Early warning control method based on scenic spot passenger flow big data time sequence learning

Abstract

This invention relates to the field of intelligent management technology and discloses an early warning and control method based on time-series learning of big data on tourist flow in scenic areas. The method includes: analyzing the time-series characteristics of meteorological forecast data during the period of sudden weather warning for the target scenic area to obtain meteorological event characteristic parameters; combining the meteorological event characteristic parameters, real-time tourist flow, and historical tourist flow database to extrapolate the influx of tourists into indoor refuge areas to obtain the estimated total tourist flow; coupling road network topology and real-time traffic status data to construct a traffic impedance diagram, assessing the congestion of potentially stressed road sections and entrance nodes to obtain an initial congestion risk index; when the risk index exceeds a safety threshold, planning the source diversion and allocation of refuge routes to obtain differentiated guidance routes; encapsulating the guidance routes and pushing them to tourists' mobile terminals, comparing the location data with the path vector trajectory to obtain a flow convergence index; when the convergence index exceeds a preset level, adjusting the timing and frequency of the push according to the excess range, and encoding evacuation guidance instructions.

Description

Technical Field

[0001] This invention relates to the field of intelligent management technology, and in particular to an early warning and control method based on time-series learning of big data on tourist flow in scenic areas. Background Technology

[0002] With the development of smart tourism and comprehensive emergency management technologies, scenic spots have generally built IoT and big data systems such as meteorological monitoring, visitor flow perception, road network monitoring, and mobile information push to cope with sudden weather events such as heavy rain, thunderstorms, and strong winds. The current common practice is to integrate weather forecasts, real-time visitor flow, and road traffic data to carry out disaster early warning, emergency shelter scheduling, and visitor evacuation guidance, thereby reducing the risks that sudden weather events may bring to visitor safety and the normal operation of the scenic spot.

[0003] Existing emergency evacuation and visitor flow guidance technologies for sudden weather events in scenic areas still have significant shortcomings. In terms of meteorological data application, most systems simply collect and use relevant data without performing multi-temporal granular reorganization of weather forecast information. They also lack spatiotemporal coupling analysis of frontal migration and precipitation echo envelopes. Especially in mountainous and waterfront scenic areas, relying solely on single-time-step meteorological thresholds to determine risk fails to accurately capture the evolutionary characteristics of precipitation front movement rates and echo morphology, resulting in insufficient quantitative accuracy of disaster severity and impact range, making it difficult to support refined allocation of evacuation flow. Regarding evacuation route planning, existing schemes mostly refer only to static road network distances or congestion status at a single moment, without considering dynamic weight allocation and push strategy adjustment based on evacuation flow load, traffic impedance, and the convergence effect of pedestrian flow timing. In canyon-type scenic areas, multiple routes are prone to concentrated influx of visitors, leading to instantaneous overload of local road sections. The system also cannot suppress or balance the push frequency based on the degree of pedestrian flow convergence, ultimately resulting in low evacuation efficiency and poor coordination between road network capacity and evacuation demand. Summary of the Invention

[0004] This invention provides an early warning and control method based on time-series learning of big data on tourist flow in scenic areas, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an early warning and control method based on time-series learning of scenic area visitor flow big data, comprising: S1. Perform time-series feature analysis on the meteorological forecast data of the target scenic area during the early warning period of sudden weather events to obtain the meteorological event feature parameters of the target scenic area; S2. Based on meteorological event characteristic parameters and real-time passenger flow data and historical passenger flow database of the target scenic area, the passenger flow influx trend of indoor shelters in the target scenic area during sudden weather events is extrapolated over time to obtain the estimated total passenger flow of indoor shelters. S3. Couple the road network topology data and real-time traffic status data of the target scenic area with traffic topology to construct the traffic impedance map of the target scenic area. Based on the estimated total passenger flow and traffic impedance map, conduct congestion assessment on potential pressure road sections and entrance nodes of the target scenic area to obtain the initial congestion risk index of the target scenic area. S4. When the initial congestion risk index exceeds the preset safety threshold, based on the traffic impedance map and the historical migration reference of the target scenic area, source diversion and allocation planning is carried out for the refuge routes of tourists in the target scenic area to obtain differentiated guidance routes. S5. Encapsulate the information of the differentiated guidance path and push the encapsulation results to the tourists' mobile terminals. Compare the real-time location data fed back by the mobile terminals with the vector trajectory of the differentiated guidance path to obtain the flow convergence index corresponding to the differentiated guidance path. S6. When the crowd flow convergence index exceeds the preset convergence level, based on the extent of the exceedance of the crowd flow convergence index, the timing and frequency of the push of differentiated guidance paths are suppressed and adjusted, and the adjustment results are encoded into instructions to obtain the evacuation guidance instructions for the target scenic area.

[0006] In a preferred embodiment, the process of obtaining the meteorological event characteristic parameters of the target scenic area is as follows: Multi-temporal granularity recombination of meteorological forecast data collected in the target scenic area during the early warning period of sudden weather events yields a meteorological forecast time series raster for the target scenic area. Based on the numerical variability distribution of adjacent time phases in the meteorological forecast time series grid, the local fluctuation boundary of the meteorological forecast time series grid is identified to obtain the weather front migration trajectory line of the target scenic area. Based on the spatial extension direction and cross-temporal displacement rate of the weather front migration trajectory, the intensity evolution trend of the weather forecast time series grid is mapped to obtain the precipitation echo envelope of the target scenic area. By comparing the morphological characteristics of the precipitation echo envelope with the historical disaster-causing weather pattern database of the target scenic area, the disaster-causing level identification code of the target scenic area is obtained. The meteorological event characteristic parameters of the target scenic area are obtained by merging and compressing the disaster level identification code, the weather front migration trajectory line, and the precipitation echo envelope.

[0007] In a preferred embodiment, the process of obtaining the estimated total number of people entering the indoor refuge area is as follows: Multidimensional feature deconstruction of meteorological event characteristic parameters yields the disaster intensity level and impact time window of the target scenic area; Based on the disaster intensity level and the impact time window, the historical visitor flow database of the target scenic area is searched under the same conditions to obtain the historical inflow ratio distribution of the target scenic area. The real-time visitor flow data of the target scenic area is used to perform site proximity allocation to obtain the visitor flow attribution results for the target scenic area; Based on the historical influx ratio distribution, the influx ratio is mapped to the passenger flow attribution results to obtain the preliminary influx volume of indoor refuge areas. The initial influx of people into indoor shelters was diverted and balanced to obtain the estimated total number of people entering the indoor shelters.

[0008] In a preferred embodiment, the process of constructing the traffic impedance map of the target scenic area is as follows: Extract road network topology data and real-time traffic status data of the target scenic area; The road network topology data is decomposed into road network elements to obtain the road segment vectors and node labels of the target scenic area. Then, the topological association of the road segment vectors and node labels is constructed to obtain the topological skeleton of the target scenic area. Based on road segment vectors and node calibration, the traffic congestion level of road segments in real-time traffic status data is analyzed to obtain the real-time traffic congestion metric value of the road segments in the target scenic area. Trip assessment is performed on real-time traffic congestion metrics to obtain the estimated travel time for the road segment under the current congestion condition; The estimated travel time is loaded as edge weights into the topological skeleton to obtain the weighted skeleton map of the road segments in the target scenic area. By adding turning time delay to the intersection nodes in the weighted skeleton diagram of the road segment, the traffic impedance diagram of the target scenic area is obtained.

