Security intelligent guiding method based on scene analysis and emergency plan triggering

Abstract

The invention relates to the technical field of safety intelligent guidance, and particularly discloses a safety intelligent guidance method based on scene analysis and emergency plan triggering. Comprising the steps of situation awareness data stream generation, combined crisis state model construction, cooperative guidance instruction generation, field effect guidance environment deployment, physical feedback signal generation, closed loop validity verification result generation and guidance environment adaptive adjustment. According to the method, situation awareness data streams are generated through fusion of a multi-source sensor, a composite crisis state model is constructed, congestion and panic levels are quantitatively evaluated, then spatial differentiation cooperative guidance instructions are generated, an acousto-optic linkage field effect guidance environment is deployed, and the guidance strategy effectiveness is verified by monitoring the physical response of a crowd, so that the guidance efficiency of the crowd is improved. The guide environment is adaptively adjusted; according to the invention, the conversion from post-event response to pre-warning is realized, the evacuation efficiency and the system reliability are improved, and an innovative solution is provided for the safety management of large public places.

Description

Technical Field

[0001] This invention relates to the field of intelligent safety guidance technology, and more specifically, to an intelligent safety guidance method based on scenario analysis and emergency plan triggering. Background Technology

[0002] In high-density gathering places such as large public transportation hubs, stadiums, and commercial complexes, crowd control during emergencies presents enormous challenges. The core issue lies in the dynamic and unpredictable nature of crowd behavior. Minor disturbances can quickly escalate into large-scale congestion or even stampedes. Traditional safety management methods struggle to detect the early stages of a crisis in real time and accurately, and are unable to quantify and assess the collective emotions and behavioral trends of the crowd, resulting in intervention measures often lagging behind the rapid deterioration of the situation.

[0003] Currently, the solutions commonly used in the industry mainly rely on manual patrols, traditional video surveillance, and pre-set static emergency plans. Security personnel assess crowd flow through on-site observation and walkie-talkie communication, and conduct evacuations via loudspeakers or manual guidance. Meanwhile, fixed emergency exit signs, evacuation route maps, and a global emergency broadcast system constitute the main guidance facilities. These methods play a role in routine crowd management, but under extreme pressure, they collectively form a passive, traditional emergency response system that relies on human experience.

[0004] However, traditional methods have significant drawbacks. First, their perception capabilities are severely insufficient. Relying on human eyes and ordinary cameras, they struggle to penetrate dense crowds and fail to capture key physical indicators such as trampling pressure and pushing force, leading to delays and blind spots in hazard identification. Second, guidance measures lack adaptability and spatial variability. Static signs and global broadcasts cannot dynamically adjust evacuation routes based on the real-time location of congestion points, and may even lead crowds to new danger zones. Uniform instructions also fail to provide effective guidance for the specific situations of people in different locations. Finally, the entire intervention process is an open-loop system, lacking a real-time feedback and verification mechanism for guidance effectiveness. Decision-makers cannot know whether evacuation instructions have been effectively implemented, nor can they dynamically optimize evacuation strategies based on the actual reactions of the crowd, resulting in low efficiency and insufficient reliability in emergency response. Summary of the Invention

[0005] In view of this, in order to solve the problems mentioned in the background technology, a safety intelligent guidance method based on scenario analysis and emergency plan triggering is proposed.

[0006] The objective of this invention can be achieved through the following technical solution: This invention provides a safety intelligent guidance method based on scenario analysis and emergency plan triggering, including the following steps: S1, Situational awareness data stream generation: Generate a multi-source fusion situational awareness data stream that represents the on-site physical environment and the state of crowd behavior.

[0007] S2. Construction of a composite crisis state model: Based on multi-source fusion of situational awareness data streams, a composite crisis state model is generated that includes the location of the danger source, the precise coordinates of the congestion bottleneck, and the risk level of panic spread.

[0008] S3. Cooperative Guidance Command Generation: Based on the composite crisis state model, generate a set of spatially differentiated cooperative guidance commands that include light effect patterns, flow speed, and audio content.

[0009] S4. Field Effect Guiding Environment Deployment: Receives spatially differentiated collaborative guidance instructions to deploy an acoustic-optical linked field effect guiding environment.

