms such as ETCP and Jieshun to obtain the real-time vacancy rate of surrounding parking lots. The Scrapy framework is used to capture the video playback volume, likes, and comments under the platform's topic tags (such as #Scenic Area Name#). The holiday passenger flow of the scenic area in the past three years is obtained through the gate card swiping records and electronic ticketing data.

[0010] Furthermore, the holiday passenger flow of the scenic spot in the past three years is imported into the Prophet model to decompose the trend term, seasonal term, and holiday effect of the time series, and generate the basic flow forecast value for the next 2 hours. Then, the traffic characteristic dataset of the scenic spot at the current moment, the actual traffic volume of the gate in the past hour, and the Prophet prediction residual are imported into the XGBoost correction model to obtain the corrected prediction value .

[0011] Furthermore, the Prophet prediction residual is , It is the actual traffic volume of the scenic area.

[0012] Prophet is an open-source library developed by Facebook for time series forecasting. It is particularly well-suited for data with significant seasonality and trends. It automatically detects seasonal patterns in data, such as daily, weekly, monthly, or annual seasonal variations. It allows users to customize the impact of special events like holidays, which is particularly important for forecasting businesses heavily impacted by holidays, such as retail and tourism. It captures long-term trends in data and allows users to specify trend change points to accommodate sudden changes in the business. It is suitable for forecasting time series data with significant seasonality and trends, such as website visits, sales data, and electricity demand.

[0013] XGBoost (eXtreme Gradient Boosting) is an ensemble learning method that builds a strong learner by combining multiple weak learners (typically decision trees). A modified model typically involves using a model like XGBoost to further adjust and optimize the predictions of a base model. XGBoost excels at handling large datasets, with efficient computational performance and memory usage. Through techniques like regularization, column sampling, and row sampling, XGBoost effectively prevents overfitting and improves model generalization. It can handle complex nonlinear relationships and interactions, making it highly effective at capturing subtle patterns in data. As a modified model: When the predictions of a base model (such as Prophet) are biased, models like XGBoost can be used to correct them and improve prediction accuracy. XGBoost can also be used as a standalone prediction model, applicable to various regression and classification problems, particularly in scenarios with complex data features and nonlinear relationships.

[0014] Furthermore, the scenic spot traffic feature dataset includes the scenic spot’s holiday visitor flow in the past three years, online reservation platforms, OTA channel bookings, weather API, traffic congestion index, social media popularity, and remaining parking spaces in surrounding parking lots.

[0015] Furthermore, a sliding window of fixed size is set to calculate the recent prediction error. The window size is adjusted according to the actual situation. At each time point, the data in the sliding window is used to calculate the prediction error of the Prophet model and the XGBoost model. The weights of the two models are dynamically adjusted according to the error size, and the model with smaller error is given a higher weight.

[0016] Furthermore, the weights of Prophet and XGBoost are dynamically adjusted based on the sliding window prediction error (MAPE):

[0017] in, is the prediction error of the Prophet model, is the prediction error of the XGBoost model.

[0018] The weights are updated every hour, and the final predicted value of scenic area traffic is output based on the dynamically adjusted weights of Prophet and XGBoost. :

[0019] Furthermore, a binocular camera is used inside the scenic area to collect images in each area of the scenic area in real time. Tourist targets are detected based on the target detection algorithm, the tourist density in the area is calculated, and the tourist appearance feature vectors are extracted based on the multi-target tracking algorithm. The cosine similarity is calculated, and the average movement speed of tourists in the area is calculated by combining the depth information of the binocular camera.

[0020] Furthermore, the final predicted values of tourist density, average tourist speed, terrain complexity and scenic area flow in the area are obtained. , analyze the congestion risk Fx of each area within the scenic area:

[0021] Furthermore, it should be noted that when analyzing the congestion risk of each area within the scenic area, all data have been normalized.

[0022] Furthermore, when the congestion risk Fx of each area within the scenic area exceeds 0.6, the area is marked as a high-risk congestion area. If there is a high-risk congestion area within the scenic area, a scenic area gate control analysis instruction is sent out.