[0009] In a preferred embodiment, the process of obtaining the initial congestion risk index of the target scenic area is as follows: The estimated total passenger flow is decomposed into flow direction to obtain the estimated passenger load of the road segment; By comparing the estimated passenger flow load with the estimated travel time in the traffic impedance diagram, the congestion threshold is obtained to identify the over-limit judgment mark of the road segment. The road sections exceeding the limit are identified by collecting the road sections that exceed the limit, and the potential pressure-bearing road sections of the target scenic area are obtained. The entrance nodes connected to the potential pressure-bearing road sections are taken as potential pressure-bearing entrance nodes. Congestion situation assessment is conducted on potentially congested road sections to obtain the initial congestion status of these sections; The initial congestion risk is assessed based on the initial pressure state to obtain the initial congestion risk index of the target scenic area.

[0010] In a preferred embodiment, the process of obtaining the differentiated induction path is as follows: The historical migration duration data of the target scenic area is retrieved routinely to obtain the historical spatiotemporal migration reference data of the target scenic area. Based on the initial congestion risk index, the risk source is traced in the estimated travel time in the traffic impedance map to obtain the congestion bottleneck location of the target scenic area. Based on the set of congestion bottleneck locations and historical spatiotemporal migration reference data, the estimated total number of visitors to indoor refuge sites in the target scenic area is used to enumerate congestion avoidance routes, thereby obtaining a set of candidate refuge routes for the target scenic area. By comparing the impedance advantages and disadvantages of the estimated total travel time of the candidate refuge routes, differentiated guidance routes for the target scenic area are obtained.

[0011] In a preferred embodiment, the process of obtaining the differentiated guidance path for the target scenic area is as follows: The estimated travel time of each candidate evacuation route in the candidate evacuation route set is accumulated segment by segment to obtain the estimated total travel time of each candidate route. Based on the estimated total travel time, the candidate paths are checked for time-distance ratio to obtain a subset of effective guiding paths for the target scenic area. By applying asymmetric weights to the subset of effective guidance paths, differentiated guidance paths for the target scenic area can be obtained.

[0012] In a preferred embodiment, the process of obtaining the pedestrian flow convergence index corresponding to the differentiated guidance path is as follows: The differentiated guidance paths are structured and encapsulated to obtain the guidance path push package for the target scenic spot; The guidance route push package is distributed to the mobile terminal of the corresponding tourist to obtain the pushed guidance route record of the target scenic spot; Within a preset response period, the real-time location point sequence fed back by the mobile terminal is continuously collected to construct the actual movement trajectory segment of tourists in the target scenic area. Vector deviation analysis is performed on the actual movement trajectory segments of tourists and the differentiated guidance path to obtain the trajectory deviation degree of the target scenic spot; Based on the same differentiated guidance path, the directional synchronicity of the actual movement trajectory segments of tourists is measured to obtain the trajectory directional convergence component of the target scenic spot. Based on the trajectory direction convergence component, the speed synchronization of the actual movement trajectory segments of tourists is extracted to obtain the flow convergence index corresponding to the differentiated guidance path.

[0013] In a preferred embodiment, the process of obtaining the trajectory deviation of the target scenic area is as follows: The positioning points in the actual movement trajectory segment of tourists are spatially orthogonally projected onto the differentiated guidance path to obtain the orthogonal reference points of the target scenic area; Based on orthogonal reference points, the distance to each location point in the actual movement trajectory segment of tourists is measured point by point to obtain the deviation distance sequence of the target scenic spot; The deviation distance sequence is averaged to obtain the trajectory deviation of the target scenic area.

[0014] In a preferred embodiment, the process of obtaining evacuation guidance instructions for the target scenic area is as follows: The difference between the population flow convergence index and the preset convergence level is extracted to obtain the convergence exceedance value of the target scenic spot; Based on the convergence exceedance value, the push interval of the differentiated guidance path is extended and adjusted to obtain the correct push timing for the target scenic spot; Based on the convergence exceedance value, the frequency of push notifications per unit time for differentiated guidance paths is reduced to obtain the corrected push frequency parameter for the target scenic area. The parameters of correction push timing and correction push frequency are encapsulated in a binary frame structure to obtain the evacuation guidance instruction frame of the target scenic area. The evacuation guidance instruction frame audio stream is decoded to obtain the evacuation guidance instructions for the target scenic area.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. First, the meteorological forecast data during the period of the sudden weather warning in the scenic area is reorganized at a multi-temporal granularity to construct a meteorological forecast time series grid. Then, based on the numerical variability distribution of adjacent time phases in the grid, the local fluctuation boundary is identified, and the migration trajectory line of the weather front is extracted. Next, according to the spatial extension direction and cross-temporal displacement rate of the front, the precipitation echo envelope is mapped and matched with the historical disaster-causing weather pattern database to generate a disaster level identification code. Finally, this information is attribute-merged and compressed to obtain standardized meteorological event characteristic parameters. Unlike traditional methods that rely solely on meteorological thresholds at a single time step to determine risk, this processing method can accurately capture the changing patterns of frontal movement and precipitation echo patterns, making the quantification of disaster level and impact range more accurate. It can also provide more reliable data support for subsequent refined allocation of evacuation flow.

[0016] 2. Based on meteorological event characteristic parameters, the historical passenger flow database is retrieved to obtain the influx ratio distribution. Combined with the real-time passenger flow field adjacent distribution results, the influx ratio mapping is carried out, and the over-limit diversion and balancing of the preliminary influx volume is carried out to obtain the estimated total passenger flow of the indoor refuge field. This dynamic passenger flow calculation method combines meteorological disaster characteristics, real-time passenger flow and historical evacuation patterns, avoids the passenger flow estimation deviation caused by traditional static allocation, effectively solves the problem of uneven passenger flow distribution and local overload in refuge fields, and ensures the rational and efficient use of indoor refuge resources in scenic areas.

[0017] 3. Based on the construction of a traffic impedance map that integrates the travel time of road segments and the turning delay of nodes, potential pressure road segments are identified and congestion risks are assessed by combining the estimated passenger flow load. For high-risk scenarios, candidate routes to avoid congestion are planned and differentiated guidance routes are formed by configuring asymmetric weights for multiple indicators. At the same time, the flow convergence index is calculated based on the actual movement trajectory of tourists. When the convergence exceeds the standard, the timing and frequency of push notifications are suppressed and evacuation guidance instructions are generated. This dynamic planning and control mechanism solves the problem that traditional static route planning is prone to causing concentrated influx of passenger flow. It realizes the dynamic coordination between road network capacity and evacuation demand, and significantly improves the emergency evacuation efficiency and road network operation stability of scenic areas with complex terrain such as canyons. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating an embodiment of the early warning and control method based on time-series learning of scenic area visitor flow big data provided by the present invention. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] This application provides an early warning and control method based on time-series learning of scenic area visitor flow big data. The execution subject of the early warning and control method based on time-series learning of scenic area visitor flow big data includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: server, terminal, etc. In other words, the early warning and control method based on time-series learning of scenic area visitor flow big data can be executed by software or hardware installed on terminal devices or server devices. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0021] Reference Figure 1 The diagram shown is a flowchart illustrating an early warning and control method based on time-series learning of scenic area visitor flow big data, according to an embodiment of the present invention. In this embodiment, the early warning and control method based on time-series learning of scenic area visitor flow big data includes: S1. Perform time-series feature analysis on the meteorological forecast data of the target scenic area during the early warning period of sudden weather events to obtain the meteorological event feature parameters of the target scenic area; In this embodiment of the invention, the process of obtaining the meteorological event characteristic parameters of the target scenic area is as follows: Multi-temporal granularity recombination of meteorological forecast data collected in the target scenic area during the early warning period of sudden weather events yields a meteorological forecast time series raster for the target scenic area. Meteorological forecast data refers to the meteorological element data such as temperature, precipitation intensity, wind speed, and air pressure continuously collected by ground meteorological monitoring equipment during the warning period of sudden weather events in the target scenic area. Multi-temporal granular reconstruction involves splitting and splicing meteorological forecast data collected at different times according to fixed time intervals. During splitting, the continuously collected meteorological data is divided into independent time segments at the minute level. During splicing, the meteorological element data corresponding to each time segment is arranged in chronological order and mapped one by one onto the spatial geographic grid of the scenic area. Finally, a spatial grid data covering the entire scenic area and arranged in a time series is formed. The meteorological forecast time series grid is formed by closely combining the scenic area's spatial geographic grid with the time series. Each grid unit contains a complete set of time-series spatial data containing meteorological element values ​​at the corresponding time point.