[0010] S5. Physical Feedback Signal Generation: During the field effect-guided environmental action, a physical feedback signal representing the effect of command execution is generated by monitoring the physical response of the crowd.

[0011] S6. Generation of Closed-Loop Validity Verification Results: Input the physical feedback signal of the pedestrian flow to generate a closed-loop validity verification result of the guidance strategy to determine whether the current guidance strategy is successful.

[0012] S7. Adaptive Adjustment of Guiding Environment: Receives the closed-loop validity verification result of the guiding strategy. When the strategy is determined to fail, it performs dynamic, closed-loop adaptive adjustment of the field effect guiding environment.

[0013] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention can achieve early and accurate early warning of potential stampede risks through the deep fusion of multi-source heterogeneous sensors. Compared with the traditional method of relying on video surveillance or manual patrol, this method can directly quantify key physical indicators such as pushing force and ground pressure within the crowd, thereby identifying the danger in the early stage of congestion and before the stampede accident occurs, greatly advancing the timing of safety intervention from post-event response to pre-event warning, and significantly improving the initiative and foresight of emergency response.

[0014] (2) The guidance strategy provided by this invention has a high degree of spatial variability and dynamic adaptability. It abandons the traditional "one-size-fits-all" emergency broadcasting model and can generate completely different guidance instructions for people in different areas based on the real-time and precise location of congestion points. This can effectively slow down the influx of people from upstream and dynamically open up and guide people to the optimal backup evacuation channels. This intelligent diversion strategy, which is adapted to local conditions and changes in real time, greatly improves the utilization efficiency of evacuation routes and effectively avoids secondary congestion caused by unclear instructions or improper routes.

[0015] (3) This invention ensures the practical effectiveness of the guidance strategy by constructing a complete closed-loop control system of "perception-decision-execution-feedback". While executing guidance instructions, the system continuously collects physical feedback signals from the crowd to quantitatively evaluate the evacuation effect, and adaptively adjusts or even completely reconstructs the guidance plan based on the evaluation results. This dynamic optimization mechanism based on real-time effect verification overcomes the shortcomings of traditional plans that lack flexibility and cannot guarantee execution effect, enabling the entire evacuation process to have the ability to self-correct and continuously optimize, significantly enhancing the reliability and robustness of the emergency system. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the method steps of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 As shown, the present invention provides a safety intelligent guidance method based on scenario analysis and emergency plan triggering, including: S1, situational awareness data stream generation: generating a multi-source fusion situational awareness data stream that characterizes the on-site physical environment and the state of crowd behavior.

[0020] In a specific embodiment of the present invention, the step of generating a multi-source fusion situational awareness data stream characterizing the on-site physical environment and the state of crowd behavior includes: scanning the crowd with millimeter-wave radar to form a group dynamic mechanical signal.

[0021] By collecting data on the load of people stepping on the ground using a pressure sensor array, a high-precision local pressure distribution map is generated.

[0022] By fusing group dynamic mechanical signals with local pressure distribution maps and aligning timestamps, a multi-source fused situational awareness data stream is generated.

[0023] It should be noted that this step aims to comprehensively collect crowd dynamics and environmental physical information within a specific area through the collaborative work of multiple sensors, and integrate this information into a unified data stream to provide raw data input for subsequent crisis analysis. The implementation process begins with selecting monitoring locations, choosing areas along the pre-defined evacuation routes where people easily converge or change direction, such as stairwells, entrances to narrow passages, or lobby exits, defining these areas as critical nodes on the evacuation routes. At these critical nodes, millimeter-wave radar, ground-embedded pressure sensor arrays, and spatial microphones are deployed and installed. When the system starts or receives a monitoring command, a central control module sends a synchronization activation signal to these three types of sensor devices. This signal ensures that all sensors begin operating at exactly the same time, laying the foundation for subsequent data timestamp alignment.

