Scenic spot people flow density detection and early warning method and system

Abstract

The invention is suitable for the technical field of people flow density detection, and provides a scenic spot people flow density detection and early warning method and system, and the method comprises the steps: collecting real-time passenger flow data through an intelligent device disposed at a key node of a scenic spot; preprocessing the collected data and then constructing a multi-dimensional feature analysis model, wherein the multi-dimensional features comprise people flow composition features, behavior pattern features and environment coupling features; constructing a sample set based on feature vectors output by the multi-dimensional feature analysis model, outputting the sample set to a deep learning model to calculate a people flow density index, and triggering graded early warning when the people flow density index exceeds a first preset threshold value; generating multi-language voice early warning information according to the graded early warning, and performing regionalized broadcasting through directional sound field equipment; and when the people flow density index exceeds a second preset threshold value, generating a tourist shunting scheme including optimized path planning, thereby effectively improving the accuracy of people flow density detection and the timeliness of early warning.

Description

Technical Field

[0001] The present invention relates to the technical field of pedestrian flow density detection, and particularly relates to a method and system for detecting and warning the pedestrian flow density in scenic spots. Background Art

[0002] Currently, the detection of pedestrian flow density in scenic spots mainly relies on single-dimensional pedestrian counting, without combining multi-dimensional feature modeling based on the unique topographic and geomorphic spatial structure of the region, the behavioral characteristics of cross-border tourists, and the ethnic festival gathering patterns. For example, there are many vertical layers in karst caves and multi-level platforms in terraced fields in the Guangxi Zhuang Autonomous Region; there are characteristics such as a sudden increase in the instantaneous pedestrian flow during the Zhuang Singing Festival. The existing density detection algorithms have low detection accuracy in complex scenarios due to the lack of consideration of multi-dimensional features. In addition, the traditional warning system lacks a dynamic threshold adjustment method and fails to conduct a coupled analysis of multiple factors such as meteorological environment and ecologically sensitive areas, resulting in untimely warnings and false alarms. These problems lead to the inability to achieve precise diversion and guidance during peak periods such as Golden Week, and frequent congestion occurs in special areas such as the stone-paved roads in ethnic villages and the border crossing channels, greatly reducing the tourist experience.

[0003] In view of this, a method and system for detecting and warning the pedestrian flow density in scenic spots are needed. Summary of the Invention

[0004] An embodiment of the present application provides a method for detecting and warning the pedestrian flow density in scenic spots, which is used to solve the problem of congestion of tourists in scenic spots caused by inaccurate detection of pedestrian flow density.

[0005] A first aspect of an embodiment of the present application provides a method for detecting and warning the pedestrian flow density in scenic spots, including:

[0006] Collecting real-time passenger flow data through intelligent devices deployed at key nodes in the scenic spot;

[0007] Preprocessing the collected data and constructing a multi-dimensional feature analysis model, where the multi-dimensional features include pedestrian flow composition features, behavior pattern features, and environmental coupling features;

[0008] Constructing a sample set based on the feature vectors output by the multi-dimensional feature analysis model, outputting the sample set to a deep learning model to calculate the pedestrian flow density index, and triggering a hierarchical warning when the pedestrian flow density index exceeds a first preset threshold;

[0009] Generating multi-language voice warning information according to the hierarchical warning and broadcasting it regionally through a directional sound field device;

[0010] When the pedestrian flow density index exceeds a second preset threshold, generating a tourist diversion plan including optimized path planning.

[0011] Furthermore, after preprocessing the collected data, a multi-dimensional feature analysis model is constructed. The multi-dimensional features include the characteristics of the flow of people composition, behavior patterns, and environmental coupling, including:

[0012] The characteristics of the flow of people composition include the clustering analysis of the tourist origin, the classification of the tourist stay duration, the identification of the group type, the stratification of the consumption ability, and the analysis of the tour preferences. The expressions are as follows:

[0013] Clustering analysis of tourist origin:

[0014]

[0015]

[0016] Among them: H origin is the entropy of the origin, which is used to measure the dispersion degree of the tourist origin. p c is the proportion of tourists from origin c. N c is the number of tourists from origin c. N total is the total number of tourists;

[0017] Classification of tourist stay duration:

[0018]

[0019] Among them: is the stay duration of the i-th tourist, is the time when the i-th tourist leaves the park, is the time when the i-th tourist enters the park;

[0020] Identification of group type:

[0021]

[0022] Among them: G is the group tightness, which is used to measure the synchronization degree of the trajectories among team members. n is the number of tourists in the team. cov(p i , p j ) is the covariance of the trajectory coordinate sequences of tourist i and tourist j, which reflects the correlation degree of their trajectories. p i and p j are the trajectory coordinate sequences of tourist i and tourist j respectively. σ i and σ j are the standard deviations of the trajectory coordinate sequences of tourist i and tourist j respectively;

[0023] Stratification of consumption ability:

[0024]

[0025] Among them: C is the consumption index, which is used to measure the consumption ability of tourists. wk is the weight of consumption type, price k is the amount of the k-th consumption, k is the total number of consumption records, T stay is the stay duration of the tourist;

[0026] Tour preference analysis:

[0027]

[0028] Among them: V prefer is the tour preference vector, reflecting the preference degree of tourists for different types of scenic spots, t nature is the stay duration of tourists at natural landscape interest points, t culture is the stay duration of tourists at ethnic culture interest points, t adventure is the stay duration of tourists at outdoor adventure interest points.

