A scenic spot queuing situation real-time sensing and dispatching management system

Abstract

The application relates to the technical field of scenic spot management, and discloses a scenic spot queuing situation real-time sensing and dispatching management system, which comprises the following steps: a video acquisition module acquires video stream data of a ticket checking entrance of a scenic spot in real time; a visual analysis module carries out human target recognition processing on the video stream data, obtains the number of people in a queue and ticket checking passage efficiency, calculates waiting time, and constructs continuous time sequence queuing data; a trend prediction module extracts the change rate of the number of people in a queue, extracts the attenuation gradient of the ticket checking passage efficiency and periodic passenger flow characteristics, constructs a multi-dimensional time sequence feature vector, determines the passenger flow load index and passage resistance coefficient of a queuing area, carries out state generation processing and iteration deduction processing, and obtains the number of people in a queue prediction result of each future time node; and an early warning and dispatching module generates congestion early warning information and a ticket checking dispatching scheme when the number of people in a queue prediction result is greater than a congestion threshold value. The application improves the accuracy of passenger flow prediction.

Description

Technical Field

[0001] This invention relates to the field of scenic area management technology, and more specifically, to a real-time perception and scheduling management system for queuing status in scenic areas. Background Technology

[0002] With the continuous development of smart scenic area construction, scenic area operation and management are gradually shifting from traditional manual management to intelligent and information-based management methods. As the main channel for tourists to enter the scenic area, the queuing status at the ticket gates directly affects tourist flow efficiency, scenic area operation order, and tourist experience. Therefore, how to obtain real-time queuing information at the ticket gates and rationally organize tourist flow based on the queuing situation has become an important research direction for smart scenic area visitor flow management.

[0003] Existing patent CN108629323A discloses an integrated method for providing chain-like travel services for tourists in scenic areas. This method utilizes cameras to capture images of queues at attractions and identifies the number of people in the queue using a deep learning algorithm. It then combines this with the tourist queue flow rate to calculate queue time and congestion levels, providing data support for optimizing scenic area services. However, the aforementioned prior art queue prediction is mainly used to assist in optimizing tourist routes. Queue analysis estimates queue time based on the current number of people in the queue, queue flow rate, and historical data. It lacks effective processing for personnel tracking, duplicate counting elimination, and passage process analysis, resulting in insufficient accuracy in calculating the number of people in the queue and passage efficiency, making it difficult to accurately predict the development of the queue situation in the near future.

[0004] Therefore, it is necessary to design a real-time perception and scheduling management system for queuing in scenic areas to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a real-time perception and scheduling management system for queuing situations in scenic areas, which aims to solve the problem that the existing technology lacks effective processing of personnel tracking, duplicate counting elimination, and passage process, resulting in insufficient accuracy in calculating the number of people queuing and passage efficiency, and difficulty in accurately predicting the development process of queuing situations in the near future.

[0006] This invention proposes a real-time queuing situation perception and scheduling management system for scenic spots, comprising: The video capture module is used to acquire real-time video stream data from the scenic area's ticket gates; The visual analysis module is used to perform human target recognition processing on the video stream data to obtain the number of people in the queue and the ticket checking efficiency; based on the historical ticket checking efficiency, the number of people in the queue, and the ticket checking efficiency, the waiting time is calculated to obtain the queuing waiting time, and continuous time series queuing data is constructed. The trend prediction module is used to extract the rate of change of the number of people queuing between adjacent time steps in the continuous time series queuing data to generate queuing number evolution characteristics; extract the decay gradient of the ticket checking efficiency to generate efficiency fluctuation characteristics; generate periodic passenger flow characteristics based on the historical passenger flow of the current time node; fuse the queuing number evolution characteristics, efficiency fluctuation characteristics, and periodic passenger flow characteristics to construct a multi-dimensional time series feature vector; based on the multi-dimensional time series feature vector, determine the passenger flow load index and passage resistance coefficient of the queuing area, and perform state generation processing to obtain an evolution state sequence; use the evolution state sequence as the initial state for iterative deduction processing to obtain the queuing number prediction results for each future time node; The early warning and scheduling module is used to compare the predicted number of people in the queue with the congestion threshold. When the predicted number of people in the queue is greater than the congestion threshold, it generates congestion early warning information and generates a ticket checking and scheduling plan.

