Optimized scheduling and safety early warning intelligent aid decision-making method based on tourist attraction people flow prediction

Abstract

The invention discloses an optimal scheduling and safety early warning intelligent aid decision-making method based on tourist attraction people flow prediction, and the method employs an advanced machine learning model to carry out the people flow prediction through integrating historical tourist data, holiday and festival information and other data sources. Based on the prediction result, the dynamic resource optimization scheduling is realized, such as reasonable arrangement of workers, facility opening and traffic guidance, so that the tourist experience is improved. Meanwhile, a safety early warning mechanism is established, and when the predicted visitor flow rate exceeds a set threshold value, management personnel can receive a monitoring result in time, issue early warning information and start an emergency plan, so that the safety of tourists is ensured. The invention aims to improve the management level and service quality of the tourist attractions through an intelligent means, and is suitable for management and operation of various tourist attractions.

Description

Technical Field

[0001] The present invention belongs to the technical field of crowd flow prediction, and in particular relates to an optimization scheduling and safety warning intelligent auxiliary decision-making method based on crowd flow prediction of tourist attractions. Background Art

[0002] With the rapid development of intelligent technology, especially driven by artificial intelligence, Internet of Things, big data, cloud computing and other technologies, the intelligent management of tourist attractions is gradually becoming an important means to improve the operational efficiency of scenic spots and the experience of tourists. As one of the core technologies in intelligent management systems, crowd flow prediction has been widely used in many fields such as scenic spot resource scheduling, passenger flow control, and safety warning; accurate crowd flow prediction can effectively prevent congestion and accidents in scenic spots, and at the same time provide a scientific basis for the opening hours of scenic spots, the scheduling of scenic spot resources, and the formulation of emergency plans.

[0003] Traditional methods of crowd flow prediction are mostly based on time series models, especially autoregressive models (AR) and long short-term memory networks (LSTM). These models predict future crowd flow by analyzing historical data. However, traditional time series prediction models are usually difficult to achieve ideal results when faced with complex crowd flow prediction problems at tourist attractions that are affected by multiple factors. The crowd flow at tourist attractions is affected by many factors. In addition to the periodic fluctuations in historical data, it is also closely related to factors such as holidays. Simply relying on historical data for prediction is difficult to accurately capture the impact of these non-temporal factors on crowd flow.

[0004] In recent years, with the rapid development of deep learning technology, prediction methods based on deep learning have shown powerful capabilities in complex time series prediction tasks. The Transformer model has been widely used in natural language processing, image processing and other fields due to its unique self-attention mechanism and the advantage of processing long sequences, and has gradually been introduced into the field of pedestrian flow prediction. Unlike traditional RNN (recurrent neural network), CNN (convolutional neural network) or LSTM models, Transformer can effectively capture long-distance dependencies in time series data through the global self-attention mechanism, thereby improving prediction accuracy. In addition, Transformer can process data of all time steps in parallel, avoiding the problem of gradient disappearance or explosion in the RNN model.

[0005] For example, the Chinese patent application with publication number CN116955953A proposed a Transformer-based bus station passenger flow prediction method. First, passenger flow data was collected at the BRT bus station, and a data set suitable for autonomous driving application scenarios was established through data processing. Then a TransFormer prediction model was constructed, which used the observation sequence as input features, adopted the encoder-decoder framework and introduced the self-attention mechanism to mine the key factors affecting future passenger flow. However, this method failed to fully consider the spatial features in the scene video data, which made it difficult to capture real-time dynamic changes and cope with fluctuations in passenger flow during special time periods.

[0006] For example, the literature [W. Zhang, C. Zhang and F. Tsung, "Transformer Based Spatial-Temporal Fusion Network for Metro Passenger Flow Forecasting," 2021 IEEE 17th International Conference on Automation Science and Engineering (CASE), Lyon, France, 2021, pp. 1515-1520] proposed a TSTFN (Transformer Based Spatial-Temporal Fusion Network) model for subway passenger flow forecasting. TSTFN contains a novel spatial-temporal synchronized self-attention layer that can simultaneously model spatial and temporal correlations. However, TSTFN mainly focuses on the spatial dependencies within the subway system, and pays less attention to external factors, which may lead to a decrease in prediction accuracy during special time periods.