[0029] Furthermore, after preprocessing the collected data, a multi-dimensional feature analysis model is constructed. The multi-dimensional features include the flow composition feature, the behavior pattern feature, and the environmental coupling feature, including:

[0030] The behavior pattern features include the high-frequency movement pattern of cross-border tourists, the analysis of the stay balance of tourists in the terraced farming area, the prediction of the behavior of tourists exploring karst caves and underground rivers, and the analysis of the gathering at the Zhuang ethnic song fair. The expressions are as follows:

[0031] High-frequency movement pattern of cross-border tourists:

[0032]

[0033] Among them: F Cross is the movement frequency of cross-border tourists, d i is the displacement of tourist i in the time window Δt, and n′ is the number of displacement data in the time window Δt;

[0034] Analysis of the stay balance of tourists in the terraced farming area:

[0035]

[0036] Among them: E balance is the stay balance index of tourists in the terraced farming area, K is the total number of terraced farming areas, N k is the number of tourists in the k-th terraced farming area, and μ is the global average density;

[0037] Prediction of the behavior of tourists exploring karst caves and underground rivers:

[0038]

[0039] Among them: P expiore is the probability that tourists carry out the behavior of exploring karst caves and underground rivers, Ilight Light intensity level selected for tourists, A oge is the age coefficient;

[0040] Analysis of the gathering of the Zhuang ethnic group's song fair:

[0041]

[0042] Among them: ρ song (x, y, t) is the spatio-temporal density of the Zhuang ethnic group's song fair at the coordinates (x, y) at time t, n″ is the number of participants in the antiphonal singing, w i is the weight of the participants in the antiphonal singing, (x i , y i ) are the coordinates of the i-th participant in the antiphonal singing, and σ = 5m is the influence radius of cultural gathering.

[0043] Furthermore, after preprocessing the collected data, a multi-dimensional feature analysis model is constructed. The multi-dimensional features include the flow composition feature, the behavior pattern feature, and the environmental coupling feature, including:

[0044] The environmental coupling feature includes flood prediction within the region, material consumption in the Zhuang brocade production experience area, early warning of the flow of people hedging at the border port, and pressure feedback in the ecological area. The expressions are as follows respectively:

[0045] Flood prediction within the region:

[0046]

[0047] Among them: T flood is the time required for the predicted flood to arrive, R vn is the hourly rainfall at the monitoring point, A basin is the river basin area, K terrain is the leakage coefficient within the region;

[0048] Material consumption in the Zhuang brocade production experience area:

[0049]

[0050] Among them: S material is the resource pressure index of the Zhuang brocade production experience area, used to measure the pressure degree borne by the material resources in this experience area, d is the number of days, U d is the consumption of Zhuang brocade wire on the d-th day, Q d is the daily material supply, N d is the face of the participants in the experience;

[0051] Early warning of the flow of people hedging at the border port:

[0052]

[0053] Among them: ΔFlow is the border crossing passenger flow hedging index, which is used to measure the imbalance degree and movement of the two-way passenger flow at the port, N in and N out are respectively the number of inbound passengers and outbound passengers within a preset time window, and V avg is the average movement speed of the passage;

[0054] Ecological area pressure feedback:

[0055]

[0056] Among them: F eco is the ecological footprint of the scenic area ecological area, n is the number of tourists, and t i is the staying duration of tourist i, D i is the minimum distance between tourist i and the ecological area, and D crit is the ecological safety distance.

[0057] Furthermore, constructing a sample set based on the feature vectors output by the multi-dimensional feature analysis model, and outputting the sample set to a deep learning model to calculate the passenger flow density index. When the passenger flow density index exceeds the first preset threshold, a hierarchical early warning is triggered, including:

[0058] Performing feature spatio-temporal encoding and temporal feature slicing on the feature vectors output by the multi-dimensional feature analysis model;

[0059] Performing feature fusion on the multi-modal features after spatio-temporal encoding and temporal feature slicing through the feature gating algorithm of the attention mechanism;

[0060] Calculating the passenger flow density index using an improved GraphSAGE model, and triggering a hierarchical early warning when the calculated passenger flow density index exceeds the first preset threshold.