[0007] Furthermore, when the visual analysis module performs human target recognition processing on the video stream data to obtain the queue length and ticket checking efficiency, it includes: Based on the video stream data, human target tracking and deduplication are performed to obtain the number of people in the queue; according to the human target displacement trajectory and passage time in the preset ticket inspection area in the video stream data, the passage efficiency is calculated to obtain the ticket inspection passage efficiency.

[0008] Furthermore, when the visual analysis module obtains the queuing time, it includes: Based on the number of people queuing and the ticket checking efficiency, a basic waiting time is calculated to obtain the basic waiting time; based on the historical ticket checking efficiency and ticket checking efficiency, an efficiency decay assessment is performed to obtain the efficiency decay compensation time; based on the basic waiting time and the efficiency decay compensation time, the time is superimposed to obtain the queuing waiting time.

[0009] Furthermore, when constructing continuous time-series queuing data, the visual analysis module includes: The number of people queuing and the ticket checking efficiency are sampled in time series to obtain initial time series data; based on the initial time series data, outlier removal and smoothing filtering are performed to obtain corrected time series data; based on the corrected time series data and time node identifiers, data alignment and splicing are performed to obtain the continuous time series queuing data.

[0010] Furthermore, when constructing a multi-dimensional time-series feature vector, the trend prediction module includes: Based on the difference in the number of people queuing between adjacent time steps and the time step interval in the continuous time series queuing data, the rate of change is calculated to obtain the queuing number evolution characteristic; based on the difference in ticket checking efficiency between adjacent time steps in the continuous time series queuing data, gradient calculation is performed to obtain the passage efficiency fluctuation characteristic; based on the current time node identifier and historical passenger flow data, concurrent passenger flow retrieval and weight allocation are performed to obtain the periodic passenger flow characteristic; based on the queuing number evolution characteristic, the passage efficiency fluctuation characteristic, and the periodic passenger flow characteristic, normalization and vector concatenation are performed to obtain the multidimensional time series feature vector.

[0011] Furthermore, when determining the passenger flow load index and traffic resistance coefficient of the queuing area, the trend prediction module includes: The passenger flow load index is obtained by performing a ratio calculation based on the number of people in the queue and the upper limit of the capacity of the queuing area; the passage resistance coefficient is obtained by performing a difference calculation and normalization based on the ticket checking efficiency and the benchmark passage efficiency of the queuing area.

[0012] Furthermore, when the trend prediction module obtains the evolutionary state sequence, it includes: Based on the passenger flow load index and the traffic resistance coefficient, a nonlinear coupling calculation is performed to obtain the queuing congestion accumulation rate; based on the queuing congestion accumulation rate, the passenger flow load index, and the traffic resistance coefficient, vector splicing processing is performed to obtain the state vector at the current time step; based on the state vectors at each time step, time-series arrangement processing is performed to obtain the evolution state sequence.

[0013] Furthermore, when the trend prediction module performs iterative deduction processing using the evolutionary state sequence as the initial state, it includes: Based on environmental factor data, interference factor generation processing is performed to obtain external interference factors; based on the terminal state vector of the evolution state sequence and the external interference factors, feature fusion processing is performed to obtain the initial inference vector; based on the initial inference vector and the state transition matrix, iterative calculation processing is performed through future time window step size to obtain the queuing number prediction results for each future time node.

[0014] Furthermore, when the early warning scheduling module generates the ticket checking scheduling plan, it includes: Based on the predicted number of people in the queue and the congestion threshold, a difference calculation is performed to obtain the congestion excess; based on the congestion excess, a scheduling level matching is performed to obtain the target scheduling level.

[0015] Furthermore, when generating the ticket checking scheduling plan, the early warning scheduling module also includes: When the target scheduling level is Level 1, the backup ticket checking channel is opened and the number of ticket checking personnel is increased; when the target scheduling level is Level 2, tourists are guided to divert and transportation connection resources are coordinated.