[0022] Based on the numerical variability distribution of adjacent time phases in the meteorological forecast time series grid, the local fluctuation boundary of the meteorological forecast time series grid is identified to obtain the weather front migration trajectory line of the target scenic area. Numerical variability distribution first calculates the change range of meteorological element values ​​between two adjacent time points in each grid cell of the weather forecast time series raster. The calculation of numerical variability follows C=A−B, where A represents the meteorological element value of the later time phase, B represents the meteorological element value of the previous time phase, and C represents the numerical variability. The preset critical range is determined based on the historical average of meteorological fluctuations in the target scenic area. Grid cells with calculated change ranges exceeding the preset critical range are marked as fluctuation cells. All marked fluctuation cells are connected sequentially according to their spatial proximity to form a continuous fluctuation boundary. Local fluctuation boundary identification involves traversing all grid cells of the weather forecast time series raster, calculating the change range of meteorological element values ​​between adjacent time phases for each grid cell, filtering out all fluctuation cells that meet the conditions, and delineating a continuous boundary according to their spatial location. The weather front migration trajectory line is formed by connecting the fluctuation boundaries determined at different time points in chronological order, which can completely reflect the continuous line data of the weather front's movement path within the scenic area.

[0023] Based on the spatial extension direction and cross-temporal displacement rate of the weather front migration trajectory, the intensity evolution trend of the weather forecast time series grid is mapped to obtain the precipitation echo envelope of the target scenic area. The spatial extension direction is the frontal extension angle determined by connecting the starting and ending coordinates of the weather front's migration trajectory. The calculation of the cross-phase displacement rate follows V=D÷T, where D represents the distance the weather front moves, T represents the unit time, and V represents the cross-phase displacement rate. The preset intensity standard is determined based on the disaster-causing precipitation intensity threshold of the target scenic area. The intensity evolution trend mapping is based on the determined spatial extension direction and the calculated cross-phase displacement rate, which completely projects the changes in precipitation intensity values ​​over time in the meteorological forecast time series grid onto the entire spatial area covered by the front. The precipitation echo envelope is formed by outlining the entire spatial area where the precipitation intensity reaches the preset intensity standard, and it is a closed spatial contour data that can accurately delineate the range of precipitation concentration areas within the scenic area.

[0024] By comparing the morphological characteristics of the precipitation echo envelope with the historical disaster-causing weather pattern database of the target scenic area, the disaster-causing level identification code of the target scenic area is obtained. The historical disaster-causing weather pattern database is a collection of morphological characteristics data, including the contours, spatial ranges, and intensity distributions of precipitation echoes corresponding to all past disaster-causing weather events in the target scenic area. These data are categorized and stored according to their disaster severity level from low to high. The morphological feature comparison involves matching the contour shape, spatial coverage area, and intensity distribution characteristics of the current precipitation echo envelope with the morphological feature data corresponding to each level in the historical disaster-causing weather pattern database one by one, and selecting the historical disaster-causing weather pattern with the highest matching degree. The disaster severity level identification code is an identifier data used to uniquely identify the degree of disaster caused by the current weather, generated according to a fixed coding rule based on the disaster severity level corresponding to the matched historical disaster-causing weather pattern.

[0025] The disaster level identification code, weather front migration trajectory line and precipitation echo envelope are merged and compressed to obtain the meteorological event characteristic parameters of the target scenic area. Attribute merging integrates the identification attributes of disaster severity level identifiers, the spatial path attributes of weather front migration trajectories, and the spatial contour attributes of precipitation echo envelopes into a unified attribute set in a fixed order. Compression removes duplicate and redundant attribute fields from the integrated attribute set, unifies the data storage and transmission format, and simplifies the data processing flow. Meteorological event feature parameters are a standardized data set that integrates all core meteorological features such as disaster severity level, front migration path, and precipitation echo range.

[0026] S2. Based on meteorological event characteristic parameters and real-time passenger flow data and historical passenger flow database of the target scenic area, the passenger flow influx trend of indoor shelters in the target scenic area during sudden weather events is extrapolated over time to obtain the estimated total passenger flow of indoor shelters. In this embodiment of the invention, the process of obtaining the estimated total passenger flow of an indoor refuge site is as follows: Multidimensional feature deconstruction of meteorological event characteristic parameters yields the disaster intensity level and impact time window of the target scenic area; Meteorological event characteristic parameters are standardized data obtained through time-series analysis and feature integration of meteorological data, containing all core information such as the disaster severity level of sudden weather, the movement path of weather fronts, and the coverage area of ​​precipitation echoes. Multidimensional feature deconstruction involves reading all attribute information within the meteorological event characteristic parameters line by line, extracting and solidifying the level attribute representing the magnitude of disaster destructive power into a disaster severity level, and continuously splicing the time attribute representing the duration of the disaster and the start and end periods of its effect into a complete time interval, which is the impact time window. The disaster severity level is a fixed level divided according to the degree of damage to the safety of tourists and facilities in the scenic area. The impact time window is the complete time range from the beginning to the end of the impact of sudden weather on the scenic area.

[0027] Based on the disaster intensity level and the impact time window, the historical visitor flow database of the target scenic area is searched under the same conditions to obtain the historical inflow ratio distribution of the target scenic area. The historical visitor flow database is a collection of data collected in advance, including the number of visitors and their proportions at various indoor shelters under different disaster intensity levels and impact time windows during all past sudden weather scenarios at the target scenic area. This data is permanently stored after being categorized and organized. The same-condition influx retrieval method retains only historical data in the historical visitor flow database that is consistent with the current disaster intensity level and impact time window duration. After removing other mismatched data, the total number of evacuees received by all indoor shelters under the historical scenario is first counted, and then the number of evacuees received by a single indoor shelter under the same scenario is counted. The influx ratio of a single indoor shelter is obtained by dividing the historical influx of visitors to a single indoor shelter by the total number of evacuees to all indoor shelters during the same period. The influx ratio of a single indoor shelter follows the formula R=N÷M, where R represents the influx ratio of a single indoor shelter, N represents the historical influx of visitors to that shelter, and M represents the total number of evacuees during the same period. The influx ratios of all indoor shelters are arranged in a fixed site number order to form a complete historical influx ratio distribution.