[0061] Furthermore, calculating the passenger flow density index using an improved GraphSAGE model, and triggering a hierarchical early warning when the calculated passenger flow density index exceeds the first preset threshold, including:

[0062] Calculation formula of the passenger flow density index:

[0063]

[0064] D density = MLP(CONCAT(h graph , F env ))

[0065] Among them: is the feature representation of node v at the l+1 layer, σ is the activation function, and W1 and W2 are trainable parameter matrices, is the feature representation of node v at the l-th layer, is the neighbor set of node v which is the average value of the feature representations of all neighbor nodes v in the neighbor set of node v at the l-th layer, is the neighbor set of node v, including three types of special neighbors: cross-border, ethnic, and environmental; D density is the calculated pedestrian flow density index, MLP is a multi-layer perceptron with a structure of 256-128-64-1, using the GELU activation function, CONCAT is a concatenation operation that concatenates the graph feature h graph and the environmental feature F env F env is an environmental feature vector including real-time humidity, the state of the stalactite growth area, etc.

[0066] Furthermore, generating multi-language voice warning information according to the hierarchical warning and performing regional broadcast through a directional sound field device includes:

[0067] Converting the warning information in each language into a voice format. When the pedestrian flow density index of the scenic area triggers the corresponding hierarchical warning, automatically send the multi-language voice warning information of the corresponding warning level to the directional sound field device in the specified area, so that the device broadcasts the voice information to the target area according to the preset parameters.

[0068] Furthermore, when the pedestrian flow density index exceeds the second preset threshold, generating a tourist diversion plan including an optimized path plan, including:

[0069] When the pedestrian flow density index exceeds the second preset threshold, use the improved ant colony algorithm to calculate the optimal diversion path;

[0070] And dynamically adjust the diversion plan according to the preset time period.

[0071] Furthermore, when the pedestrian flow density index exceeds the second preset threshold, using the improved ant colony algorithm to calculate the optimal diversion path, including:

[0072] F = min(λ1∑d ij +λ2maxρ seg +λ3∑S eco )

[0073] where: F is the optimal path, ∑d ij is the sum of the distances of each section in the path, d ij is the distance of the section from position i to position j, λ1 is the weight corresponding to d ij maxρ seg is the maximum pedestrian flow density in each section of the path, λ2 is the weight corresponding to maxρ seg ∑S ecoIt is the total weight of the ecologically sensitive areas passed by the path, S eco It is the weight of the ecologically sensitive area, and λ3 is for S eco The corresponding weight.

[0074] The second aspect of the embodiment of the present application provides a scenic area crowd density detection and early warning system, including:

[0075] A real-time passenger flow data acquisition unit, which is used to collect real-time passenger flow data through intelligent devices deployed at key nodes in the scenic area;

[0076] A multi-dimensional feature analysis model construction unit, which is used to preprocess the collected data and construct a multi-dimensional feature analysis model. The multi-dimensional features include the composition characteristics of the crowd, the behavior pattern characteristics, and the environmental coupling characteristics;

[0077] A crowd density index calculation unit, which is used to construct a sample set based on the feature vector output by the multi-dimensional feature analysis model, output the sample set to a deep learning model to calculate the crowd density index, and trigger a hierarchical early warning when the crowd density index exceeds a first preset threshold;

[0078] An early warning information generation unit, which is used to generate multi-language voice early warning information according to the hierarchical early warning and broadcast it regionally through a directional sound field device;

[0079] A tourist diversion plan generation unit, which is used to generate a tourist diversion plan including optimized path planning when the crowd density index exceeds a second preset threshold.

[0080] From the above technical solutions, it can be seen that the embodiments of the present invention have the following advantages:

[0081] Based on the tourist data obtained by the present invention, a multi-dimensional feature analysis is constructed. By classifying the tourist sources, the detection coverage rate can be improved based on the tourist distribution under complex terrains; an improved deep learning algorithm is used to effectively solve the misjudgment problem of traditional algorithms in scenarios of high-frequency movement of cross-border tourist groups and instantaneous gatherings during ethnic festivals by dynamically constructing a cross-border tourist association subgraph and a national culture feature enhancement module; when it is determined that there is a crowded situation, multi-language directional sound field early warnings are generated; when the crowd density index reaches a certain value, a tourist diversion plan including an optimized path is generated to quickly evacuate the crowd and reduce congestion. Description of the Drawings

[0082] Figure 1 It is a schematic flowchart of an embodiment of a scenic area crowd density detection and early warning method in the present invention. Detailed Embodiments

[0083] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present 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 only used to explain the present invention and are not used to limit the present invention.

[0084] In this embodiment, the method for detecting and warning the crowd density in the scenic area is used to reduce the congestion in the scenic area and improve the tourist experience. The implementation method in this embodiment can be implemented in the system, can be implemented on the server, or can be implemented on the terminal, and no specific limitation is made.