[0016] Compared with existing technologies, the advantages of this invention are as follows: By recognizing, tracking, and deduplicating human targets in video streams, the problem of duplicate counting is avoided. Combined with traffic trajectory analysis to improve ticket checking efficiency, the acquisition of queue size and traffic efficiency becomes more accurate and reliable, thereby improving the quality of basic data perception. Through time-series sampling, anomaly removal, smoothing filtering, and data alignment, discrete detection results are transformed into continuous time-series queue data, which can fully depict the dynamic process of queue status changes over time, providing a high-quality data foundation for subsequent predictions. By introducing a historical traffic efficiency and current efficiency decay assessment mechanism, the basic waiting time is dynamically compensated and corrected, making the waiting time calculation take into account both real-time performance and historical trends, improving the accuracy and stability of the estimation results. By integrating the rate of change in queue size, traffic efficiency fluctuation characteristics, and periodic passenger flow characteristics, a multi-dimensional time-series feature vector is constructed. Furthermore, by introducing a load index and a traffic resistance coefficient, a quantitative expression of the multi-factor coupled state of the queuing system is achieved, enhancing the system's ability to model complex passenger flow scenarios. Iterative deduction based on evolutionary state sequences, incorporating external disturbance factors and state transition mechanisms, enables the prediction process to not only rely on historical trends but also simulate evolutionary paths under environmental disturbances, thereby improving the robustness and accuracy of short- and medium-term passenger flow predictions. By comparing the predicted queue length with congestion thresholds, early warning information is automatically generated and a tiered scheduling strategy is triggered. The number of ticket gates opened, staffing levels, and passenger diversion plans can be dynamically adjusted according to different congestion levels, achieving a shift from "passive response" to "proactive control." Attached Figure Description

[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a functional block diagram of the scenic area queuing situation real-time perception and scheduling management system provided in an embodiment of the present invention; Figure 2 The logical flowchart of real-time perception and scheduling management of queuing status in scenic areas provided in this embodiment of the invention. Detailed Implementation

[0018] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey its scope to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features described herein can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0019] In some embodiments of this application, see Figure 1-2 As shown, a real-time perception and scheduling management system for queuing in scenic areas is proposed, including: The video capture module is used to acquire real-time video stream data from the scenic area's ticket gates; The visual analysis module is used to perform human target recognition processing on video stream data to obtain the number of people in queue and ticket checking efficiency; based on historical ticket checking efficiency, number of people in queue and ticket checking efficiency, the waiting time is calculated to obtain the queuing waiting time, and continuous time series queuing data is constructed. The trend prediction module is used to extract the rate of change of the number of people queuing between adjacent time steps in continuous time series queuing data to generate queuing number evolution characteristics; extract the decay gradient of ticket checking efficiency to generate efficiency fluctuation characteristics; generate periodic passenger flow characteristics based on the historical passenger flow of the current time node; fuse the queuing number evolution characteristics, efficiency fluctuation characteristics, and periodic passenger flow characteristics to construct a multi-dimensional time series feature vector; based on the multi-dimensional time series feature vector, determine the passenger flow load index and passage resistance coefficient of the queuing area, and perform state generation processing to obtain an evolutionary state sequence; use the evolutionary state sequence as the initial state for iterative deduction processing to obtain the queuing number prediction results for each future time node; The early warning and scheduling module compares the predicted number of people in the queue with the congestion threshold. When the predicted number of people in the queue is greater than the congestion threshold, it generates a congestion warning and a ticket checking and scheduling plan.

[0020] In this embodiment, the video acquisition module is deployed at a fixed monitoring point or on a high-level overhead structure at the scenic area's ticket gate to continuously collect video stream data of the ticket gate area. The video stream data can be a real-time high-definition video stream, output at a fixed frame rate to ensure the continuity and stability of subsequent human target recognition and trajectory analysis. The video acquisition module can be combined with a multi-camera collaborative acquisition method to cover different ticket gates and queuing areas to reduce information loss caused by obstruction.

[0021] When the visual analysis module performs human target recognition on video stream data, it first identifies human targets in each video frame based on a deep learning object detection model to obtain initial human detection results. Subsequently, a multi-target tracking algorithm is used to correlate the trajectories of human targets in consecutive frames to obtain the temporal trajectory of each human target. Simultaneously, duplicate detection results of the same human target in different frames are deduplicated to avoid affecting queue headcount statistics, thereby obtaining the effective number of people queuing within the current queuing area.

[0022] After obtaining the trajectory of the human target, spatial constraints are further applied to the human target based on the preset ticket checking area, retaining only human targets located in the queuing area and in a waiting state as valid queuing objects. Based on this, the passage time of a single human target is calculated according to the dwell time in the queuing area and the passage time after entering the ticket checking area. The passage times of all human targets are statistically summarized to obtain the average passage efficiency per unit time, thus obtaining the ticket checking passage efficiency parameter.