[0028] The real-time visitor flow data of the target scenic area is used to perform site proximity allocation to obtain the visitor flow attribution results for the target scenic area; Real-time visitor flow data is collected in real time by visitor flow monitoring terminals deployed throughout the target scenic area. This data includes the spatial location of each visitor, the total number of visitors in each area of ​​the scenic area, and the spatial distribution density of visitors. Proximity allocation involves pre-delineating a dedicated service coverage area for each indoor shelter according to the scenic area's geographical layout. This area is centered on the indoor shelter and bounded by the shortest walking distance for each visitor. Each visitor's real-time location is compared with the service coverage area of ​​each indoor shelter, assigning each visitor to the corresponding indoor shelter within their designated coverage area. The number of visitors within the service coverage area of ​​each indoor shelter is then cumulatively counted. The cumulative results for all indoor shelters are then summarized to obtain the visitor flow allocation result. This result provides complete data recording the total number of visitors in the real-time service area corresponding to each indoor shelter.

[0029] Based on the historical influx ratio distribution, the influx ratio is mapped to the passenger flow attribution results to obtain the preliminary influx volume of indoor refuge areas. The inflow ratio mapping involves retrieving the total number of visitors in the service area of ​​each indoor shelter in real time, then retrieving the corresponding inflow ratio of that indoor shelter in the historical inflow ratio distribution, multiplying the total number of visitors in the current shelter by its own historical inflow ratio, and obtaining the theoretical inflow value of the indoor shelter without considering the site capacity limit. All the theoretical inflow values ​​of all indoor shelters are summed up to obtain the preliminary inflow volume of the indoor shelter. The preliminary inflow volume is the indoor shelter passenger flow value calculated by combining historical patterns and real-time passenger flow without correcting for site carrying capacity limit.

[0030] The initial influx of people into indoor shelters was diverted and balanced to obtain the estimated total number of people entering the indoor shelters. Overload diversion and balancing involves determining the maximum number of visitors a site can accommodate in advance based on its building area, safety evacuation standards, and facility configuration. The initial influx of visitors to each indoor refuge site is compared to its maximum capacity. If the initial influx is less than or equal to the maximum capacity, it is used as the final estimated visitor flow for that site. If the initial influx exceeds the maximum capacity, the maximum capacity is subtracted from the initial influx to obtain the excess visitor flow. Then, all surrounding indoor refuge sites that have not exceeded their maximum capacity are analyzed. For the refuge areas, calculate the available capacity of each surrounding site, add up the available capacity of all surrounding sites to get the total available capacity, then divide the available capacity of each surrounding site by the total available capacity to get the diversion ratio of each site, and finally multiply the over-limit passenger flow value by the diversion ratio of each surrounding site to get the number of diverted passengers that the surrounding site should be able to handle. Add the number of diverted passengers to the original estimated passenger flow of the surrounding sites, and fix the estimated passenger flow of the over-limit sites to the maximum capacity. Sum up the adjusted passenger flow values ​​of all indoor refuge areas to get the total estimated passenger flow of the target scenic area's indoor refuge areas.

[0031] S3. Couple the road network topology data and real-time traffic status data of the target scenic area with traffic topology to construct the traffic impedance map of the target scenic area. Based on the estimated total passenger flow and traffic impedance map, conduct congestion assessment on potential pressure road sections and entrance nodes of the target scenic area to obtain the initial congestion risk index of the target scenic area. In this embodiment of the invention, the process of obtaining the initial congestion risk index of the target scenic area is as follows: Extract road network topology data and real-time traffic status data of the target scenic area; The road network topology data is the basic road data of the entire area, which is stored in a dedicated geographic information database after the scenic area completed the full-area road survey using professional surveying equipment. It includes the spatial location of all roads, road connection relationships, road physical length, road width, and precise coordinates of intersections. The real-time traffic status data is the real-time perception data collected by pedestrian monitoring cameras and infrared passenger flow sensors deployed on each road section. It includes the number of pedestrians per unit area, the actual movement speed of pedestrians, and whether there are any pedestrians lingering. Extraction involves retrieving all road network topology data completely from the scenic area's geographic information database and simultaneously retrieving all real-time traffic status data from the real-time monitoring terminal. The integrity of both types of data is checked one by one. Invalid data with missing spatial coordinates, no valid traffic status, and duplicate collection are removed. All valid and complete data are retained for subsequent processing.

[0032] The road network topology data is decomposed into road network elements to obtain the road segment vectors and node labels of the target scenic area. Then, the topological association of the road segment vectors and node labels is constructed to obtain the topological skeleton of the target scenic area. Road network element decomposition involves breaking down the complete road network topology data into indivisible single road units and node units formed by road intersections, according to the independent segmentation rules of the physical roads in the scenic area. Road segment vectors provide a complete spatial digital description of each single road unit, recording linear spatial data such as the spatial coordinates of the road's starting point, ending point, physical length, and orientation angle. Node labeling assigns a unique spatial location number and fixed coordinate mark to each road intersection node and the starting and ending points of the road segment, ensuring that each node has a unique and identifiable identifier within the scenic area. Topology association construction precisely connects the starting and ending points of each road segment vector with the corresponding numbered node labels according to the actual road connection relationships recorded in the road network topology data, so that all road segment vectors and node labels form an interconnected overall network structure, which is the topology skeleton.

[0033] Based on road segment vectors and node calibration, the traffic congestion level of road segments in real-time traffic status data is analyzed to obtain the real-time traffic congestion metric value of the road segments in the target scenic area. Traffic status analysis first precisely delineates the physical coverage area of ​​each individual road based on the spatial coverage range corresponding to the road segment vector. Then, it matches the pedestrian number and movement speed information collected from the real-time traffic status data to the physical coverage area of ​​the corresponding road segment. A fixed quantitative classification standard for the congestion level of the road segment is pre-set. The state where the number of pedestrians per unit area is less than the minimum standard and the movement speed is higher than the fastest standard is recorded as low congestion level. The state where the number of pedestrians per unit area is at the middle standard and the movement speed is at the medium standard is recorded as medium congestion level. The state where the number of pedestrians per unit area is higher than the highest standard and the movement speed is lower than the slowest standard is recorded as high congestion level. The three congestion levels are converted into continuously increasing quantitative values ​​in sequence. This value is the real-time traffic congestion metric, which is the only quantitative data that represents the current pedestrian congestion status of a single road segment.

[0034] Trip assessment is performed on real-time traffic congestion metrics to obtain the estimated travel time for the road segment under the current congestion condition; Trip assessment first determines the average pedestrian speed corresponding to the real-time traffic congestion metric. Low congestion corresponds to a fixed, relatively fast average pedestrian speed, medium congestion corresponds to a fixed, moderate average pedestrian speed, and high congestion corresponds to a fixed, relatively slow average pedestrian speed. Then, combined with the physical length of the road recorded in the road segment vector, the time required for a pedestrian to walk from the start point to the end point of the road segment is calculated. This time is the estimated travel time. The estimated travel time is calculated according to T=L÷V, where T represents the estimated travel time, L represents the physical length of the road segment, and V represents the average pedestrian speed under the current congestion state.

[0035] The estimated travel time is loaded as edge weights into the topological skeleton to obtain the weighted skeleton map of the road segments in the target scenic area. Edge weight loading involves using the estimated travel time calculated for each road segment as the exclusive weight information for that road segment, and adding it one by one to the road segment vectors with the same number in the topology skeleton. This ensures that each road segment vector in the topology skeleton carries its own estimated travel time weight. The road network with travel time weights formed after all road segments have been weighted is the road segment weighted skeleton diagram.