[0007] From the above analysis, we can see that although the existing methods have been able to solve these problems to a certain extent, there are still some challenges. For example, how to efficiently integrate spatial information with time series information and how to fully explore the correlation between different data sources are still the focus and difficulty of current research. Summary of the invention

[0008] In view of the above, the present invention provides an intelligent auxiliary decision-making method for optimized scheduling and safety warning based on crowd flow prediction at tourist attractions, which can simultaneously process video data, time series data and the influence of external factors, further improve the accuracy of crowd flow prediction and the generalization ability of the model, and provide strong technical support for the intelligent management of scenic spots.

[0009] An optimization scheduling and safety warning intelligent auxiliary decision-making method based on tourist attraction crowd flow prediction includes the following steps:

[0010] (1) Collect historical traffic data, scenic spot video data, and holiday information and pre-process these data;

[0011] (2) Constructing a passenger flow prediction model, which includes:

[0012] The Trifusion coding module encodes historical traffic data, holiday information, and side information of scenic spot video data (including camera position and viewing angle) into fixed-dimensional vectors through the embedding layer, and then fuses the vectors obtained by encoding these three types of data information into Trifusion coding;

[0013] The pre-trained CNN network is used to extract spatial features including the number and density of people flow in scenic spot video data;

[0014] LSTM network is used to further extract spatial features to capture the dynamic changes and trend fluctuations of tourist flow and obtain spatiotemporal features containing tourist flow information;

[0015] Transformer module, predicting the flow of people in the scenic spot in each time period according to the spatiotemporal characteristics and Trifusion coding;

[0016] (3) using the data information preprocessed in step (1) to train the above-mentioned pedestrian flow prediction model;

[0017] (4) By inputting the real-time video data of the scenic spot into the trained crowd flow prediction model, the crowd flow data of the scenic spot in the current time period can be predicted and output, and then the scenic spot can be given a safety warning and crowd flow dispatch based on the crowd flow data predicted by the model.

[0018] Furthermore, the historical passenger flow data in step (1) includes the number of tourists in each period of each day during the same period in the past few years and the number of tourists on holidays. The scenic area video data collects the tourist flow inside and outside the scenic area in real time through multiple cameras, and the holiday information integrates the information of national statutory holidays and school winter and summer vacations.

[0019] Furthermore, the preprocessing of data information in step (1) includes data synchronization, format standardization, video frame denoising and blurry frame removal, wherein data synchronization is to process missing or delayed data through time alignment algorithm and interpolation method, to ensure that each frame of video data and its side information, historical traffic data and holiday information have a timestamp, and to ensure the timing of the data stream; format standardization is to scale each frame of video data to a fixed size according to the required unified resolution, to ensure the standardization of video input, and to avoid resolution differences caused by different cameras; video frame denoising is to use Gaussian filtering algorithm to remove static background noise in video data, and the quality of each frame of video data is evaluated by calculating its standard deviation and contrast, and low-quality frames are automatically removed; blurry frame removal is to automatically detect and remove blurry frames through image clarity evaluation methods such as Laplace transform, to ensure that each frame of video data input to the model is clear and valid.

[0020] Furthermore, the CNN network adopts the ResNet50 deep convolutional neural network, which is pre-trained on the dataset ImageNet, extracts high-level features of the image through forward propagation, and then uses the pre-trained network for transfer learning. The first several layers at the front of the network are frozen to retain their common features, and then only the layers at the back of the network are trained to adapt to the characteristics of the scenic spot video data.

[0021] Furthermore, the process of training the crowd flow prediction model in step (3) is as follows:

[0022] 3.1 Initialize model parameters, including the bias vector and weight matrix of each layer, learning rate and optimizer;

[0023] 3.2 Input historical passenger flow data, holiday information, scenic spot video data and side information into the model, and the model forward propagates the output to obtain the corresponding prediction result, i.e., passenger flow data, and calculate the loss function between the prediction result and the label (i.e., historical passenger flow data);

[0024] 3.3 According to the loss function, the optimizer is used to iteratively update the model parameters through the gradient descent method until the loss function converges and the training is completed.