[0023] When calculating queue waiting time, the visual analysis module first establishes a basic queuing model based on the current number of people in the queue and the ticket checking efficiency. The basic waiting time is obtained by calculating the ratio of the number of people in the queue to the throughput capacity per unit time. This basic waiting time reflects the time required for the queue to clear naturally under the condition that the current throughput capacity remains constant.

[0024] To further consider the dynamic changes in traffic efficiency, the visual analysis module introduces historical ticket checking efficiency as a benchmark reference value and compares it with the current ticket checking efficiency to obtain the difference in efficiency changes. Based on this, attenuation modeling is applied to the difference in efficiency changes to characterize the impact of decreased or fluctuating traffic capacity on queuing time, and the efficiency attenuation compensation time is calculated accordingly to correct the deviation of the base waiting time.

[0025] Finally, the base waiting time and the efficiency decay compensation time are superimposed to obtain the corrected queuing waiting time. This queuing waiting time not only reflects the current static queue size but also incorporates factors related to changes in throughput capacity.

[0026] When constructing continuous time-series queuing data, the visual analysis module performs periodic time-series sampling on the number of people queuing and ticket checking efficiency. The sampling interval can be configured to be on the order of seconds or minutes depending on the actual scenario to form an initial time-series data set. The initial time-series data contains the original observations of the number of people queuing and the efficiency at different time points.

[0027] The initial time-series data undergoes preprocessing, including outlier removal and smoothing filtering. Outlier removal removes abrupt data points caused by occlusion, recognition errors, or network jitter; smoothing filtering reduces the impact of short-term random fluctuations on the overall trend, making data changes more consistent with the actual queuing state evolution, thus obtaining corrected time-series data.

[0028] Further time alignment processing is performed on the corrected time series data according to the time node identifier, and data from different sources or different sampling periods are uniformly mapped to the standard time axis. The queuing number data and ticket checking efficiency data are structured and spliced ​​together to form a unified data representation format, thus obtaining continuous time series queuing data.

[0029] Continuous time series queuing data serves as the basic input data for the subsequent trend prediction module. It can fully reflect the continuous evolution of the queuing status at the scenic spot's ticket gate over time, providing high-quality data support for subsequent queuing trend prediction, congestion analysis, and scheduling decisions.

[0030] In a more specific implementation, the trend prediction module uses continuous time series queuing data as the basic input data. The continuous time series queuing data includes at least a sequence of queuing numbers and a sequence of ticket checking efficiency obtained by sampling at a uniform time interval. A unique time identifier is attached to each time node to ensure the temporal consistency and traceability of subsequent feature calculations.

[0031] The trend prediction module first performs differential calculations on the continuous time series queuing data for adjacent time steps to obtain the changing trend of the number of people in the queue. Specifically, it first performs differential calculations on the number of people in the queue between two adjacent time sampling points to obtain the absolute change in the number of people in the queue within adjacent time steps; then, it divides the absolute change by the corresponding time interval to obtain the rate of change per unit time; to reduce the impact of sudden anomaly detection errors, it also performs a moving average processing on the rate of change of multiple consecutive time steps to obtain a smoothed curve of the rate of change of the number of people in the queue; finally, it uses this rate of change curve as a feature of the evolution of the number of people in the queue to characterize the growth rate or decline rate of the queue size.

[0032] The trend prediction module performs dynamic gradient analysis on ticket checking efficiency to obtain its fluctuation characteristics. Specifically, it first calculates the difference between adjacent time points in the continuous time series to obtain the increment of efficiency change; then, it processes the absolute value of the increment to reflect the fluctuation amplitude, while retaining the sign to distinguish between upward and downward trends; further, it smooths the fluctuation series using an exponentially weighted moving average method, giving higher weight to recent data to enhance sensitivity to changes in real-time capacity; finally, it obtains the efficiency fluctuation characteristics to characterize the stability and decay trend of the ticket checking system's capacity.