[0024] Once activated, the millimeter-wave radar begins emitting high-frequency electromagnetic waves into its designated coverage area and senses the environment by receiving and analyzing signals reflected from the crowd. By calculating the Doppler shift of the signals, the movement speed and direction of different individuals within the crowd can be precisely measured. After data processing and averaging, a vector field describing the overall movement trend of the crowd is formed. Simultaneously, by analyzing the gradient changes in the velocity vector field—that is, areas where the speed of the crowd at the rear is significantly faster than that in front—the intensity of the pushing and shoving can be identified and quantified. This speed and pushing trend data is integrated into a continuous time-series data set, constituting a dynamic mechanical signal of the crowd.

[0025] Simultaneously, an array of pressure sensors embedded in the ground also begins to operate. This array consists of hundreds or even thousands of small pressure sensor cells laid in a grid beneath the ground, capable of sensing the physical pressure applied from above in real time. When people step on the ground, each sensor cell outputs an electrical signal proportional to the pressure magnitude. The system collects the signal values ​​from all cells in real time and combines these discrete pressure data points into a two-dimensional data matrix. This matrix, rendered in pseudo-color, visually displays the pressure distribution across the entire area, forming a high-precision local pressure distribution map, with highlighted areas representing locations of concentrated pressure.

[0026] Finally, the system fuses the crowd dynamics signals from millimeter-wave radar with the local pressure distribution maps from a pressure sensor array. The core of this fusion lies in time alignment; since the sensors are activated synchronously, the system timestamps each frame of the crowd dynamics signal and each local pressure distribution map with millisecond precision. The data processing center packages the two types of data with the same timestamp into a single data unit. This data unit comprehensively records the macroscopic movement state and microscopic physical pressure distribution of the crowd at a specific moment. These data units are arranged chronologically, forming a continuous data stream—a multi-source fusion situational awareness data stream characterizing the on-site physical environment and crowd behavior. The spatial microphone is synchronously activated in this step and begins collecting audio data, but its data stream is temporarily independent and used for analysis in subsequent steps.

[0027] Key nodes in evacuation routes refer to geographical locations that decisively impact pedestrian flow efficiency in emergency evacuation scenarios, such as intersections, width changes, or exits. Millimeter-wave radar is a sensor that uses millimeter-wave electromagnetic waves for detection. Its function is to non-contactly perceive the distance, speed, and angle of targets over a large area, and in this case, it is used to capture the macroscopic motion characteristics of crowds. A pressure sensor array is a device composed of multiple pressure sensing units arranged in a matrix, embedded in the ground, and its function is to measure and record detailed pressure distribution within its coverage area in real time. A spatial microphone is a device deployed within the monitoring area to pick up ambient sound, and its function is to collect acoustic information from the scene. The crowd dynamics signal is a structured time-series dataset. Each record contains a timestamp, a vector field describing the average movement speed of the crowd, and a trend index quantifying the degree of internal pushing within the crowd. This index is based on the analysis of over one hundred simulated evacuation drill videos. The local pressure distribution map is a two-dimensional data matrix, where each element represents the pressure magnitude at the corresponding location on the ground, measured in Pascals. Its high precision is reflected in the physical spacing between sensor units being less than or equal to 0.1 meters. The multi-source fusion situational awareness data stream is a composite data structure that uses timestamps as the primary key to link group dynamic mechanical signals and local pressure distribution maps at the same moment, forming a continuous data sequence that provides a comprehensive and synchronized data foundation for situational analysis.

[0028] S2. Construction of a composite crisis state model: Based on multi-source fusion of situational awareness data streams, a composite crisis state model is generated that includes the location of the danger source, the precise coordinates of the congestion bottleneck, and the risk level of panic spread.

[0029] In a specific embodiment of the present invention, the step of generating a composite crisis state model that includes the location of the hazard source, the precise coordinates of the congestion bottleneck, and the risk level of panic spread includes: identifying the congestion bottleneck by cross-validating the stagnant movement region in the group dynamics signal with the high-load hotspot region in the local pressure distribution map.

[0030] By analyzing the dynamic mechanical signals of the population in the upstream area of ​​the congestion bottleneck and combining them with the abnormal acoustic features in the audio signals collected by the spatial microphone, the worsening trend of congestion and the level of panic among the population are quantitatively assessed.