[0085] Embodiment 1

[0086] Please refer to Figure 1 , an embodiment of a method for detecting and warning the crowd density in a scenic area in the present invention includes the following steps:

[0087] S11. Collect real-time passenger flow data through intelligent devices deployed at key nodes in the scenic area;

[0088] In this embodiment, various intelligent devices are deployed at key node positions in the scenic area such as entrances, exits, popular scenic spots, narrow passages, transfer points, etc. These devices can be cameras, which can use video analysis technology to identify and count the number of tourists; Wi-Fi probes, which can count the number of people and analyze the movement trajectories by detecting the Wi-Fi signals of devices such as tourists' mobile phones; Bluetooth beacons, which can identify tourists' devices and obtain location information based on Bluetooth low-power technology; turnstiles, which have a counting function and record the number of tourists entering and leaving a specific area, etc. The intelligent devices continuously collect real-time passenger flow data in the scenic area, including information such as the number of tourists, locations, and movement directions, providing a comprehensive data basis for the subsequent steps.

[0089] S12. Preprocess the collected data and then construct a multi-dimensional feature analysis model. The multi-dimensional features include crowd composition features, behavior pattern features, and environmental coupling features;

[0090] Clean the collected original passenger flow data to remove noise data, abnormal data, and duplicate data; perform data format conversion and normalization processing for subsequent analysis. For example, remove duplicates from the object detection results in the video data and uniformly calibrate the timestamps collected by different devices. Construct a multi-dimensional feature analysis model based on the preprocessed data, where the multi-dimensional features include crowd composition features, behavior pattern features, and environmental coupling features.

[0091] The crowd composition features include cluster analysis of tourist origin areas, classification of tourist stay durations, identification of group types, stratification of consumption capabilities, and analysis of tourist preferences. The expressions are as follows:

[0092] Cluster analysis of tourist origin areas:

[0093]

[0094] Among them: H origin is the entropy of the source area, used to measure the dispersion degree of the tourist source area, and p c is the proportion of tourists from source area c, and N c is the number of tourists from source area c, and N total is the total number of tourists;

[0095] Classification of tourist stay duration:

[0096]

[0097] Among them: is the stay duration of the i-th tourist, is the time when the i-th tourist leaves the park, is the time when the i-th tourist enters the park;

[0098] Recognition of group types:

[0099]

[0100] Among them: G is the group tightness, used to measure the synchronization degree of the trajectories among group members, n is the number of tourists in the group, and cov(p i , p j ) is the covariance of the trajectory coordinate sequences of tourist i and tourist j, reflecting the correlation degree of their trajectories, and p i and p j are the trajectory coordinate sequences of tourist i and tourist j respectively, and σ i and σ j are the standard deviations of the trajectory coordinate sequences of tourist i and tourist j respectively;

[0101] Stratification of consumption ability:

[0102]

[0103] Among them: C is the consumption index, used to measure the consumption ability of tourists, w k is the weight of the consumption type, price k is the amount of the k-th consumption, K is the total number of consumption records, and T stay is the stay duration of the tourist;

[0104] Analysis of tour preferences:

[0105]

[0106] Among them: V prefer is the tour preference vector, reflecting the preference degree of tourists for different types of scenic spots, and t nature is the stay duration of the tourist at the natural landscape type of interest points, and tculture The residence time of tourists at ethnic culture - related interest points is \(t\). adventure The residence time of tourists at outdoor adventure - related interest points is \(t'\).

[0107] The behavior pattern features include the high - frequency movement pattern of cross - border tourists, the equilibrium analysis of tourists' residence in terraced farming areas, the prediction of cave and underground river exploration behavior, and the aggregation analysis of the Zhuang ethnic group's song fair. Their expressions are as follows:

[0108] High - frequency movement pattern of cross - border tourists:

[0109]

[0110] Where: \(F\) Cross is the movement frequency of cross - border tourists, \(d\) i is the displacement of tourist \(i\) within the time window \(\Delta t\), and \(n'\) is the number of displacement data within the time window \(\Delta t\);

[0111] Equilibrium analysis of tourists' residence in terraced farming areas:

[0112]

[0113] Where: \(E\) balance is the equilibrium index of tourists' residence in terraced farming areas, \(K\) is the total number of terraced fields divided into zones, \(N\) k is the number of tourists in the \(k\) - th terraced field zone, and \(\mu\) is the global average density;

[0114] Prediction of cave and underground river exploration behavior:

[0115]

[0116] Where: \(P\) expiore is the probability that tourists carry out cave and underground river exploration behavior, \(I\) light is the light intensity level selected by tourists, \(A\) oge is the age coefficient;

[0117] Aggregation analysis of the Zhuang ethnic group's song fair:

[0118]

[0119] Where: \(\rho\) song \((x,y,t)\) is the spatio - temporal density of the Zhuang ethnic group's song fair at the coordinates \((x,y)\) at time \(t\), \(n''\) is the number of song - singing participants, \(w\) i is the weight of song - singing participants, \((x\) i ,y i ) is the coordinate of the \(i\) - th song - singing participant, and \(\sigma = 5m\) is the cultural aggregation influence radius.