[0036] Add a turning time delay to the intersection nodes in the weighted skeleton map of the road segment to obtain the traffic impedance map of the target scenic area; The turning delay addition is based on the actual pedestrian traffic conditions at scenic area intersections. It determines the pedestrian waiting time for different turning directions at each intersection node. There is no waiting time for straight-ahead directions, while left-turn and right-turn directions have fixed waiting times set according to the intersection width and pedestrian traffic rules. The turning waiting time value corresponding to each intersection node is accurately added to the intersection node with the same number in the road segment weighting skeleton map. This makes the road network include both the estimated travel time of the road segment and the turning waiting time of the node. The complete travel time network after integration is the travel impedance map.

[0037] The estimated total passenger flow is decomposed into flow direction to obtain the estimated passenger load of the road segment; The site flow decomposition first determines the necessary routes for all tourists to reach each indoor shelter. Then, the estimated total passenger flow for each indoor shelter is allocated segment by segment according to the order in which tourists travel from their current location to the shelter. All passenger flow allocated to each necessary route segment is summed up to obtain the estimated passenger flow load for that segment. The estimated passenger flow load is the total number of evacuees that a single route segment will need to accommodate in the future.

[0038] By comparing the estimated passenger flow load with the estimated travel time in the traffic impedance diagram, the congestion threshold is obtained to identify the over-limit judgment mark of the road segment. The congestion threshold comparison is based on the design capacity standards of scenic area roads. A threshold for passenger flow capacity and a threshold for travel time are set for each road segment. The estimated passenger flow load of the road segment is compared with the threshold, and the estimated travel time is also compared with the threshold. If both the estimated passenger flow load and the estimated travel time are greater than the threshold, the road segment is considered to be in an over-limit state. If either the estimated passenger flow load or the estimated travel time is less than or equal to the threshold, or the estimated travel time is less than or equal to the threshold, the road segment is considered not to be in an over-limit state. This clear determination result is the over-limit identification indicator.

[0039] The road sections exceeding the limit are identified by collecting the road sections that exceed the limit, and the potential pressure-bearing road sections of the target scenic area are obtained. The entrance nodes connected to the potential pressure-bearing road sections are taken as potential pressure-bearing entrance nodes. The over-limit road section collection involves traversing all road sections in the scenic area to identify over-limit indicators, collecting and organizing all road sections judged to be in an over-limit state to form a unified road section set. This road section set is the potential pressure road section. Then, each potential pressure road section is directly connected to the scenic area entrance, passage entrance, and area connection entrance node. All these directly connected entrance nodes are collected to form a fixed node set. This node set is the potential pressure entrance node.

[0040] Congestion situation assessment is conducted on potentially congested road sections to obtain the initial congestion status of these sections; Congestion assessment involves extracting the estimated passenger load and estimated travel time for each potential congestion-prone road segment. Fixed congestion score calculation rules are pre-defined: a fixed score is added for each unit exceeding the passenger load threshold, and a fixed score is added for each unit exceeding the estimated travel time threshold. The two scores for each road segment are added together to obtain the single-segment congestion score. Then, the congestion scores for all single road segments in the potential congestion-prone road segments are summed up. The total score obtained is the initial congestion state, which is a quantitative value representing the overall congestion level of the potential congestion-prone road segments in the scenic area.

[0041] The initial congestion risk is assessed based on the initial pressure state to obtain the initial congestion risk index of the target scenic area; Congestion risk rating involves pre-setting a fixed congestion risk level range corresponding to the initial pressure state, accurately matching the calculated initial pressure state with the risk level range, determining the unique risk level to which the initial pressure state belongs, and then converting the risk level into a standardized continuous index value. This index value is the initial congestion risk index of the target scenic area. The initial congestion risk index is the only standardized data that represents the overall initial congestion risk level of the target scenic area.

[0042] S4. When the initial congestion risk index exceeds the preset safety threshold, based on the traffic impedance map and the historical migration reference of the target scenic area, source diversion and allocation planning is carried out for the refuge routes of tourists in the target scenic area to obtain differentiated guidance routes. In this embodiment of the invention, the process of obtaining the differentiated induction path is as follows: The historical migration duration data of the target scenic area is retrieved routinely to obtain the historical spatiotemporal migration reference data of the target scenic area. Historical migration duration data comprises all historical records of actual walking time, route selection, and inter-regional migration time of tourists from each departure area to each indoor shelter within the scenic area under normal tourist conditions without sudden weather interference, road congestion, or visitor flow control. This data is continuously collected and stored by the scenic area's comprehensive visitor flow monitoring equipment. Normal migration time retrieval involves filtering from the scenic area's dedicated historical visitor flow database, retaining only valid migration duration data under normal conditions, and eliminating invalid entries with walking times exceeding the normal range, missing route information, or incorrect data collection. All valid migration duration values ​​from the same departure area to the same indoor shelter are summed, and the total sum is divided by the total number of valid data entries in that group to obtain the average normal migration time from that area to that shelter. Based on the correspondence between the scenic area's spatial divisions and the indoor shelter numbers, the average normal migration times from all areas to all shelters are arranged in an orderly manner, forming a complete spatiotemporal data set including spatial location, migration duration, and corresponding shelter, which constitutes the historical spatiotemporal migration reference.

[0043] Based on the initial congestion risk index, the risk source is traced in the estimated travel time in the traffic impedance map to obtain the congestion bottleneck location of the target scenic area. The initial congestion risk index is a standardized quantitative value representing the overall road congestion risk level of the target scenic area, obtained through prior congestion assessment. The traffic impedance map is a complete road traffic time network data including the estimated travel time for all road segments in the scenic area and the turning delays for all intersections. Risk source tracing involves dividing the initial congestion risk index by the total number of road segments within the scenic area to obtain the basic score for local congestion risk for each road segment. The basic score for local congestion risk for each road segment is then multiplied by the estimated travel time for that road segment in the traffic impedance map to obtain the maximum risk level for that road segment. The final local congestion risk score is calculated according to the formula F=J×T, where F represents the local congestion risk score, J represents the basic local congestion risk score, and T represents the estimated travel time of the road segment. The highest critical standard for the local congestion risk score is set in advance based on the historical peak congestion of roads in the scenic area. All road segments with local congestion risk scores reaching the highest critical standard, as well as the intersection nodes directly connected to the road segments, are screened out. The precise spatial coordinates of these road segments and nodes are summarized, and the resulting set of spatial locations is the congestion bottleneck location.

[0044] Based on the set of congestion bottleneck locations and historical spatiotemporal migration reference data, the estimated total number of visitors to indoor refuge sites in the target scenic area is used to enumerate congestion avoidance routes, thereby obtaining a set of candidate refuge routes for the target scenic area. The congestion bottleneck location set is the precise spatial coordinate set of all high-congestion-risk road sections and intersections within the scenic area; the historical spatiotemporal migration reference is the average migration time data from each area to each refuge site under normal conditions; the congestion avoidance path enumeration is based on the tourist departure area corresponding to the estimated total tourist flow for each indoor refuge site, with the hard screening rule that the tourist walking path does not pass through any road section or intersection in the congestion bottleneck location set; based on the actual road connectivity relationship of the scenic area's road network topology, starting from the tourist departure area, the normally passable road sections and nodes are connected in sequence, and all independent walking paths that can safely reach the corresponding indoor refuge site are sorted out one by one; all independent walking paths that meet the congestion avoidance rules are uniformly summarized to form a complete path set, which is the candidate refuge path set.