[0031] In a specific embodiment of the present invention, the specific method for quantitatively assessing the worsening trend of congestion and the level of panic among the population is as follows: the worsening trend score of congestion is calculated based on the weighted sum of the rate of change of the average movement speed of the population in the upstream area of ​​the congestion bottleneck and the pushing trend index.

[0032] The panic level score of the crowd is calculated by weighting the maximum sound decibel value collected by the spatial microphone with the number of identified screaming events.

[0033] Based on the quantitative assessment results, a composite crisis state model is generated.

[0034] It should be noted that this step receives and processes the multi-source fusion situational awareness data stream generated in the first step, which represents the on-site physical environment and the state of crowd behavior. Through in-depth analysis and data fusion, it aims to identify potential dangers, assess the degree of crisis, and ultimately build a structured data model to describe the current crisis state.

[0035] The processing flow begins with the input of the data stream. The system continuously receives crowd dynamics signals and local pressure distribution maps, each containing a synchronization timestamp. For each data unit, the system performs cross-validation to pinpoint congestion points. It first searches for stagnant areas in the crowd dynamics signals, achieved by setting a speed threshold, for example, marking areas with a movement speed below 0.5 meters per second as stagnant. Simultaneously, it searches for high-load hotspots in the corresponding local pressure distribution maps, achieved by setting a pressure threshold, for example, marking areas with pressure values ​​exceeding 3000 Pascals as high-load. When a geographical location is marked as both a stagnant area and a high-load hotspot at the same timestamp, the system identifies the center coordinates of this location as a congestion bottleneck with a potential risk of stampede.

[0036] After identifying the congestion bottleneck, the system shifts its focus to the upstream area, i.e., the direction from which the crowd originates. The system retrieves the dynamic mechanical signals of the crowd in this upstream area, analyzing the rate of change of crowd movement speed over time and the increase in the pushing trend index to determine whether the congestion is easing or worsening. Simultaneously, the system retrieves audio signals collected by spatial microphones deployed in the area, extracting abnormal acoustic features through signal processing algorithms. These features include whether the sound decibel level exceeds the preset 85-decibel safety threshold, whether there are human screams with frequencies above 2000 Hz, and whether the instantaneous rate of change in sound loudness is drastic. Combining the physical worsening trend of congestion with signs of panic in the audio, the system quantitatively assesses the worsening trend of congestion and the level of panic among the crowd, providing a numerical score for each.

[0037] Finally, the system integrates the analysis results from the first two steps to generate a structured data object, namely a composite crisis state model. This model explicitly includes textual descriptions of the hazard sources, such as "escalator exit"; precise coordinates of congestion bottlenecks, such as a numerical pair containing latitude and longitude or internal venue coordinates; and a panic spread risk level that integrates the trend of congestion worsening and the emotional state of the crowd, such as a five-level assessment from "Level 1 - No Risk" to "Level 5 - Extremely High Risk". This model provides a clear and quantitative basis for subsequent decision-making and intervention.

[0038] This step involves a quantitative assessment of the congestion worsening trend and the level of panic. To ensure the reproducibility of the technology, the following exemplary assessment formula is introduced. The formula for calculating the congestion worsening trend D_trend is D_trend = w1*(V_avg_prev - V_avg_curr) / Δt + w2*P_trend, where D_trend represents the congestion worsening trend score. V_avg_curr and V_avg_prev represent the average movement speed of the crowd in the upstream area of ​​the congestion bottleneck at the current and previous moments, respectively, in meters per second. Δt is the time interval between the two moments, in seconds. P_trend is the pushing trend index extracted from the crowd dynamics signal. w1 and w2 are dimensionless weighting coefficients used to balance the contribution of speed changes and pushing trends to the overall assessment. They are set based on statistical analysis of 50 real congestion event videos, for example, setting w1 = 0.6 and w2 = 0.4.

[0039] The formula for calculating the panic level P_level is: P_level = c1 * (dB_max - dB_thresh) + c2 * N_scream, where P_level represents the panic level score of the crowd and is dimensionless. dB_max is the maximum decibel value of sound collected by the spatial microphone during the assessment period. dB_thresh is a preset decibel threshold for normal ambient sound, such as 70 decibels. N_scream is the number of screaming events identified through audio features. c1 and c2 are dimensionless weighting coefficients used to adjust the importance of decibel value and number of screams in panic assessment. Their settings are based on psychoacoustic experimental data, for example, setting c1 = 0.7 and c2 = 0.3. To ensure that the formula results are within a standardized level range, the calculation results can be normalized.