[0120] The environmental coupling features include flood prediction within the region, material consumption in the Zhuang brocade production experience area, warning of the impact of the flow of people at the border crossing, and pressure feedback in the ecological area. The expressions are as follows:

[0121] Flood prediction within the region:

[0122]

[0123] Where: T flood is the time required for the predicted flood to arrive, R vn is the hourly rainfall at the monitoring point, A basin is the river basin area, K terrain is the leakage coefficient within the region;

[0124] Material consumption in the Zhuang brocade production experience area:

[0125]

[0126] Where: S material is the resource pressure index of the Zhuang brocade production experience area, used to measure the pressure on the material resources in this experience area, d is the number of days, U d is the consumption of Zhuang brocade wire on the d-th day, Q d is the daily material supply, N d is the number of faces participating in the experience;

[0127] Warning of the impact of the flow of people at the border crossing:

[0128]

[0129] Where: ΔFlow is the border crossing people flow impact index, used to measure the imbalance and movement of the two-way people flow at the border crossing, N in and N out are the number of inbound and outbound passengers within the preset time window respectively, V avg is the average moving speed of the passage;

[0130] Pressure feedback in the ecological area:

[0131]

[0132] Where: F eco is the ecological footprint of the scenic area's ecological area, n is the number of tourists, t i is the staying duration of tourist i, D i is the minimum distance between tourist i and the ecological area, D crit is the ecological security distance.

[0133] S13. Construct a sample set based on the feature vectors output by the multi-dimensional feature analysis model, and output the sample set to the deep learning model to calculate the crowd density index. When the crowd density index exceeds the first preset threshold, a hierarchical warning is triggered;

[0134] 1. Perform feature spatio-temporal encoding and temporal feature slicing on the feature vectors output by the multi-dimensional feature analysis model;

[0135] After the multi-dimensional feature analysis model processes the data, it outputs the feature vectors corresponding to each time point or time period. These feature vectors contain information in many aspects such as the composition of the crowd, behavior patterns, and environmental coupling. Integrate these feature vectors to construct a sample set as the input data of the deep learning model. Integrate the above feature matrix with the temporal data (the features of the previous 5-minute window) into a spatio-temporal tensor. First, perform spatial encoding of Zhuang culture on the scenic area: Each 10m×10m grid is assigned a unique code, and the cultural sensitive area (such as the song fairground) is additionally assigned a weight factor of 1.5. Construct a sliding time window (window size = 5min, step size = 1min) to extract the feature trend: ΔF t = R t ―0.6F t―1 ―0.4F t―2 , and the coefficient setting reflects the suddenness of the tourist flow in Guangxi (such as the concentrated arrival of cross-border buses).

[0136] 2. Perform feature fusion on the multi-modal features processed by the feature gating algorithm of the attention mechanism for the spatio-temporal encoding and temporal feature slicing;

[0137] Based on the above feature analysis, calculate the weights of each feature. The calculation formula for the feature importance weight:

[0138]

[0139] where: W a ∈R 9×1 is a trainable parameter matrix.

[0140] Perform cross-modal interaction and fusion on each feature:

[0141]

[0142] where: is the Hadamard product.

[0143] 3. Use the improved GraphSAGE model to calculate the crowd density index. When the calculated crowd density index exceeds the first preset threshold, a hierarchical warning is triggered.

[0144] Calculate the crowd density index using an improved GraphSAGE model, which includes a domain sampling strategy, i.e., cross-region sampling for cross-border tourist nodes and cultural association sampling for tourists wearing Zhuang ethnic costumes. The aggregation function and the crowd density calculation formula are as follows:

[0145]

[0146] D density = MLP(CONCAT(h graph , F env ))

[0147] Where: is the feature representation of node v at the l+1 layer, σ is the activation function, W1 and W2 are trainable parameter matrices, is the feature representation of node v at the l layer, is the set of neighbors of node v The average value of the feature representations of all neighbor nodes v in the l layer in the set, is the neighbor set of node v, including three types of special neighbors: cross-border, ethnic, and environmental; D density is the calculated crowd density index, MLP is a multi-layer perceptron with a structure of 256-128-64-1, using the GELU activation function, CONCAT is a concatenation operation that concatenates the graph feature h graph and the environmental feature F env concatenate, F env is an environmental feature vector including real-time humidity, the state of stalactite growth areas, etc.