[0025] Furthermore, in the step (4), safety warning and crowd flow dispatching are carried out for the scenic spot based on the crowd flow data predicted by the model, specifically including: presetting the tourist carrying capacity threshold and warning rules of different risk levels in the scenic spot; receiving and analyzing the output results of the crowd flow prediction model in real time, dynamically monitoring the crowd flow prediction data of the scenic spot in each time period, and judging whether there are safety hazards in the scenic spot in combination with the tourist carrying capacity threshold in the scenic spot; when the predicted crowd flow approaches or exceeds the tourist carrying capacity threshold, automatically triggering the safety warning mechanism, and prompting the warning information to the relevant management personnel of the scenic spot; recording the key data in the safety dispatching process and feeding it back to the system for optimizing the subsequent dispatching rules and the crowd flow prediction model; performing data analysis on safety dispatching and emergency management, summarizing the dispatching results and optimizing the future dispatching rules and models, and further optimizing the tourist carrying capacity threshold and warning rules through data analysis, so as to improve the system's emergency response speed and the level of tourist safety protection.

[0026] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor is used to execute the computer program to implement the above-mentioned optimized scheduling and safety warning intelligent auxiliary decision-making method based on tourist attraction crowd flow prediction.

[0027] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned optimized scheduling and safety warning intelligent auxiliary decision-making method based on tourist attraction crowd flow prediction.

[0028] The present invention proposes a method based on CNN+LSTM+Transformer combined with multimodal information TriFusion encoding, which effectively captures the correlation between these multi-dimensional information by encoding historical tourist flow data, holiday information and side information, and provides strong support for accurate crowd flow prediction of tourist attractions. The main technical features of the present invention are as follows:

[0029] Multimodal data fusion: The present invention jointly encodes historical tourist flow, holiday information and side information, so that the model can fully consider the impact of multiple factors on the flow of people, and not just rely on a single time series or spatial feature, thereby improving the generalization ability of the model.

[0030] CNN+LSTM+Transformer architecture: This paper uses CNN to extract spatial features in video frames, LSTM to process the temporal features of historical tourist traffic, and Transformer to perform global temporal modeling and cross-modal feature fusion, which can effectively capture the complex relationship between multiple factors; this architecture effectively solves the limitations of a single method and provides a full range of modeling capabilities for spatiotemporal information.

[0031] Effective integration of holiday information and side information: Through position encoding and embedding layers, holiday information, camera perspective and other auxiliary information are effectively integrated, so that the model can more accurately simulate the complex changes in the flow of people in actual scenarios; for example, the flow of tourists on holidays and weekends usually fluctuates significantly, and traditional methods are difficult to effectively capture these influences. However, the present invention uses TriFusion encoding to make full use of these influencing factors.

[0032] Accurate crowd flow prediction: By combining historical tourist flow and external factors, the present invention can achieve accurate crowd flow prediction for tourist attractions and provide real-time decision support. This technology can not only provide early warning information about tourist flow for scenic area management decisions, but also help optimize scenic area resource scheduling, improve tourist experience, reduce congestion and enhance safety.

[0033] Based on the above technical features, the present invention has the following beneficial technical effects compared with the prior art:

[0034] 1. Improved accuracy: The present invention can effectively improve the accuracy of pedestrian flow prediction by integrating multiple information sources, especially combining visual data and historical traffic data. Traditional time series models and methods based on historical data cannot accurately capture the interaction of multi-dimensional information when facing complex scenarios, while the present invention significantly improves the prediction effect through multimodal feature fusion.

[0035] 2. Powerful time series modeling capabilities: This paper uses the self-attention mechanism of the Transformer model to process data with longer time spans and capture complex time dependencies. Compared with traditional LSTM or RNN, Transformer can effectively avoid gradient vanishing and long-distance dependency problems, and enhance the stability and accuracy of prediction.

[0036] 3. Flexibility and scalability: The TriFusion encoding method adopted in the present invention has good flexibility and scalability, and can integrate auxiliary information to support customized predictions in different scenarios.

[0037] 4. Optimize scenic area management and scheduling: The technology of the present invention can provide real-time decision support for scenic area management, help managers estimate changes in tourist flow, optimize scenic area resource allocation, and conduct effective safety warnings and risk predictions; especially during holidays and peak periods, it can predict fluctuations in passenger flow in advance, help scenic spots take preventive measures, and ensure the safety and experience of tourists. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a schematic diagram of the network structure of the pedestrian flow prediction model of the present invention.

[0039] Figure 2Schematic diagram of the data preprocessing process of the present invention.