[0033] The trend prediction module further retrieves historical passenger flow data from the historical database for the same time period or similar dates based on the current time node identifier to construct periodic passenger flow characteristics. Specifically, firstly, time attribute information is extracted based on the current time node, including hourly segments, weekday attributes, and holiday identifiers; then, a set of historical passenger flow samples matching the current time attributes is retrieved from the historical passenger flow database, such as historical daily passenger flow for the same hourly segment or historical data for the same holiday type; then, multiple retrieved historical samples are assigned different weight coefficients according to time similarity or date proximity, and weighted summation is performed to obtain the basic periodic passenger flow value; this basic periodic passenger flow value can also be standardized to eliminate numerical differences under different time scales, thereby forming periodic passenger flow characteristics. The weight coefficients are calculated based on the matching degree between historical data and the current time, for example, by determining the contribution ratio of each historical sample through a time distance decay function or a normalized weight allocation method based on similarity scores, and then forming periodic passenger flow characteristics through weighted averaging.

[0034] The trend prediction module unifies the characteristics of queue length evolution, traffic efficiency fluctuation, and periodic passenger flow to construct a multi-dimensional time-series feature vector. Specifically, firstly, the characteristics of queue length evolution, traffic efficiency fluctuation, and periodic passenger flow are normalized separately. The normalization method can be min-max normalization or Z-score standardization to unify the numerical range of different features. Then, the three types of features are concatenated along the feature dimension according to a preset feature order to form an initial feature vector. Noise reduction or feature smoothing can also be performed on the concatenated feature vector to reduce the impact of local outliers on the overall prediction. Finally, a multi-dimensional time-series feature vector is obtained for subsequent state modeling.

[0035] Based on multidimensional time-series feature vectors, the trend prediction module calculates the passenger flow load index of the queuing area. Specifically, it first obtains the number of people queuing at the current time point and reads the pre-set upper limit of the queuing area capacity, which is the maximum number of people that can be accommodated based on the physical space of the ticket gate, the width of the passage, and the safety carrying capacity standards. Then, it calculates the ratio of the number of people queuing to the upper limit of capacity to obtain the basic load ratio. To avoid linear distortion when the load ratio is close to saturation, the basic load ratio is also processed by nonlinear mapping, such as using a piecewise function or an exponential enhancement function, to make the changes in the high-load area more sensitive. Finally, the passenger flow load index is obtained to characterize the congestion level of the queuing area.

[0036] The trend prediction module further calculates the traffic resistance coefficient. Specifically, it first obtains the current ticket checking efficiency and the preset benchmark efficiency, which is the average throughput capacity statistically obtained based on historical normal operating conditions. Then, it calculates the difference between the current efficiency and the benchmark efficiency to obtain the efficiency deviation. Further, it normalizes the efficiency deviation to map it to a uniform numerical range. To enhance the sensitivity to a decrease in throughput, negative deviations can be amplified so that the resistance coefficient increases faster when throughput decreases. Finally, the traffic resistance coefficient is obtained, which is used to characterize the degree of decrease in the throughput capacity of the ticket checking system.

[0037] The trend prediction module uses a nonlinear coupling calculation based on the passenger flow load index and the traffic resistance coefficient to generate the queuing congestion accumulation rate. Specifically, firstly, based on the current time step, the passenger flow load index and the traffic resistance coefficient are combined to calculate a basic congestion intensity index. For example, an exponential amplification factor is introduced after multiplying the load index and the resistance coefficient for nonlinear mapping to enhance sensitivity under high load conditions. The congestion intensity index is used to characterize the degree of congestion accumulation in the queuing system per unit time. Subsequently, the congestion intensity index, passenger flow load index, and traffic resistance coefficient are vector-concatenated to form a state vector at the current time step. In the time dimension, the state vectors of multiple consecutive time steps are sequentially arranged while maintaining time index consistency, thus forming a complete evolutionary state sequence used to describe the state trajectory of the queuing system over time.

[0038] When the trend prediction module extrapolates future states, it first acquires external environmental factor data, including but not limited to weather changes, holiday attributes, emergency information, and traffic connection status. Based on this data, it then models interference factors to generate an external interference factor vector. Specifically, it first collects environmental factor data, including weather levels, temperature, rainfall, holiday types, emergency indicators, and traffic congestion indices. Next, it encodes various environmental factors, converting qualitative information into numerical vectors. Then, it assigns weight coefficients based on the degree of influence of different factors on the queuing system, weighting and fusing the environmental factors across all dimensions to obtain a unified external interference factor vector, used to characterize the impact of the external environment on the evolution of the queuing state.