[0023] Furthermore, after receiving the gate control analysis instruction of the scenic area, based on the congestion risk Fx of each area within the scenic area and the final predicted value of the scenic area flow , determine the frequency of scenic area gate opening:

[0024] Set according to the carrying capacity of the scenic area.

[0025] An artificial intelligence-based scenic area security management system, comprising: The scenic area traffic forecast module obtains multi-source data from multiple channels and generates basic traffic forecast values for the next 2 hours based on the Prophet model. Then, the traffic characteristic dataset of the scenic spot at the current moment and the Prophet prediction residual are imported into the XGBoost correction model to obtain the corrected prediction value Based on the prediction error of the sliding window, the weights of Prophet and XGBoost are dynamically adjusted to output the final predicted value of the scenic area flow. ; The congestion analysis module within the scenic area uses a binocular camera to collect real-time images of each area in the scenic area, analyzes the tourist density and average speed of the area, and analyzes the congestion risk of each area within the scenic area based on the complexity of the regional terrain and the final predicted value of the scenic area flow. It then sends out the scenic area gate control analysis instructions; The scenic area gate control module receives the scenic area gate control analysis instruction, based on the congestion risk Fx of each area in the scenic area and the final predicted value of the scenic area flow , determine the frequency of opening the scenic area gates.

[0026] (3) Beneficial effects The present invention provides a scenic area security management system and method based on artificial intelligence, which has the following beneficial effects: 1. Obtain multi-source data from multiple channels and generate basic traffic forecasts for the next 2 hours based on the Prophet model Then, the traffic characteristic dataset of the scenic spot at the current moment and the Prophet prediction residual are imported into the XGBoost correction model to obtain the corrected prediction value Based on the prediction error of the sliding window, the weights of Prophet and XGBoost are dynamically adjusted to output the final predicted value of the scenic area flow. , automatically adjust the weights of the two models. This mechanism stabilizes the model MAPE below 5% in short-term predictions, which is far superior to a single model, achieving high precision and strong adaptability in scenic area traffic prediction.

[0027] 2. Use binocular cameras inside the scenic area to collect real-time images of each area in the scenic area, analyze the tourist density and average speed in the area, and analyze the congestion risk of each area in the scenic area based on the complexity of the regional terrain and the final predicted value of the scenic area flow. Send out scenic area gate control analysis instructions to the outside, and warn of congestion and accident risks in advance, achieve flow balance, and avoid congestion.

[0028] 3. After receiving the gate control analysis instruction of the scenic area, based on the congestion risk Fx of each area in the scenic area and the final predicted value of the scenic area flow By determining the frequency of gate opening in scenic spots, we can timely discover the congestion risks in various areas and adjust the frequency of gate opening. We can allocate security personnel more reasonably and ensure that there are sufficient security forces to maintain order during peak hours, thereby reducing the risk of safety accidents such as congestion and stampedes. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a flow chart of a scenic area security management method based on artificial intelligence according to the present invention; Figure 2 This is a structural diagram of an artificial intelligence-based scenic area security management system of the present invention. DETAILED DESCRIPTION

[0030] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0031] See also Figure 1 The present invention provides a scenic area security management method based on artificial intelligence, comprising the following steps: Step 1: Obtain multi-source data from multiple channels and generate a basic traffic forecast for the next 2 hours based on the Prophet model. Then, the traffic characteristic dataset of the scenic spot at the current moment and the Prophet prediction residual are imported into the XGBoost correction model to obtain the corrected prediction value Based on the prediction error of the sliding window, the weights of Prophet and XGBoost are dynamically adjusted to output the final predicted value of the scenic area flow. .

[0032] The step 1 includes the following: Step 101: Obtain multi-source data from multiple channels, unify the timestamps of different data sources, aggregate data at a 15-minute granularity, and construct a scenic spot traffic feature dataset, including holiday visitor flow, online reservation platforms, OTA channel bookings, weather APIs, traffic congestion indexes, social media popularity, and remaining parking spaces in surrounding parking lots over the past three years.