[0040] The congestion bottleneck is a specific geographical coordinate point identified by the system, characterized by extremely low crowd movement speed and extremely high physical pressure, posing a high risk of stampedes. Abnormal acoustic features refer to special sound signals that differ from normal background noise, including but not limited to sudden increases in volume, high-frequency screams or cries; these features are extracted from the original audio stream using audio signal processing techniques. The congestion worsening trend is a quantitative indicator used to describe the speed and severity of congestion deteriorating over time. The crowd panic level is a quantitative indicator used to assess the collective emotional state of the crowd; a higher level indicates greater panic and more unpredictable behavior. The composite crisis state model is a structured dataset that encapsulates core information about the crisis event, including the location of the danger, its precise coordinates, and the severity level of the risk, serving as direct input for subsequent emergency response.

[0041] S3. Cooperative Guidance Command Generation: Based on the composite crisis state model, generate a set of spatially differentiated cooperative guidance commands that include light effect patterns, flow speed, and audio content.

[0042] In a specific embodiment of the present invention, the step of generating a set of spatially differentiated collaborative guidance instructions that include light effect modes, flow speeds and audio content includes: matching and activating a dynamic diversion strategy from a preset contingency plan library based on the coordinates of congestion bottleneck points and the risk level of panic spread in the composite crisis state model.

[0043] Based on the dynamic diversion strategy, visual deceleration guidance instructions are generated for the upstream area of ​​the congestion bottleneck, and visual acceleration guidance instructions are generated for a preset backup safety path.

[0044] Based on the dynamic diversion strategy, targeted voice reassurance instructions with different content are generated for different areas, and together with visual deceleration guidance instructions and visual acceleration guidance instructions, they form spatially differentiated collaborative guidance instructions.

[0045] It should be noted that this step aims to transform the abstract crisis assessment generated in the second step into a concrete and actionable on-site guidance plan. The process begins by receiving the composite crisis state model output from the previous step. Once the system acquires this model, it immediately analyzes its internal data, focusing on extracting two core parameters: the precise coordinates of congestion bottlenecks and the risk level of panic spread.

[0046] These two parameters are used as query criteria to match a contingency plan database stored internally. This database is pre-configured by safety management experts based on building layout and various potential emergency situations, and contains a large number of dynamic diversion strategies. Each dynamic diversion strategy corresponds to a specific scenario, defined by the location of congestion and the risk level. The matching process is a precise search; the system looks for preset entries that perfectly match the precise coordinates of the current congestion bottleneck and the risk level of panic spread. Once a matching entry is found, the system activates the corresponding dynamic diversion strategy.

[0047] Initiating a dynamic traffic diversion strategy essentially involves loading a pre-defined set of rules and parameters. Based on these rules, the system generates specific guidance instructions. First, for the area upstream of the congestion bottleneck—the area from which crowds are heading—the system generates visual deceleration guidance instructions. This instruction is a data packet defining the lighting effects that the smart light strips in that area should display, such as color, brightness, and flashing pattern, as well as the speed and direction of the light effect's flow. Next, the system queries the building's electronic map to find one or more pre-planned alternative safe paths that bypass the current congestion bottleneck. For these alternative safe paths, the system generates visual acceleration guidance instructions, also in the form of data packets defining the lighting effects and faster flow speeds of the smart light strips along these paths.

[0048] While generating visual instructions, the dynamic triage strategy also generates targeted voice reassurance instructions with different content based on the function of different areas. For example, in the area upstream of the congestion bottleneck, the instructions may be reassuring, asking people to remain calm and slow down; while at the entrance of the alternative safety route, the instructions may be guiding, instructing people to turn and follow the new directions. These voice instructions will include the audio content to be played, the target broadcast area, and the suggested volume.