[0039] The trend prediction module fuses the terminal state vector of the evolutionary state sequence with external interference factors. The fusion method can be vector weighted superposition or feature concatenation mapping to obtain an initial inference vector, which serves as the initial input condition for future state prediction. For example, factors such as weather, holidays, and traffic conditions are numerically encoded and then weighted and fused using preset weight coefficients or based on historical impact statistics to form a calculable unified external interference factor vector. Specifically, the terminal state vector in the evolutionary state sequence is first used as the current system state input; then, the external interference factor vector is fused with the terminal state vector, either through concatenation and linear mapping or weighted superposition mapping, to form the initial inference vector; then, the initial inference vector is input into a preset state transition matrix model, and the state vector for the next time step is obtained through matrix multiplication; a nonlinear activation function can also be introduced in each iteration to simulate the nonlinear growth characteristics of the queuing system under high load; by repeatedly iterating and calculating multiple time steps, the state sequence for multiple future time nodes is obtained, and the corresponding queuing number prediction results are extracted from it.

[0040] The state transition matrix is ​​obtained by extracting training samples from historical continuous time series queuing data through a sliding time window method and fitting parameters using least squares regression or multiple linear regression methods. The mapping relationship between the state vector and the state at the next time step is used as the learning objective, thereby forming a fixed-dimensional set of state transition matrix parameters, which is updated or periodically retrained during online operation.

[0041] Through the above-mentioned stepwise feature extraction, state modeling and iterative deduction process, the prediction results can not only rely on historical trend changes, but also comprehensively consider changes in traffic capacity, periodic patterns of passenger flow and external environmental disturbances. This enables multi-factor coupled prediction of queuing patterns at scenic spot ticket gates, improving the stability and accuracy of predictions in complex passenger flow scenarios.

[0042] The early warning and scheduling module is used to make real-time risk assessments on the queuing number prediction results output by the trend prediction module, and generate graded early warning information and corresponding scheduling strategies when potential congestion risks are detected, thereby realizing dynamic resource control and emergency response at the scenic area's ticket gates.

[0043] The early warning and scheduling module first obtains the predicted queue length for each future time point and simultaneously reads the pre-configured congestion threshold. The congestion threshold is determined based on a comprehensive analysis of the maximum capacity of the ticket gate, safe passage standards, and historical peak passenger flow statistics. For example, the congestion threshold can be defined as a fixed percentage of the maximum capacity of the ticket gate area, or determined by taking quantiles from historical peak passenger flow statistics. After obtaining the above data, the system performs a time-by-time comparison and analysis between the predicted queue length and the congestion threshold.

[0044] When the predicted queue size at any future time point exceeds the congestion threshold, the system calculates the congestion excess. Specifically, it subtracts the congestion threshold from the predicted queue size to obtain the difference exceeding the safe carrying capacity. This difference is used to quantify the severity of congestion and serves as a core input parameter for subsequent scheduling level classification. If the predicted queue size does not exceed the congestion threshold, normal queuing is maintained based on the current ticket checking scheme.

[0045] The early warning and scheduling module performs scheduling level matching based on congestion exceedance. Specifically, multiple congestion level intervals are pre-defined; for example, level 1 corresponds to a mild exceedance, level 2 to a moderate exceedance, and level 3 to a severe exceedance. The system determines the target scheduling level corresponding to the current time point by comparing the congestion exceedance with the threshold of each level interval step by step, thereby realizing the mapping from quantitative indicators to hierarchical scheduling strategies.

[0046] When the target scheduling level is detected to meet the triggering conditions, the system automatically generates a congestion warning. The congestion warning includes at least the current time, the predicted number of people in the queue, the congestion excess, the congestion level, and the risk duration window. It is pushed in real time through the scenic area scheduling platform or management terminal to prompt management personnel to make intervention decisions.

[0047] When generating a ticket checking schedule, the early warning scheduling module first calls the corresponding preset scheduling strategy library according to the target scheduling level. The scheduling strategy library stores resource scheduling rules and execution action sets corresponding to different levels in advance to ensure the consistency and executability of the scheduling response.

[0048] When the target scheduling level is Level 1, the system implements a light scheduling strategy, including automatically opening backup ticket checking channels and expanding the capacity of the current ticket checking windows. Simultaneously, the system sends personnel replenishment instructions to the on-site management terminal, dispatching pre-set mobile ticket checking personnel to the corresponding ticket checking gates for support, thereby improving local throughput capacity without affecting overall operations.