[0033] Through WeChat official accounts, mini-programs and other channels, the number of reservations and the entry verification rate are counted in real time. APIs such as Hefeng Weather and Caiyun Weather are called to obtain real-time weather data such as temperature, precipitation probability, and wind speed. Through the open platforms of Gaode Map and Baidu Map, the real-time congestion level and travel time of roads around the scenic area are obtained. It is connected with parking platforms such as ETCP and Jieshun to obtain the real-time vacancy rate of surrounding parking lots. The Scrapy framework is used to capture the video playback volume, likes, and comments under the platform's topic tags (such as #Scenic Area Name#). The holiday passenger flow of the scenic area over the past three years is obtained through gate card swiping records and electronic ticketing data.

[0034] Step 102: Import the holiday passenger flow of the scenic spot in the past three years into the Prophet model to decompose the trend term, seasonal term, and holiday effect of the time series, and generate the basic flow forecast value for the next 2 hours. Then, the traffic characteristic dataset of the scenic spot at the current moment, the actual traffic volume of the gate in the past hour, and the Prophet prediction residual are imported into the XGBoost correction model to obtain the corrected prediction value .

[0035] Among them, the Prophet prediction residual is , It is the actual traffic volume of the scenic area.

[0036] Prophet is an open-source library developed by Facebook for time series forecasting. It is particularly well-suited for data with significant seasonality and trends. It automatically detects seasonal patterns in data, such as daily, weekly, monthly, or annual seasonal variations. It allows users to customize the impact of special events like holidays, which is particularly important for forecasting businesses heavily impacted by holidays, such as retail and tourism. It captures long-term trends in data and allows users to specify trend change points to accommodate sudden changes in the business. It is suitable for forecasting time series data with significant seasonality and trends, such as website visits, sales data, and electricity demand.

[0037] XGBoost (eXtreme Gradient Boosting) is an ensemble learning method that builds a strong learner by combining multiple weak learners (typically decision trees). A modified model typically involves using a model like XGBoost to further adjust and optimize the predictions of a base model. XGBoost excels at handling large datasets, with efficient computational performance and memory usage. Through techniques like regularization, column sampling, and row sampling, XGBoost effectively prevents overfitting and improves model generalization. It can handle complex nonlinear relationships and interactions, making it highly effective at capturing subtle patterns in data. As a modified model: When the predictions of a base model (such as Prophet) are biased, models like XGBoost can be used to correct them and improve prediction accuracy. XGBoost can also be used as a standalone prediction model, applicable to various regression and classification problems, particularly in scenarios with complex data features and nonlinear relationships.

[0038] Step 103: Set a fixed-size sliding window for calculating the recent forecast error. The window size can be adjusted based on actual conditions to balance the use of historical information with the capture of recent changes. For example, the sliding window can be selected to cover the last 7, 14, or 30 days of data.

[0039] At each time point, the prediction errors of the Prophet and XGBoost models are calculated using the data within the sliding window. The weights of the two models are dynamically adjusted based on the size of the error, with the model with the smaller error being given a higher weight. Weight calculations can be performed using methods such as the inverse error method and the least squares method. For example, the inverse error method can calculate weights based on the model's prediction error, with models with smaller errors receiving higher weights.

[0040] Dynamically adjust the weights of Prophet and XGBoost based on the sliding window prediction error (MAPE):

[0041] in, is the prediction error of the Prophet model, is the prediction error of the XGBoost model.

[0042] The weights are updated every hour, and the final predicted value of scenic area traffic is output based on the dynamically adjusted weights of Prophet and XGBoost. :

[0043] When using, combine the contents in steps 101 to 103: Obtain multi-source data from scenic spots from multiple channels and generate basic traffic forecasts for the next 2 hours based on the Prophet model Then, the traffic characteristic dataset of the scenic spot at the current moment and the Prophet prediction residual are imported into the XGBoost correction model to obtain the corrected prediction value Based on the prediction error of the sliding window, the weights of Prophet and XGBoost are dynamically adjusted to output the final prediction value of the scenic area flow. , automatically adjust the weights of the two models. This mechanism stabilizes the model MAPE below 5% in short-term predictions, which is far superior to a single model, achieving high precision and strong adaptability in scenic area traffic prediction.