[0045] The process of comparing the impedance advantages and disadvantages of the estimated total travel time of candidate refuge routes to obtain differentiated guidance routes for the target scenic area is as follows: The estimated travel time of each candidate evacuation route in the candidate evacuation route set is accumulated segment by segment to obtain the estimated total travel time of each candidate route. The segment-by-segment accumulation method first identifies all independent road segments included in each candidate path within the candidate refuge path set. Then, it sequentially extracts the estimated travel time for each independent road segment from the traffic impedance diagram. The extracted estimated travel times for all road segments are then accumulated sequentially according to the travel order of the path to obtain the estimated total travel time for the entire candidate path from start to finish. The calculation of the estimated total travel time follows the formula Z = A1 + A2 + ... + A n In the formula, Z represents the estimated total travel time, and A1 to A n This represents the estimated travel time for each independent segment of the candidate route.

[0046] Based on the estimated total travel time, the candidate paths are checked for time-distance ratio to obtain a subset of effective guiding paths for the target scenic area. The time-distance ratio verification involves measuring the actual total physical length of each candidate path using professional geographic surveying data from the scenic area. The estimated total travel time for each candidate path is divided by its actual total physical length to obtain the time-distance ratio. A fixed upper limit for the acceptable ratio is pre-set based on the typical walking efficiency of tourists in the scenic area. The time-distance ratio of each candidate path is then compared with this upper limit. Candidate paths with a time-distance ratio less than or equal to the upper limit are considered valid paths that meet the travel efficiency requirements. All candidate paths deemed valid are then aggregated to form the effective directional path subset.

[0047] Asymmetric weighting is applied to a subset of effective guidance paths to obtain differentiated guidance paths for the target scenic area; The asymmetric weighting configuration first calculates four core indicators for each path in the effective guidance path subset: estimated total travel time, time-to-distance ratio, path physical width, and path maximum passenger capacity. A fixed quantitative scoring standard is set for each indicator: a shorter estimated total travel time corresponds to a higher score; a smaller time-to-distance ratio corresponds to a higher score; a wider path physical width corresponds to a higher score; and a larger path maximum passenger capacity corresponds to a higher score. The scores of the four indicators for each path are summed to obtain the overall score for that path. The sum of the overall scores of all paths in the effective guidance path subset is calculated, and the overall score of a single path is divided by the sum of the overall scores of all paths to obtain the passenger flow allocation weight for that path. Paths are matched according to their weight values ​​from largest to smallest, and all effective paths carrying independent passenger flow allocation weights are integrated to form the differentiated guidance path for the target scenic area.

[0048] S5. Encapsulate the information of the differentiated guidance path and push the encapsulation results to the tourists' mobile terminals. Compare the real-time location data fed back by the mobile terminals with the vector trajectory of the differentiated guidance path to obtain the flow convergence index corresponding to the differentiated guidance path. In this embodiment of the invention, the process of obtaining the crowd flow convergence index corresponding to the differentiated guidance path is as follows: The differentiated guidance paths are structured and encapsulated to obtain the guidance path push package for the target scenic spot; Differentiated guidance paths are multi-branch congestion avoidance and disaster mitigation paths planned by the target scenic area for different tourist groups. They include a sequence of spatial coordinates of the path, the number of the road segments, the coordinates of the turning nodes, the estimated travel time, and the weight of the passenger flow allocation. Structured encapsulation involves classifying and sorting all the information in the differentiated guidance path according to the fixed data format received by the mobile terminal, deleting duplicate coordinate and node information, unifying the numerical precision of latitude and longitude coordinates and the encoding format of road segment nodes, and integrating all the sorted effective information into an independent data packet that can be transmitted through a wireless network. This independent data packet is the guidance path push packet for the target scenic area.

[0049] The guidance route push package is distributed to the mobile terminal of the corresponding tourist to obtain the pushed guidance route record of the target scenic spot; Targeted distribution involves matching the real-time location of tourists with the differentiated guidance paths assigned by the system, sending guidance path push packages to the corresponding tourists' mobile terminals, and simultaneously recording the path number, tourist identification, precise push time, and receiving terminal device number of each push. All push-related information is summarized and stored in chronological order to form a complete push data record, which is the push guidance path record of the target scenic spot.

[0050] Within a preset response period, the real-time location point sequence fed back by the mobile terminal is continuously collected to construct the actual movement trajectory segment of tourists in the target scenic area. The preset response period is a fixed observation time length set in advance after tourists receive the guided path and begin to move along the path. The real-time positioning point sequence is the real-time latitude and longitude coordinate data of tourists automatically collected and transmitted by the mobile terminal at fixed time intervals. Continuous collection means that all positioning coordinate data transmitted by tourists are received uninterruptedly throughout the entire preset response period. The discrete coordinate points are arranged in sequence according to the time sequence of coordinate collection. Based on the correspondence between time and spatial coordinates, the discrete coordinate points are connected into a continuous spatial movement line. This continuous spatial movement line is the actual movement trajectory segment of tourists in the target scenic area.

[0051] The process of performing vector deviation analysis on the actual movement trajectory segment of tourists and the differentiated guidance path to obtain the trajectory deviation degree of the target scenic spot is as follows: The positioning points in the actual movement trajectory segment of tourists are spatially orthogonally projected onto the differentiated guidance path to obtain the orthogonal reference points of the target scenic area; Spatial orthogonal projection involves projecting each real-time location coordinate point in the tourist's actual movement trajectory segment vertically onto the central axis of the corresponding differentiated guidance path. The coordinate point that falls precisely on the central axis of the differentiated guidance path after projection is the orthogonal reference point of the target scenic area.

[0052] Based on orthogonal reference points, the distance to each location point in the actual movement trajectory segment of tourists is measured point by point to obtain the deviation distance sequence of the target scenic spot; Point-to-point distance measurement calculates the straight-line distance between each tourist's real-time location coordinates and the corresponding orthogonal reference point according to the spatial straight-line distance calculation rules. All the calculated distance values ​​are arranged in the order of time when the location points were collected, and the resulting continuous set of distance values ​​is the deviation distance sequence of the target scenic area.

[0053] The deviation distance sequence is averaged to obtain the trajectory deviation of the target scenic area; The mean aggregation method first adds up the deviation distance values ​​corresponding to all positioning points in the deviation distance sequence to obtain the total deviation distance value of all positioning points. Then, the total deviation distance value is divided by the total number of positioning points in the deviation distance sequence to obtain the average deviation distance value, which is the trajectory deviation of the target scenic area. The trajectory deviation is calculated according to P=(S1+S2+…+S…). n ) / n, where P is the trajectory deviation, S1, S2, ..., S nThese represent the deviation distance values ​​corresponding to the 1st to nth real-time positioning points in the deviation distance sequence, where n is the total number of positioning points in the deviation distance sequence. The calculation logic of this formula is as follows: first, all individual deviation distance values ​​in the deviation distance sequence are summed to obtain the total deviation distance value; then, the total deviation distance value is used as the dividend, and the total number of positioning points in the deviation distance sequence is used as the divisor for division. The final average deviation distance value is the trajectory deviation.

[0054] Based on the same differentiated guidance path, the directional synchronicity of the actual movement trajectory segments of tourists is measured to obtain the trajectory directional convergence component of the target scenic spot. The directional synchronicity measurement first extracts the real-time movement direction angle values ​​of all tourists' actual movement trajectory segments under the same differentiated guidance path. Then, it calculates the difference between the movement direction angle value of each tourist and the standard direction angle value of the differentiated guidance path. All angle difference values ​​are added together to obtain the total angle difference value. The total angle difference value is divided by the total number of tourists under the path to obtain the average angle difference value. The smaller the average angle difference value, the higher the synchronicity of tourists' movement direction. The average angle difference value is standardized and transformed into a range of zero to one. The transformed value is the trajectory direction convergence component of the target scenic spot.