[0049] When the target scheduling level is Level 2, the system implements a medium scheduling strategy. Under this strategy, the system not only adjusts ticket gate resources but also triggers a visitor flow guidance mechanism. For example, it pushes flow guidance information through the scenic area's broadcast system, electronic guidance screens, or mobile terminals to guide visitors to other ticket gates or backup entrances to alleviate queuing pressure. Simultaneously, the system collaborates with the transportation resource management module to optimize the scheduling of external shuttle vehicles, such as adjusting departure frequencies or guiding visitors to enter the park during off-peak hours, to reduce instantaneous visitor flow pressure at the entrance.

[0050] Through the aforementioned early warning comparison, excess quantity calculation, level matching, and hierarchical scheduling execution process, the early warning scheduling module can achieve early identification and rapid response to the risk of congestion at the scenic area's ticket gates. This transforms the system from a traditional post-event processing mode to a prediction-based proactive scheduling mode, thereby significantly improving the scenic area's operational efficiency and safety management capabilities during peak periods.

[0051] In summary, by using video stream human target recognition, tracking, and deduplication to avoid duplicate counting, and combining this with flow trajectory analysis to improve ticket checking efficiency, the acquisition of queue size and flow efficiency becomes more accurate and reliable, thus improving the quality of basic data perception. Through time-series sampling, anomaly removal, smoothing filtering, and data alignment, discrete detection results are transformed into continuous time-series queue data, which can fully depict the dynamic process of queue status changes over time, providing a high-quality data foundation for subsequent predictions. By introducing a historical flow efficiency and current efficiency decay assessment mechanism, the basic waiting time is dynamically compensated and corrected, ensuring that waiting time calculation takes into account both real-time performance and historical trends, improving the accuracy and stability of the estimation results. By integrating the rate of change in queue size, flow efficiency fluctuation characteristics, and periodic passenger flow characteristics, a multi-dimensional time-series feature vector is constructed. Furthermore, by introducing a load index and flow resistance coefficient, a quantitative expression of the multi-factor coupled state of the queuing system is achieved, enhancing the system's ability to model complex passenger flow scenarios. Iterative deduction based on evolutionary state sequences, incorporating external disturbance factors and state transition mechanisms, enables the prediction process to not only rely on historical trends but also simulate evolutionary paths under environmental disturbances, thereby improving the robustness and accuracy of short- and medium-term passenger flow predictions. By comparing the predicted queue length with congestion thresholds, early warning information is automatically generated and a tiered scheduling strategy is triggered. The number of ticket gates opened, staffing levels, and passenger diversion plans can be dynamically adjusted according to different congestion levels, achieving a shift from "passive response" to "proactive control."

[0052] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the present invention.

Claims

1. A real-time perception and scheduling management system for queuing status in scenic areas, characterized in that, include: The video capture module is used to acquire real-time video stream data from the scenic area's ticket gates; The visual analysis module is used to perform human target recognition processing on the video stream data to obtain the number of people in the queue and the ticket checking efficiency; based on the historical ticket checking efficiency, the number of people in the queue, and the ticket checking efficiency, the waiting time is calculated to obtain the queuing waiting time, and continuous time series queuing data is constructed. The trend prediction module is used to extract the rate of change of the number of people queuing between adjacent time steps in the continuous time series queuing data to generate queuing number evolution characteristics; extract the decay gradient of the ticket checking efficiency to generate efficiency fluctuation characteristics; and generate periodic passenger flow characteristics based on the historical passenger flow of the current time point. The queuing number evolution characteristics, traffic efficiency fluctuation characteristics, and periodic passenger flow characteristics are fused together to construct a multi-dimensional time series feature vector; based on the multi-dimensional time series feature vector, the passenger flow load index and traffic resistance coefficient of the queuing area are determined, and state generation processing is performed to obtain an evolution state sequence; using the evolution state sequence as the initial state, iterative deduction processing is performed to obtain the queuing number prediction results for each future time node; The early warning and scheduling module is used to compare the predicted number of people in the queue with the congestion threshold. When the predicted number of people in the queue is greater than the congestion threshold, it generates congestion early warning information and generates a ticket checking and scheduling plan.