[0049] Ultimately, the system integrates the generated visual deceleration guidance commands, visual acceleration guidance commands, and all directional voice reassurance commands into a unified, structured set of commands. This set is a spatially differentiated collaborative guidance command that includes light effect modes, flow speeds, and audio content; it is packaged and ready to be sent to the hardware execution device in the next stage.

[0050] The technical terminology used in the connecting steps has already been explained and does not need to be repeated here. The contingency plan library is a structured database storing various emergency plans. Each plan exists as a dynamic diversion strategy, associated with specific congestion bottleneck coordinates and panic spread risk levels. Its settings are based on simulations of over 50 different emergency evacuation scenarios for the building. The dynamic diversion strategy is a specific set of action rules that, once activated, automatically generates combined instructions for slowing down, accelerating, and calming people based on the crisis situation. Visual slowing guidance instructions are control commands that instruct lighting equipment in specific areas to produce visual signals that warn and persuade people to slow down. Alternative safety routes are alternative channels reserved during building planning for crowd control when the main evacuation routes are blocked; this information is stored in the system's building information model. Visual accelerating guidance instructions are control commands that instruct lighting equipment in specific areas to produce visual signals that attract people and guide them to pass quickly. Directional voice calming instructions are control commands that specify specific audio playback devices, content, and volume, designed to deliver calming or guiding information to people in specific areas. Spatially differentiated collaborative guidance instructions are a final output composite data package that contains detailed parameters for all visual and auditory guidance instructions for different spatial locations, ensuring the synergy and regional adaptability of guidance measures.

[0051] S4. Field Effect Guiding Environment Deployment: Receives spatially differentiated collaborative guidance instructions to deploy an acoustic-optical linked field effect guiding environment.

[0052] In a specific embodiment of the present invention, the step of deploying an acoustic-optical linkage field effect guidance environment includes: driving a smart light strip in the upstream area of ​​the congestion bottleneck to execute a visual deceleration guidance command to construct a visual deceleration field.

[0053] Drive the intelligent light strip on the backup safety path to execute visual acceleration guidance commands to construct a visual acceleration field.

[0054] Activate directional speakers to accurately deliver directional voice reassurance commands, which, together with visual deceleration and visual acceleration fields, form a sound-light-interactive field effect guidance environment.

[0055] It's important to note that this step transforms the abstract instructions generated in step three into a perceptible guidance environment in the physical world. Its core lies in precisely driving various intelligent devices deployed on-site to achieve a non-coercive, immersive influence on crowd behavior. The process begins with a central control system receiving spatially differentiated collaborative guidance instructions generated in the previous step. This set of instructions is parsed by the system and broken down into specific control commands for different devices and areas.

[0056] First, the system filters visual deceleration guidance instructions for the area upstream of the congestion bottleneck based on the identifiers in the command. This instruction includes the target smart light strip's device ID, the color code to be displayed, the lighting effect mode, and a slow flow speed parameter. The control system sends this command via wired or wireless network to the controllers of the smart light strips deployed on the ground in the area upstream of the congestion bottleneck. Upon receiving the command, these controllers immediately activate the LED light strips they manage, causing them to display an overall yellow color and flow at the slow speed set in the command, in the opposite direction to the direction of the crowd's movement. This visual effect psychologically conveys a warning and a signal to the crowd behind that they need to slow down, thus creating a visual deceleration field.

[0057] While executing the deceleration command, the system also parses a visual acceleration guidance command for the backup safety path. This command also includes the target smart light strip's device ID, color code, light effect mode, and a rapid flow speed parameter. The control system sends this command to the smart light strip controllers laid on the backup safety path. These light strips then display a striking green and flow at a faster speed along the path towards the safety exit. This dynamic green light flow has a strong attraction and directional indication, effectively guiding the crowd's gaze and footsteps, prompting them to converge on this new evacuation route, thus creating a visual acceleration field.