[0055] Based on the trajectory direction convergence component, the speed synchronization of the actual movement trajectory segment of tourists is extracted to obtain the flow convergence index corresponding to the differentiated guidance path. Speed ​​synchronization extraction first extracts the real-time movement speed values ​​of all actual movement trajectory segments under the same differentiated guidance path. Then, it calculates the difference between each tourist's movement speed value and the standard movement speed value of the differentiated guidance path. All speed difference values ​​are summed to obtain the total speed difference value. The total speed difference value is divided by the total number of tourists under that path to obtain the average speed difference value. The smaller the average speed difference value, the higher the synchronization of tourist movement speed. The speed synchronization value is multiplied by the trajectory direction convergence component. The resulting comprehensive value is the flow convergence index corresponding to the differentiated guidance path. The flow convergence index is calculated according to Q=K×H, where Q is the flow convergence index corresponding to the differentiated guidance path; K is the trajectory direction convergence component; and H is the speed synchronization value. The calculation logic of this formula is to use the trajectory direction convergence component value as the first multiplier and the speed synchronization value as the second multiplier, and perform multiplication. By integrating the synchronization characteristics of tourist movement direction and speed through multiplication, the final value obtained is the flow convergence index representing the overall degree of flow convergence.

[0056] S6. When the crowd flow convergence index exceeds the preset convergence level, based on the extent of the crowd flow convergence index exceeding the standard, the timing and frequency of the push of differentiated guidance paths are suppressed and adjusted, and the adjustment results are encoded into instructions to obtain the evacuation guidance instructions for the target scenic area. In this embodiment of the invention, the process of obtaining evacuation guidance instructions for the target scenic area is as follows: The difference between the population flow convergence index and the preset convergence level is extracted to obtain the convergence exceedance value of the target scenic spot; The crowd flow convergence index is a standardized quantitative value obtained through prior trajectory analysis, representing the degree of uniformity in the direction and speed of tourist movement under the same differentiated guidance path. The preset convergence level is a fixed critical value set in advance according to the scenic area's passenger flow evacuation safety control standards, used to determine whether the crowd flow convergence state exceeds the safe range. The difference extraction is to use the crowd flow convergence index as the minuend and the preset convergence level as the subtrahend, and directly perform a subtraction operation to obtain the difference between the two values. This difference is the convergence exceedance value of the target scenic area. Its calculation logic is to subtract the preset safe convergence critical value from the actual monitored crowd flow convergence index value, and finally obtain the specific quantitative difference value of the crowd flow convergence exceeding the safety standard.

[0057] Based on the convergence exceedance value, the push interval of the differentiated guidance path is extended and adjusted to obtain the correct push timing for the target scenic spot; The convergence exceedance value reflects the extent to which the convergence index of people exceeds the preset safety threshold. The push interval is the original time interval for the differentiated guidance path to send information to tourists' mobile terminals. The extension adjustment is a pre-established positive adjustment rule between the convergence exceedance value and the push interval. For every fixed unit increase in the convergence exceedance value, the push interval is extended by a fixed duration. The original push interval is added to the extension duration calculated according to the rule to obtain the adjusted final push interval. The specific push time point corresponding to this final push interval is the corrected push timing for the target scenic spot.

[0058] Based on the convergence exceedance value, the frequency of push notifications per unit time for differentiated guidance paths is reduced to obtain the corrected push frequency parameter for the target scenic area. The convergence exceedance value is a core quantitative value reflecting the degree of convergence of crowd flow. The number of push notifications per unit time is the original number of times the differentiated guidance path sends information to tourists' mobile terminals per unit time. The frequency reduction adjustment is a pre-established inverse adjustment rule between the convergence exceedance value and the number of push notifications per unit time. For every fixed increase in the convergence exceedance value, the number of push notifications per unit time is reduced by a fixed number. The final push notification number is obtained by subtracting the reduction calculated according to the rule from the original number of push notifications per unit time. This final push notification number is the corrected push notification frequency parameter for the target scenic spot. Its calculation logic is to subtract the number of push notifications that need to be reduced based on the convergence exceedance value from the original number of push notifications per unit time of the path, and finally obtain the corrected push notification frequency value adapted to on-site management.

[0059] The parameters of correction push timing and correction push frequency are encapsulated in a binary frame structure to obtain the evacuation guidance instruction frame of the target scenic area. The corrected push timing is the optimal push time point of the differentiated guidance path obtained after extension adjustment. The corrected push frequency parameter is the optimal number of pushes per unit time of the differentiated guidance path obtained after frequency reduction adjustment. The binary frame structure encapsulation is based on the fixed binary data frame format commonly used by scenic area management equipment. The corrected push timing value is written into the time parameter field of the data frame, and the corrected push frequency parameter value is written into the frequency parameter field of the data frame. This completes the ordered encoding and integration of the two types of data, forming a complete data frame that can be transmitted and identified. This complete data frame is the evacuation guidance instruction frame of the target scenic area.

[0060] The audio bitstream of the evacuation guidance instruction frame is decoded to obtain the evacuation guidance instruction for the target scenic area; The evacuation guidance instruction frame is a binary encoded data carrier containing parameters for correcting the timing and frequency of the push. Audio stream decoding converts the binary encoded data of the evacuation guidance instruction frame into audio encoded signals one by one, and then converts the audio encoded signals into clear and playable voice text content. This voice text content can be directly used by the scenic area's broadcasting equipment to perform evacuation guidance operations, which is the evacuation guidance instruction for the target scenic area.

[0061] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0062] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0063] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for early warning and control based on time-series learning of big data on tourist flow in scenic areas, characterized in that the method... include: S1. Perform time-series feature analysis on the meteorological forecast data of the target scenic area during the early warning period of sudden weather events to obtain the meteorological event feature parameters of the target scenic area; S2. Based on meteorological event characteristic parameters and real-time passenger flow data and historical passenger flow database of the target scenic area, the passenger flow influx trend of indoor shelters in the target scenic area during sudden weather events is extrapolated over time to obtain the estimated total passenger flow of indoor shelters. S3. Couple the road network topology data and real-time traffic status data of the target scenic area with traffic topology to construct the traffic impedance map of the target scenic area. Based on the estimated total passenger flow and traffic impedance map, conduct congestion assessment on potential pressure road sections and entrance nodes of the target scenic area to obtain the initial congestion risk index of the target scenic area. S4. When the initial congestion risk index exceeds the preset safety threshold, based on the traffic impedance map and the historical migration reference of the target scenic area, source diversion and allocation planning is carried out for the refuge routes of tourists in the target scenic area to obtain differentiated guidance routes. S5. Encapsulate the information of the differentiated guidance path and push the encapsulation results to the tourists' mobile terminals. Compare the real-time location data fed back by the mobile terminals with the vector trajectory of the differentiated guidance path to obtain the flow convergence index corresponding to the differentiated guidance path. S6. When the crowd flow convergence index exceeds the preset convergence level, based on the extent of the exceedance of the crowd flow convergence index, the timing and frequency of the push of differentiated guidance paths are suppressed and adjusted, and the adjustment results are encoded into instructions to obtain the evacuation guidance instructions for the target scenic area.