2. The scenic area queuing situation real-time perception and scheduling management system according to claim 1, characterized in that, When the visual analysis module performs human target recognition processing on the video stream data to obtain the number of people queuing and ticket checking efficiency, it includes: Based on the video stream data, human target tracking and deduplication are performed to obtain the number of people in the queue; according to the human target displacement trajectory and passage time in the preset ticket inspection area in the video stream data, the passage efficiency is calculated to obtain the ticket inspection passage efficiency.

3. The scenic area queuing situation real-time perception and scheduling management system according to claim 2, characterized in that, When the visual analysis module obtains the queuing time, it includes: Based on the number of people queuing and the ticket checking efficiency, a basic waiting time is calculated to obtain the basic waiting time; based on the historical ticket checking efficiency and ticket checking efficiency, an efficiency decay assessment is performed to obtain the efficiency decay compensation time; based on the basic waiting time and the efficiency decay compensation time, the time is superimposed to obtain the queuing waiting time.

4. The scenic area queuing situation real-time perception and scheduling management system according to claim 3, characterized in that, When constructing continuous time series queuing data, the visual analysis module includes: The number of people queuing and the ticket checking efficiency are sampled in time series to obtain initial time series data; based on the initial time series data, outlier removal and smoothing filtering are performed to obtain corrected time series data; based on the corrected time series data and time node identifiers, data alignment and splicing are performed to obtain the continuous time series queuing data.

5. The scenic area queuing situation real-time perception and scheduling management system according to claim 4, characterized in that, When constructing a multi-dimensional time-series feature vector, the trend prediction module includes: Based on the difference in the number of people queuing between adjacent time steps and the time step interval in the continuous time series queuing data, the rate of change is calculated to obtain the queuing number evolution characteristic; based on the difference in ticket checking efficiency between adjacent time steps in the continuous time series queuing data, gradient calculation is performed to obtain the passage efficiency fluctuation characteristic; based on the current time node identifier and historical passenger flow data, concurrent passenger flow retrieval and weight allocation are performed to obtain the periodic passenger flow characteristic; based on the queuing number evolution characteristic, the passage efficiency fluctuation characteristic, and the periodic passenger flow characteristic, normalization and vector concatenation are performed to obtain the multidimensional time series feature vector.

6. The scenic area queuing situation real-time perception and scheduling management system according to claim 5, characterized in that, The trend prediction module, when determining the passenger flow load index and passage resistance coefficient of the queuing area, includes: The passenger flow load index is obtained by performing a ratio calculation based on the number of people in the queue and the upper limit of the capacity of the queuing area; the passage resistance coefficient is obtained by performing a difference calculation and normalization based on the ticket checking efficiency and the benchmark passage efficiency of the queuing area.

7. The scenic area queuing situation real-time perception and scheduling management system according to claim 6, characterized in that, When the trend prediction module obtains the evolutionary state sequence, it includes: Based on the passenger flow load index and the traffic resistance coefficient, a nonlinear coupling calculation is performed to obtain the queuing congestion accumulation rate; based on the queuing congestion accumulation rate, the passenger flow load index, and the traffic resistance coefficient, vector splicing processing is performed to obtain the state vector at the current time step; based on the state vectors at each time step, time-series arrangement processing is performed to obtain the evolution state sequence.

8. The scenic area queuing situation real-time perception and scheduling management system according to claim 7, characterized in that, When the trend prediction module performs iterative deduction processing using the evolutionary state sequence as the initial state, it includes: Based on environmental factor data, interference factor generation processing is performed to obtain external interference factors; based on the terminal state vector of the evolution state sequence and the external interference factors, feature fusion processing is performed to obtain the initial inference vector; based on the initial inference vector and the state transition matrix, iterative calculation processing is performed through future time window step size to obtain the queuing number prediction results for each future time node.

9. The scenic area queuing situation real-time perception and scheduling management system according to claim 8, characterized in that, When the early warning scheduling module generates the ticket checking scheduling plan, it includes: Based on the predicted number of people in the queue and the congestion threshold, a difference calculation is performed to obtain the congestion excess; based on the congestion excess, a scheduling level matching is performed to obtain the target scheduling level.

10. The scenic area queuing situation real-time perception and scheduling management system according to claim 9, characterized in that, When the early warning scheduling module generates the ticket checking scheduling plan, it also includes: When the target scheduling level is Level 1, the backup ticket checking channel is opened and the number of ticket checking personnel is increased; when the target scheduling level is Level 2, tourists are guided to divert and transportation connection resources are coordinated.

Images