[0044] Step 2: Use a binocular camera inside the scenic area to collect real-time images of each area in the scenic area, analyze the tourist density and average speed of the area, combine the complexity of the regional terrain and the final predicted value of the scenic area flow to analyze the congestion risk of each area in the scenic area, and send out scenic area gate control analysis instructions.

[0045] The second step includes the following: Step 201: Use a binocular camera to capture real-time images of each area within the scenic area. Use a target detection algorithm to detect tourist targets and calculate the tourist density within the area. Use a multi-target tracking algorithm to extract tourist appearance feature vectors, calculate cosine similarity, and combine the binocular camera depth information to calculate the average tourist movement speed within the area. Pre-label the area types on the GIS map to obtain terrain complexity, such as assigning a weight of 1.8 to plank roads and 1.0 to squares.

[0046] Step 202: Obtain the final predicted values of tourist density, average tourist speed, terrain complexity and scenic area flow in the area , analyze the congestion risk Fx of each area within the scenic area:

[0047] It should be noted that when analyzing the congestion risk of each area within the scenic area, all data have been normalized.

[0048] Step 203: When the congestion risk Fx of each area in the scenic area exceeds 0.6, the area is marked as a high-risk congestion area. If a high-risk congestion area exists in the scenic area, a scenic area gate control analysis instruction is sent out.

[0049] When using, combine the contents in steps 201 to 203: A binocular camera is used inside the scenic area to collect real-time images of each area in the scenic area, analyze the tourist density and average speed in the area, and analyze the congestion risk of each area in the scenic area based on the complexity of the regional terrain and the final predicted value of the scenic area flow. The scenic area gate control analysis instructions are sent out to the outside to provide early warning of congestion and accident risks, achieve flow balance, and avoid congestion.

[0050] Step 3: After receiving the gate control analysis instruction of the scenic area, based on the congestion risk Fx of each area in the scenic area and the final predicted value of the scenic area flow , determine the frequency of opening the scenic area gates.

[0051] The step three includes the following: Step 301: After receiving the gate control analysis instruction of the scenic area, based on the congestion risk Fx of each area in the scenic area and the final predicted value of the scenic area flow , determine the frequency of scenic area gate opening:

[0052] Set according to the carrying capacity of the scenic area.

[0053] When used, combine the content of step 301: After receiving the scenic area gate control analysis command, based on the current congestion risk Fx of each area in the scenic area and the final predicted value of the scenic area flow By determining the frequency of gate opening in scenic spots, we can timely discover the congestion risks in various areas and adjust the frequency of gate opening. We can allocate security personnel more reasonably and ensure that there are sufficient security forces to maintain order during peak hours, thereby reducing the risk of safety accidents such as congestion and stampedes.

[0054] See also Figure 2 The present invention provides a scenic area security management system based on artificial intelligence, comprising: The scenic area traffic forecast module obtains multi-source data from multiple channels and generates basic traffic forecast values for the next 2 hours based on the Prophet model. Then, the traffic characteristic dataset of the scenic spot at the current moment and the Prophet prediction residual are imported into the XGBoost correction model to obtain the corrected prediction value Based on the prediction error of the sliding window, the weights of Prophet and XGBoost are dynamically adjusted to output the final predicted value of the scenic area flow. .

[0055] The congestion analysis module within the scenic area uses a binocular camera to collect real-time images of each area in the scenic area, analyzes the tourist density and average speed of the area, combines the complexity of the regional terrain and the final predicted value of the scenic area flow to analyze the congestion risk of each area within the scenic area, and sends out scenic area gate control analysis instructions.

[0056] The scenic area gate control module receives the scenic area gate control analysis instruction, based on the congestion risk Fx of each area in the scenic area and the final predicted value of the scenic area flow , determine the frequency of opening the scenic area gates.