2. The early warning and control method based on time-series learning of scenic area visitor flow big data as described in claim 1, characterized in that, The process of obtaining the meteorological event characteristic parameters of the target scenic area is as follows: Multi-temporal granularity recombination of meteorological forecast data collected in the target scenic area during the early warning period of sudden weather events yields a meteorological forecast time series raster for the target scenic area. Based on the numerical variability distribution of adjacent time phases in the meteorological forecast time series grid, the local fluctuation boundary of the meteorological forecast time series grid is identified to obtain the weather front migration trajectory line of the target scenic area. Based on the spatial extension direction and cross-temporal displacement rate of the weather front migration trajectory, the intensity evolution trend of the weather forecast time series grid is mapped to obtain the precipitation echo envelope of the target scenic area. By comparing the morphological characteristics of the precipitation echo envelope with the historical disaster-causing weather pattern database of the target scenic area, the disaster-causing level identification code of the target scenic area is obtained. The meteorological event characteristic parameters of the target scenic area are obtained by merging and compressing the disaster level identification code, the weather front migration trajectory line, and the precipitation echo envelope.

3. The early warning and control method based on time-series learning of scenic area visitor flow big data as described in claim 1, characterized in that, The process of obtaining the estimated total number of visitors to indoor shelters is as follows: Multidimensional feature deconstruction of meteorological event characteristic parameters yields the disaster intensity level and impact time window of the target scenic area; Based on the disaster intensity level and the impact time window, the historical visitor flow database of the target scenic area is searched under the same conditions to obtain the historical inflow ratio distribution of the target scenic area. The real-time visitor flow data of the target scenic area is used to perform site proximity allocation to obtain the visitor flow attribution results for the target scenic area; Based on the historical influx ratio distribution, the influx ratio is mapped to the passenger flow attribution results to obtain the preliminary influx volume of indoor refuge areas. The initial influx of people into indoor shelters was diverted and balanced to obtain the estimated total number of people entering the indoor shelters.

4. The early warning and control method based on time-series learning of scenic area visitor flow big data as described in claim 1, characterized in that, The process of constructing the traffic impedance map of the target scenic area is as follows: Extract road network topology data and real-time traffic status data of the target scenic area; The road network topology data is decomposed into road network elements to obtain the road segment vectors and node labels of the target scenic area. Then, the topological association of the road segment vectors and node labels is constructed to obtain the topological skeleton of the target scenic area. Based on road segment vectors and node calibration, the traffic congestion level of road segments in real-time traffic status data is analyzed to obtain the real-time traffic congestion metric value of the road segments in the target scenic area. Trip assessment is performed on real-time traffic congestion metrics to obtain the estimated travel time for the road segment under the current congestion condition; The estimated travel time is loaded as edge weights into the topological skeleton to obtain the weighted skeleton map of the road segments in the target scenic area. By adding turning time delay to the intersection nodes in the weighted skeleton diagram of the road segment, the traffic impedance diagram of the target scenic area is obtained.

5. The early warning and control method based on time-series learning of scenic area visitor flow big data as described in claim 4, characterized in that, The process of obtaining the initial congestion risk index of the target scenic area is as follows: The estimated total passenger flow is decomposed into flow direction to obtain the estimated passenger load of the road segment; By comparing the estimated passenger flow load with the estimated travel time in the traffic impedance diagram, the congestion threshold is obtained to identify the over-limit judgment mark of the road segment. The road sections exceeding the limit are identified by collecting the road sections that exceed the limit, and the potential pressure-bearing road sections of the target scenic area are obtained. The entrance nodes connected to the potential pressure-bearing road sections are taken as potential pressure-bearing entrance nodes. Congestion situation assessment is conducted on potentially congested road sections to obtain the initial congestion status of these sections; The initial congestion risk is assessed based on the initial pressure state to obtain the initial congestion risk index of the target scenic area.

6. The early warning and control method based on time-series learning of scenic area visitor flow big data as described in claim 1, characterized in that, The process of obtaining the differentiated induction path is as follows: The historical migration duration data of the target scenic area is retrieved routinely to obtain the historical spatiotemporal migration reference data of the target scenic area. Based on the initial congestion risk index, the risk source is traced in the estimated travel time in the traffic impedance map to obtain the congestion bottleneck location of the target scenic area. Based on the set of congestion bottleneck locations and historical spatiotemporal migration reference data, the estimated total number of visitors to indoor refuge sites in the target scenic area is used to enumerate congestion avoidance routes, thereby obtaining a set of candidate refuge routes for the target scenic area. By comparing the impedance advantages and disadvantages of the estimated total travel time of the candidate refuge routes, differentiated guidance routes for the target scenic area are obtained.

7. The early warning and control method based on time-series learning of scenic area visitor flow big data as described in claim 6, characterized in that, The process of obtaining the differentiated guidance path for the target scenic area is as follows: The estimated travel time of each candidate evacuation route in the candidate evacuation route set is accumulated segment by segment to obtain the estimated total travel time of each candidate route. Based on the estimated total travel time, the candidate paths are checked for time-distance ratio to obtain a subset of effective guiding paths for the target scenic area. By applying asymmetric weights to the subset of effective guidance paths, differentiated guidance paths for the target scenic area can be obtained.

8. The early warning and control method based on time-series learning of scenic area visitor flow big data as described in claim 1, characterized in that, The process of obtaining the population flow convergence index corresponding to the differentiated guidance path is as follows: The differentiated guidance paths are structured and encapsulated to obtain the guidance path push package for the target scenic spot; The guidance route push package is distributed to the mobile terminal of the corresponding tourist to obtain the pushed guidance route record of the target scenic spot; Within a preset response period, the real-time location point sequence fed back by the mobile terminal is continuously collected to construct the actual movement trajectory segment of tourists in the target scenic area. Vector deviation analysis is performed on the actual movement trajectory segments of tourists and the differentiated guidance path to obtain the trajectory deviation degree of the target scenic spot; Based on the same differentiated guidance path, the directional synchronicity of the actual movement trajectory segments of tourists is measured to obtain the trajectory directional convergence component of the target scenic spot. Based on the trajectory direction convergence component, the speed synchronization of the actual movement trajectory segments of tourists is extracted to obtain the flow convergence index corresponding to the differentiated guidance path.

9. The early warning and control method based on time-series learning of scenic area visitor flow big data as described in claim 8, characterized in that, The process of obtaining the trajectory deviation of the target scenic area is as follows: The positioning points in the actual movement trajectory segment of tourists are spatially orthogonally projected onto the differentiated guidance path to obtain the orthogonal reference points of the target scenic area; Based on orthogonal reference points, the distance to each location point in the actual movement trajectory segment of tourists is measured point by point to obtain the deviation distance sequence of the target scenic spot; The deviation distance sequence is averaged to obtain the trajectory deviation of the target scenic area.

10. The early warning and control method based on time-series learning of scenic area visitor flow big data as described in claim 1, characterized in that, The process of receiving evacuation guidance instructions for the target scenic area is as follows: The difference between the crowd flow convergence index and the preset convergence level is extracted to obtain the convergence exceedance value of the target scenic area; Based on the convergence exceedance value, the push interval of the differentiated guidance path is extended and adjusted to obtain the correct push timing for the target scenic spot; Based on the convergence exceedance value, the frequency of push notifications per unit time for differentiated guidance paths is reduced to obtain the corrected push frequency parameter for the target scenic area. The parameters of correction push timing and correction push frequency are encapsulated in a binary frame structure to obtain the evacuation guidance instruction frame of the target scenic area. The evacuation guidance instruction frame audio stream is decoded to obtain the evacuation guidance instructions for the target scenic area.

Images