[0148] Assume the threshold range is 0.6. When the detected crowd density exceeds 0.6, a first-level alarm is generated. And a correction is made for the cross-border factor, i.e., when the proportion of cross-border tourists is greater than 50%, the threshold is automatically lowered by 20%.

[0149] S14. Generate multi-language voice warning messages according to the graded warning and broadcast them regionally through a directional sound field device;

[0150] In this embodiment, the warning level can be divided into different levels according to the degree of exceeding the threshold, such as mild congestion (yellow warning), moderate congestion (orange warning), severe congestion (red warning), etc. Each level corresponds to different countermeasures and emergency levels.

[0151] Convert the warning information in each language into a voice format. When the scenic area crowd density index triggers the corresponding graded warning, automatically send the voice warning information in multiple languages at the corresponding warning level to the directional sound field devices in the specified area, so that the devices broadcast the voice information to the target area according to the preset parameters. According to different warning levels, generate corresponding multiple languages, such as Chinese, English, Japanese, Korean, Vietnamese, etc., considering the voice warning information of the main tourist sources and international tourists in the scenic area, and the content includes the current congestion situation, possible risk warnings, and suggestions for tourists to take actions, such as passing in an orderly manner and avoiding gathering. Use the directional sound field devices to accurately broadcast the voice warning information to specific areas within the scenic area, that is, areas with high crowd density or likely to be congested, to ensure that tourists can receive the warning information in a timely manner without disturbing other areas.

[0152] S15. When the crowd density index exceeds the second preset threshold, generate a tourist diversion plan including optimized path planning.

[0153] 1. When the crowd density index exceeds the second preset threshold, use the improved ant colony algorithm to calculate the optimal diversion path;

[0154] 2. And dynamically adjust the diversion plan according to the preset time period.

[0155] In this embodiment, set the second preset threshold. When the crowd density index exceeds this higher-level threshold, it indicates that the congestion situation in the scenic area has been very serious and may affect the experience and safety of tourists. At this time, according to the real-time passenger flow data, scenic area map, scenic spot carrying capacity and other information, generate a tourist diversion plan including optimized path planning.

[0156] The calculation formula for using the improved ant colony algorithm to calculate the optimal diversion path is as follows:

[0157] F = min(λ1∑d ij + λ2maxρ seg + λ3∑S eco )

[0158] Where: F is the optimal path, ∑d ij is the sum of the distances of each section of the path, d ij is the distance of the section from position i to position j, λ1 is the weight corresponding to d ij , maxρ seg is the maximum crowd density in each section of the path, λ2 is the weight corresponding to maxρ seg , ∑S eco is the total weight of the ecological sensitive areas passed by the path, S eco is the weight of the ecological sensitive area, and λ3 is the weight corresponding to S eco .

[0159] Dynamically adjust the calculation formula of the diversion plan according to the preset time period as follows:

[0160]

[0161] Where: ΔT is the preset update period, T base = 5 min, which is shortened to 3 min in the rainy season.

[0162] The diversion plan will plan reasonable evacuation or detour routes for tourists in different areas, guide tourists to relatively less crowded areas to relieve the congestion. At the same time, inform tourists of the diversion plan in a timely manner through channels such as the scenic area's APP, electronic display screens, and broadcasts to ensure that tourists can be diverted in an orderly manner according to the planned routes.

[0163] Embodiment 2

[0164] An embodiment of a scenic area crowd density detection and early warning system in the present invention includes the following steps:

[0165] A real-time passenger flow data acquisition unit, which is used to acquire real-time passenger flow data through intelligent devices deployed at key nodes in the scenic area;

[0166] A multi-dimensional feature analysis model construction unit, which is used to preprocess the collected data and construct a multi-dimensional feature analysis model. The multi-dimensional features include the crowd composition feature, behavior pattern feature, and environmental coupling feature;

[0167] A crowd density index calculation unit, which is used to construct a sample set based on the feature vectors output by the multi-dimensional feature analysis model, output the sample set to a deep learning model to calculate the crowd density index, and trigger a hierarchical early warning when the crowd density index exceeds the first preset threshold;

[0168] An early warning information generation unit, which is used to generate multi-language voice early warning information according to the hierarchical early warning and perform regionalized broadcasts through a directional sound field device;

[0169] A tourist diversion plan generation unit, which is used to generate a tourist diversion plan including an optimized route plan when the crowd density index exceeds the second preset threshold.

[0170] Those of ordinary skill in the art can realize that the units of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the components of each example have been generally described according to their functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0171] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored, etc. In addition, in each embodiment of the present invention, each functional unit can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0172] If the above integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs that can store program codes.

[0173] It can be understood that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of each embodiment of the present invention, and they should all be covered within the scope of the claims and the description of the present invention.