[0040] Figure 3 Schematic diagram of the data input process of the pedestrian flow prediction model of the present invention.

[0041] Figure 4 It is a schematic diagram of the process of providing safety warning for scenic spots according to the prediction results of the model in the present invention. DETAILED DESCRIPTION

[0042] In order to describe the present invention more specifically, the technical solution of the present invention is described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0043] This implementation provides a network model based on CNN, LSTM, and Transformer, while integrating historical tourist flow data, holiday information, and side information, so that the model can make a more accurate prediction of the flow of people in the scenic area based on some external auxiliary information.

[0044] like Figure 2 As shown, the present invention first pre-processes the input data, mainly including video data, historical tourist flow data, holiday information and side information.

[0045] The video data comes from the video stream transmitted by the scenic spot camera. After each frame of video is collected, it needs to be preprocessed to remove noise and outliers to ensure the validity of the data. After noise filtering and anomaly detection, the video data is further normalized to adjust the resolution of the video frame to adapt it to the input requirements of the subsequent model. 70% of the preprocessed data is divided into training sets and 30% into test sets to train the Transformer model.

[0046] like Figure 1 As shown, after the video frame is preprocessed, the spatial features such as the number and density of pedestrians are extracted through the pre-trained CNN network model. This embodiment uses ResNet50 as the baseline model of the CNN layer. ResNet is a deep convolutional neural network that has strong feature extraction capabilities when processing complex visual tasks and can effectively identify spatial information such as pedestrians and scene changes in video frames.

[0047] For the CNN layer, ResNet50 is first pre-trained on the large-scale image classification dataset ImageNet, and the high-level features of the image are extracted through forward propagation. Then, the pre-trained model is used for transfer learning. The first few layers of the network are frozen (retaining their common features), and then only the latter few layers of the network are trained to adapt to the image features in the present invention, such as pedestrians in video frames; in this way, the common features learned by the pre-trained model on the ImageNet dataset can be fully utilized, and the model can be adapted to the specific task of crowd flow prediction through fine-tuning.

[0048] These spatial features are then used as input to the LSTM layer. LSTM can capture the temporal information in real-time data, thereby not only extracting features in a single space or time for the video. Since LSTM can effectively capture the temporal features in the time series, the Transformer input no longer requires additional position information encoding. The output of LSTM already contains enough temporal information, allowing the Transformer to directly process these features and perform global temporal modeling.

[0049] like Figure 3 As shown in the figure, the historical tourist flow data is encoded through the embedding layer and converted into a vector of fixed dimension; the holiday information (such as whether it is a holiday, the type of holiday, etc.) uses the sine and cosine functions to convert the date information into periodic features to help the model capture the periodic pattern of the date; the side information (such as camera position, viewing angle, etc.) is also encoded and converted into a vector through the learnable embedding layer. Finally, the three encoding vectors of side information holiday information and historical tourist flow data are added together to form a Trifusion encoding and standardized to provide additional external context information for subsequent models.

[0050] The Trifusion code and real-time tourist flow data are input into the Transformer to further improve the model's understanding ability; the final output of the Transformer is a prediction result representing the flow of people, and the output layer maps the output of the Transformer to the final prediction value through a fully connected layer, specifically the predicted flow range in each time period. This implementation uses the mean square error (MSE) or weighted loss function to train the model, and optimizes the model parameters by minimizing the loss function, so that the output prediction result is as close to the actual tourist flow as possible.

[0051] The prediction output is a range, which represents the flow of tourists in the scenic area within a period of time. Through training and optimization, the model can accurately predict the changes in tourist flow in the scenic area and provide decision support for resource scheduling, passenger flow control, etc. Figure 4As shown, according to the prediction results of the scenic area flow, according to the pre-set safety warning rules, the scenic area personnel are dispatched in time to ensure the safety of tourists and reduce the potential risks of the scenic area.

[0052] According to the prediction results and preset thresholds, the staff will optimize the tourist flow, including reminding tourists to divert and dispatching personnel, etc., specifically including the following aspects:

[0053] (1) Monitoring of prediction results: The system receives and analyzes the output results of the crowd flow prediction module in real time, dynamically monitors the crowd flow prediction data of the scenic area at different time periods, and determines whether there are safety hazards based on the tourist carrying capacity threshold in the scenic area.