[0057] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0058] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0059] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A scenic area security management method based on artificial intelligence, characterized by: The steps include: Obtain multi-source data from scenic spots from multiple channels and generate basic traffic forecasts for the next 2 hours based on the Prophet model Then, the traffic characteristic dataset of the scenic spot at the current moment and the Prophet prediction residual are imported into the XGBoost correction model to obtain the corrected prediction value Based on the prediction error of the sliding window, the weights of Prophet and XGBoost are dynamically adjusted to output the final prediction value of the scenic area flow. ; Use binocular cameras inside the scenic area to collect real-time images of each area in the scenic area, analyze the tourist density and average speed of the area, and analyze the congestion risk of each area in the scenic area based on the complexity of the regional terrain and the final predicted value of the scenic area flow. Then send out the scenic area gate control analysis instructions; After receiving the gate control analysis instruction of the scenic area, based on the congestion risk Fx of each area in the scenic area and the final predicted value of the scenic area flow , determine the frequency of opening the scenic area gates.

2. The artificial intelligence-based scenic area security management method according to claim 1, characterized in that: Import the holiday passenger flow of the scenic spot in the past three years into the Prophet model to decompose the trend term, seasonal term, and holiday effect of the time series, and generate the basic flow forecast value for the next 2 hours Then, the traffic characteristic dataset of the scenic spot at the current moment, the actual traffic volume of the gate in the past hour, and the Prophet prediction residual are imported into the XGBoost correction model to obtain the corrected prediction value .

3. The artificial intelligence-based scenic area security management method according to claim 1, characterized in that: The scenic spot traffic feature dataset includes holiday visitor flow in the past three years, online reservation platforms, OTA channel bookings, weather API, traffic congestion index, social media popularity, and remaining parking spaces in surrounding parking lots.

4. The artificial intelligence-based scenic area security management method according to claim 1, characterized in that: A fixed-size sliding window is set to calculate the recent prediction error. The window size is adjusted according to the actual situation. At each time point, the data in the sliding window is used to calculate the prediction error of the Prophet model and the XGBoost model. The weights of the two models are dynamically adjusted according to the error size, and the model with smaller error is given a higher weight.

5. The scenic area security management method based on artificial intelligence according to claim 1 is characterized in that: A binocular camera is used inside the scenic area to collect images in each area of the scenic area in real time. Tourist targets are detected based on the target detection algorithm, the tourist density in the area is calculated, and the tourist appearance feature vectors are extracted based on the multi-target tracking algorithm. The cosine similarity is calculated, and the average movement speed of tourists in the area is calculated by combining the depth information of the binocular camera.

6. The artificial intelligence-based scenic area security management method according to claim 1, characterized in that: When analyzing the congestion risk of each area within the scenic area, all data have been normalized.

7. The artificial intelligence-based scenic area security management method according to claim 1, characterized in that: When the congestion risk Fx of each area within the scenic area exceeds 0.6, the area is marked as a high-risk congestion area. If there is a high-risk congestion area within the scenic area, a scenic area gate control analysis instruction is sent out.

8. An artificial intelligence-based scenic area security management system, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The scenic area traffic forecast module obtains multi-source data from multiple channels and generates basic traffic forecast values for the next 2 hours based on the Prophet model. Then, the traffic characteristic dataset of the scenic spot at the current moment and the Prophet prediction residual are imported into the XGBoost correction model to obtain the corrected prediction value Based on the prediction error of the sliding window, the weights of Prophet and XGBoost are dynamically adjusted to output the final prediction value of the scenic area flow. ; The congestion analysis module within the scenic area uses a binocular camera to collect real-time images of each area in the scenic area, analyzes the tourist density and average speed of the area, and analyzes the congestion risk of each area within the scenic area based on the complexity of the regional terrain and the final predicted value of the scenic area flow. It then sends out the scenic area gate control analysis instructions; The scenic area gate control module receives the scenic area gate control analysis instruction, based on the congestion risk Fx of each area in the scenic area and the final predicted value of the scenic area flow , determine the frequency of opening the scenic area gates.

Images