Claims

1. A method for detecting and warning the density of people in a scenic area, characterized in that: include: Collect real-time passenger flow data through smart devices deployed at key nodes in scenic spots; After preprocessing the collected data, a multi-dimensional feature analysis model is constructed. The multi-dimensional features include crowd flow composition features, behavior pattern features, and environmental coupling features. Constructing a sample set based on the feature vector output by the multi-dimensional feature analysis model, outputting the sample set to a deep learning model to calculate a crowd density index, and triggering a graded warning when the crowd density index exceeds a first preset threshold; Generate multi-language voice warning information according to the graded warning, and broadcast it regionally through directional sound field equipment; When the crowd density index exceeds a second preset threshold, a tourist diversion plan including optimized path planning is generated.

2. The method for detecting and warning the density of people in a scenic area according to claim 1, characterized in that: The collected data is preprocessed to construct a multi-dimensional feature analysis model, and the multi-dimensional features include crowd flow composition features, behavior pattern features, and environmental coupling features, including: The characteristics of the flow of people include cluster analysis of tourists’ origin, classification of tourists’ length of stay, identification of group types, stratification of consumption capacity and analysis of sightseeing preferences, and the expressions are as follows: Cluster analysis of tourists’ origins: Where: H origin is the source entropy, which is used to measure the dispersion of tourists’ sources, p c is the proportion of tourists from origin c, N c is the number of tourists from origin c, N total is the total number of tourists; Tourist stay duration classification: in: is the length of stay of the i-th tourist, is the departure time of the i-th tourist, is the entry time of the i-th visitor; Team Type Identification: Where: G is the team closeness, which is used to measure the degree of synchronization between team members’ trajectories, n is the number of tourists in the team, cov(p i ,p j ) is the covariance of the trajectory coordinate sequence of tourist i and tourist j, reflecting the degree of correlation between the two trajectories, p i and p j are the trajectory coordinate sequences of tourist i and tourist j respectively, σ i and σ j are the standard deviations of the trajectory coordinate sequences of tourists i and j, respectively; Consumption capacity stratification: Where: C is the consumption index, which is used to measure the consumption capacity of tourists, w k is the consumption type weight, price k is the amount of the kth consumption, K is the total number of consumption records, T stay The length of stay of tourists; Tour preference analysis: Where: V prefer is the tourist preference vector, reflecting tourists’ preference for different types of attractions, t nature is the length of time tourists stay at natural landscape points of interest, t culture is the length of time tourists stay at ethnic and cultural points of interest, t adventure The length of time that tourists stay at outdoor adventure points of interest.

3. The method for detecting and warning the density of people flow in scenic spots according to claim 2, characterized in that: The collected data is preprocessed to construct a multi-dimensional feature analysis model, and the multi-dimensional features include crowd flow composition features, behavior pattern features, and environmental coupling features, including: The behavioral pattern characteristics include the high-frequency movement pattern of cross-border tourists, the equilibrium analysis of tourists’ stay in terraced farming areas, the prediction of cave and underground river exploration behavior, and the aggregation analysis of Zhuang ethnic group singing festivals, and the expressions are as follows: High-frequency movement patterns of cross-border tourists: Among them: F Cross is the movement frequency of cross-border tourists, d i is the displacement of tourist i in the time window Δt, n′ is the number of displacement data in the time window Δt; Analysis of tourist stay equilibrium in terraced farming areas: Where: E balance is the tourist stay equilibrium index in the terraced farming area, K is the total number of terraced partitions, N k is the number of tourists in the kth terrace partition, μ is the global average density; Prediction of cave and underground river exploration behavior: Where: P expiore The probability of tourists exploring caves and underground rivers, I light Light intensity level selected for visitors, A oge is the age coefficient; Analysis of Zhuang Nationality Song Festival: Where: song (x,y,t) is the spatiotemporal density of the Zhuang Nationality Song Festival at the coordinate (x,y) at time t, n″ is the number of participants in the song, w i is the weight of the singing participant, (x i ,y i ) is the coordinate of the i-th singing participant, σ=5m is the influence radius of cultural aggregation.

4. The method for detecting and warning the density of people flow in scenic spots according to claim 2, characterized in that: The collected data is preprocessed to construct a multi-dimensional feature analysis model, and the multi-dimensional features include crowd flow composition features, behavior pattern features, and environmental coupling features, including: The environmental coupling characteristics include regional flood prediction, material consumption in the Zhuang brocade production experience area, border port passenger flow hedging warning, and ecological zone pressure feedback, and the expressions are as follows: Flood forecast for the area: Where: T flood To predict the time required for flood to arrive, R vn is the hourly rainfall at the monitoring point, A basin is the river basin area, K terrain is the leakage coefficient in the area; Material consumption in the Zhuang Brocade Making Experience Area: Where: S material is the resource pressure index of the Zhuang Brocade Making Experience Area, which is used to measure the pressure level of the material resources in the experience area. d is the number of days, and U is d is the consumption of Zhuangjin wire on the dth day, Q d is the daily material supply, N d For the faces of people participating in the experience; Warning of crowd flow at border ports: Where: ΔFlow is the hedging index of the border port flow, which is used to measure the imbalance and movement of the two-way flow of people at the port, N in and N out are the number of inbound and outbound people in the preset time window, V avg is the average moving speed of the channel; Ecoregion Pressure Feedback: Among them: F eco is the ecological footprint of the scenic area, n is the number of tourists, t i is the length of stay of tourist i, D i is the minimum distance between tourist i and ecological zone, D crit It is the ecological safety distance.