[0054] (2) Setting of safety warning rules: The system presets the threshold of the number of tourists in the scenic area and the warning rules of different risk levels, including:

[0055] 2.1 Safe tourist volume threshold: the maximum number of tourists allowed in different areas of the scenic area, such as setting different thresholds for the tourist capacity of the core area and the secondary area, or setting a maximum passenger flow for the entire scenic area;

[0056] 2.2 Warning level: The warning level is set according to the predicted changes in tourist flow, including normal, mild, severe and extremely severe risk levels, and the corresponding response measures are different.

[0057] (3) Warning information release: When the predicted tourist flow approaches or exceeds the safe tourist volume threshold, the system automatically triggers the safety warning mechanism and prompts the relevant scenic area managers with warning information, including:

[0058] 3.1 Scenic area managers may take preventive measures based on monitoring results, such as increasing security personnel and setting up flow control facilities in key areas;

[0059] 3.2 Management personnel may issue reminders to tourists about excessive traffic by updating billboards, using broadcasting systems or mobile applications to notify tourists and suggest them to avoid high traffic areas during off-peak hours.

[0060] (4) Dispatch feedback mechanism: The system records key data during the safety dispatch process and feeds it back to the safety warning module for optimizing subsequent dispatch rules and passenger flow prediction models, including:

[0061] 4.1 Execution time and effect evaluation of scheduling measures;

[0062] 4.2 Real-time tourist flow in different areas and the response time of corresponding control measures;

[0063] 4.3 Statistical data of visitor feedback and emergency response are used to adjust warning thresholds and emergency strategies.

[0064] (5) Analysis and optimization of dispatch results: The system conducts data analysis on safety dispatch and emergency management, summarizes dispatch results and optimizes future dispatch rules and models. Through data analysis, it further optimizes the tourist carrying capacity threshold and early warning rules, thereby improving the system's emergency response speed and tourist safety level.

[0065] In summary, the present invention first uses CNN to extract features from visual information in video frames, especially to extract spatial features such as human flow, scene density, and pedestrian distribution from camera video streams; then uses LSTM to perform time series modeling on historical tourist flow data to capture dynamic changes and trend fluctuations in human flow. In addition, the present invention also encodes multi-dimensional features of information such as historical tourist flow, holiday information, and camera viewing angles, and fuses them into TriFusion encoding to enhance the multimodal learning ability of the model and improve the accuracy and robustness of predictions.

[0066] In summary, the present invention successfully solves the problem of insufficient modeling of a single data source in the prior art by combining CNN+LSTM+Transformer with TriFusion encoding, and can efficiently integrate spatiotemporal information and external factors to provide a more accurate prediction of the flow of people at tourist attractions; this technology has broad application prospects and practical value in the fields of intelligent management of scenic spots, resource scheduling, safety warning, etc.

[0067] The above description of the embodiments is to facilitate the understanding and application of the present invention by those skilled in the art. It is obvious that those skilled in the art can easily make various modifications to the above embodiments and apply the general principles described herein to other embodiments without creative work. Therefore, the present invention is not limited to the above embodiments. Improvements and modifications made by those skilled in the art to the present invention based on the disclosure of the present invention should be within the protection scope of the present invention.

Claims

1. An optimization scheduling and safety warning intelligent auxiliary decision-making method based on tourist attraction crowd flow prediction, comprising the following steps: (1) Collect historical traffic data, scenic spot video data, and holiday information and pre-process these data; (2) Constructing a passenger flow prediction model, which includes: The Trifusion coding module encodes the historical traffic data, holiday information, and side information of scenic spot video data into vectors of fixed dimensions through the embedding layer, and then fuses the vectors obtained by encoding these three types of data information into Trifusion coding; The pre-trained CNN network is used to extract spatial features including the number and density of people flow in scenic spot video data; LSTM network is used to further extract spatial features to capture the dynamic changes and trend fluctuations of tourist flow and obtain spatiotemporal features containing tourist flow information; Transformer module, predicting the flow of people in the scenic spot in each time period according to the spatiotemporal characteristics and Trifusion coding; (3) using the data information preprocessed in step (1) to train the above-mentioned pedestrian flow prediction model; (4) By inputting the real-time video data of the scenic spot into the trained crowd flow prediction model, the crowd flow data of the scenic spot in the current time period can be predicted and output, and then the scenic spot can be given a safety warning and crowd flow dispatch based on the crowd flow data predicted by the model.