5. The method for detecting and warning the density of people flow in scenic spots according to claim 1, characterized in that: The step of constructing a sample set based on the feature vector output by the multi-dimensional feature analysis model, outputting the sample set to a deep learning model to calculate a crowd density index, and triggering a graded warning when the crowd density index exceeds a first preset threshold value includes: Performing feature spatiotemporal coding and temporal feature slicing on the feature vector output by the multi-dimensional feature analysis model; The feature gating algorithm of the attention mechanism is used to fuse the multimodal features processed by spatiotemporal coding and temporal feature slicing; The improved GraphSAGE model is used to calculate the crowd density index, and a graded warning is triggered when the calculated crowd density index exceeds a first preset threshold.

6. The method for detecting and warning the density of people flow in scenic spots according to claim 5, characterized in that: The improved GraphSAGE model is used to calculate the crowd density index, and when the calculated crowd density index exceeds a first preset threshold, a graded warning is triggered, including: The calculation formula of the crowd density index is: D density =MLP(CONCAT(h graph ,F env )) in: is the feature representation of node v at layer l+1, σ is the activation function, W1 and W2 are trainable parameter matrices, is the feature representation of node v at layer l, is the neighbor set of node v The average value of the feature representation of all neighbor nodes v in layer l, is the neighbor set of node v, including three special neighbors: cross-border, ethnic, and environmental; D density To calculate the crowd density index, MLP is a multi-layer perceptron with a structure of 256-128-64-1, using the GELU activation function, and CONCAT is a concatenation operation. graph and environmental characteristics F env Splicing, F env is an environmental feature vector including real-time humidity, state of stalactite growth area, etc.

7. The method for detecting and warning the density of people flow in scenic spots according to claim 1, characterized in that: The generating of multi-language voice warning information according to the graded warning and performing regionalized broadcasting through directional sound field equipment includes: The warning information in each language is converted into voice format. When the crowd density index of the scenic spot triggers the corresponding graded warning, the multi-language voice warning information of the corresponding warning level is automatically sent to the directional sound field equipment in the designated area, so that the equipment broadcasts the voice information to the target area according to the preset parameters.

8. The method for detecting and warning the density of people flow in scenic spots according to claim 1, characterized in that: When the crowd density index exceeds the second preset threshold, a tourist diversion plan including optimized path planning is generated, including: When the crowd density index exceeds a second preset threshold, an improved ant colony algorithm is used to calculate an optimal diversion path; And dynamically adjust the diversion plan according to the preset time period.

9. The method for detecting and warning the density of people flow in scenic spots according to claim 8, characterized in that: When the crowd density index exceeds the second preset threshold, using an improved ant colony algorithm to calculate the optimal diversion path includes: F=min(λ1∑d ij +λ2maxρ seg +λ3∑S eco ) Where: F is the optimal path, ∑d ij is the sum of the distances of each section in the path, d ij is the distance from position i to position j, λ1 is d ij The corresponding weight, maxρ seg is the maximum flow density in each section of the path, λ2 is maxρ seg The corresponding weight, ∑S eco is the sum of the weights of the ecologically sensitive areas that the path passes through, S eco is the weight of ecologically sensitive areas, λ3 is S eco The corresponding weight.

10. A scenic spot crowd density detection and early warning system, characterized in that: The method for detecting and warning the crowd density in a scenic spot according to any one of claims 1 to 9 comprises: Real-time passenger flow data collection unit, used to collect real-time passenger flow data through smart devices deployed at key nodes in the scenic area; A multi-dimensional feature analysis model building unit is used to build a multi-dimensional feature analysis model after pre-processing the collected data. The multi-dimensional features include crowd flow composition features, behavior pattern features and environmental coupling features. A crowd density index calculation unit, used to construct a sample set based on the feature vector output by the multi-dimensional feature analysis model, output the sample set to a deep learning model to calculate the crowd density index, and trigger a graded warning when the crowd density index exceeds a first preset threshold; A warning information generating unit, used to generate voice warning information in multiple languages ​​according to the graded warning, and to broadcast it regionally through a directional sound field device; A tourist diversion plan generating unit is used to generate a tourist diversion plan including optimized path planning when the crowd density index exceeds a second preset threshold.

Images