2. The method for optimizing scheduling and intelligent auxiliary decision-making for safety warning based on crowd flow prediction of tourist attractions according to claim 1 is characterized by: The historical passenger flow data in step (1) includes the number of tourists in each period of each day during the same period in the past few years and the number of tourists on holidays. The scenic area video data collects the tourist flow inside and outside the scenic area in real time through multiple cameras. The holiday information integrates the information of national statutory holidays and school winter and summer vacations.

3. The method for optimizing scheduling and intelligent auxiliary decision-making for safety warning based on crowd flow prediction of tourist attractions according to claim 1 is characterized by: The preprocessing of data information in step (1) includes data synchronization, format standardization, video frame denoising and blur frame removal, wherein data synchronization is to process missing or delayed data through time alignment algorithm and interpolation method, to ensure that each frame of video data and its side information, historical traffic data and holiday information have a timestamp, and to ensure the timing of the data stream; format standardization is to scale each frame of video data to a fixed size according to the required unified resolution, to ensure the standardization of video input, and to avoid resolution differences caused by different cameras; video frame denoising is to use Gaussian filtering algorithm to remove static background noise in video data, and the quality of each frame of video data is evaluated by calculating its standard deviation and contrast, and low-quality frames are automatically removed; blur frame removal is to automatically detect and remove blurry frames through image clarity evaluation methods such as Laplace transform, to ensure that each frame of video data input to the model is clear and valid.

4. The method for optimizing scheduling and intelligent auxiliary decision-making for safety warning based on crowd flow prediction of tourist attractions according to claim 1 is characterized by: The CNN network uses the ResNet50 deep convolutional neural network, which is pre-trained on the ImageNet dataset. The high-level features of the image are extracted through forward propagation, and then the pre-trained network is used for transfer learning. The front several layers of the network are first frozen to retain their common features. Then only the later layers of the network are trained to adapt to the characteristics of scenic spot video data.

5. The method for optimizing scheduling and intelligent auxiliary decision-making for safety warning based on crowd flow prediction of tourist attractions according to claim 1 is characterized by: The process of training the crowd flow prediction model in step (3) is as follows: 3.1 Initialize model parameters, including the bias vector and weight matrix of each layer, learning rate and optimizer; 3.2 Input historical passenger flow data, holiday information, scenic spot video data and side information into the model, forward propagate the model output to obtain the corresponding prediction result, i.e., passenger flow data, and calculate the loss function between the prediction result and the label; 3.3 According to the loss function, the optimizer is used to iteratively update the model parameters through the gradient descent method until the loss function converges and the training is completed.

6. The method for optimizing scheduling and intelligent auxiliary decision-making for safety warning based on crowd flow prediction of tourist attractions according to claim 1 is characterized by: In the step (4), the scenic area is provided with a safety warning and crowd flow dispatch according to the crowd flow data predicted by the model, specifically including: presetting the tourist carrying capacity threshold and warning rules of different risk levels in the scenic area; receiving and analyzing the output results of the crowd flow prediction model in real time, dynamically monitoring the crowd flow prediction data of the scenic area in each period, and judging whether there are safety hazards in the scenic area in combination with the tourist carrying capacity threshold in the scenic area; when the predicted crowd flow approaches or exceeds the tourist carrying capacity threshold, automatically triggering the safety warning mechanism, and prompting the warning information to the relevant management personnel of the scenic area; recording the key data in the safety dispatch process and feeding it back to the system for optimizing the subsequent dispatch rules and the crowd flow prediction model; performing data analysis on safety dispatch and emergency management, summarizing the dispatch results and optimizing the future dispatch rules and models, and further optimizing the tourist carrying capacity threshold and warning rules through data analysis, so as to improve the system's emergency response speed and the level of tourist safety protection.

7. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: The processor is used to execute the computer program to implement the optimization scheduling and safety warning intelligent auxiliary decision-making method based on tourist attraction crowd flow prediction as described in any one of claims 1 to 6.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, it realizes the optimization scheduling and safety warning intelligent auxiliary decision-making method based on tourist attraction passenger flow prediction as described in any one of claims 1 to 6.

Images