[0058] Simultaneously, the system also processes directional voice reassurance commands. These commands include the coordinates of the target area, the audio content to be played, and the specified volume. The control system activates directional speakers deployed in different areas and sends the corresponding audio commands there. Utilizing sound wave focusing technology, the directional speakers concentrate sound energy into a narrow beam for propagation, thus accurately delivering reassurance messages such as "There's a crowd ahead, please slow down" to people in congested areas without interfering with those in adjacent alternative passages. Similarly, guidance messages such as "Please follow the green light strip" are accurately delivered to the entrance of new passages. These precisely delivered sound and light guidance effects work together to create a coordinated sound and light field effect guidance environment, managing crowd behavior collaboratively and differentiatedly through the dual effects of vision and hearing.

[0059] The technical terms used in the connecting steps have already been explained and do not need to be repeated here. A smart light strip is a programmable linear light-emitting diode lighting device characterized by its ability to be embedded in the ground and withstand foot traffic, and its ability to dynamically change color, brightness, and lighting effects. A visual deceleration field is a perception area created through specific lighting effects, such as slowly flowing warm light, which subtly encourages people within the area to slow down their movement using principles of visual psychology. A visual acceleration field is a perception area created through specific lighting effects, such as rapidly flowing cool light, which attracts people's attention and clearly indicates their direction and speed of movement. A directional speaker is a loudspeaker device that can focus sound onto a specific direction and area, characterized by rapid sound attenuation outside the designated area, thus achieving precise sound transmission. A field-effect guided environment is a comprehensive, immersive guided system that organically combines various guided elements such as visual and acoustic fields, and its function is to influence the macroscopic behavior of people by creating an environmental atmosphere with a clear guiding intent.

[0060] S5. Physical Feedback Signal Generation: During the field effect-guided environmental action, a physical feedback signal representing the effect of command execution is generated by monitoring the physical response of the crowd.

[0061] In a specific embodiment of the present invention, the step of generating a physical feedback signal of the flow of people characterizing the execution effect of the instruction includes: using a pressure sensor array to monitor the degree of reduction of the physical load of the crowd in the visual deceleration field and the gradient of the change in the trampling pressure of the crowd in the visual acceleration field.

[0062] Using millimeter-wave radar, we can track the easing of the crowd pushing trend in the visually decelerating field, as well as the increase in the overall movement speed of the crowd in the visually accelerating field.

[0063] The system integrates the degree of reduction in physical load, the gradient of changes in trampling pressure, the easing of pushing trends, and the increase in movement speed to generate physical feedback signals of pedestrian flow.

[0064] It should be noted that the core task of this step is to quantify and evaluate the actual effectiveness of the guidance measures through continuous data collection during the operation of the field-effect guidance environment constructed in step four. This is the perception link of a closed-loop feedback system, and its implementation process begins the moment the guidance command is issued and continues continuously.

[0065] In the visual deceleration field, the area upstream of the congestion bottleneck, the system continuously utilizes an array of pressure sensors embedded in the ground. These sensors constantly measure pressure data within their coverage area, and the system compares this real-time data with baseline pressure data before the guidance environment was activated. By calculating the difference between the current average physical load and the baseline value, the system can obtain a quantified degree of reduction in physical load. For example, it calculates the average of all sensor readings across the entire area and tracks the percentage decrease in this average over time. Simultaneously, pressure sensor arrays located in the visual acceleration field, the newly opened backup safety path, are also active. The system analyzes the data from these sensors, focusing not on the absolute value of pressure, but on the changes in pressure over time and space. Specifically, the system calculates the rate of change of pressure readings along the direction of pedestrian movement, i.e., the gradient of trampling pressure. A positive and stable gradient indicates that the crowd is continuously and orderly passing through the path, rather than forming new stagnations.

[0066] Simultaneously, millimeter-wave radar continuously tracks these two key areas. Within the visual deceleration field, the radar continuously analyzes the crowd's moving speed vector field and calculates the pushing trend index. The system compares the current pushing trend index with the peak index before guidance, thus quantifying the easing of the pushing trend. A continuously decreasing index indicates that the pressure from behind on the front crowd is decreasing, and the deceleration guidance is effective. Within the visual acceleration field, the millimeter-wave radar focuses on measuring the overall moving speed of the crowd. The system calculates the average speed of the crowd in this area and compares it with the almost zero speed before guidance, obtaining a clear value indicating an increase in moving speed. This value directly reflects the effectiveness of the diversion.

[0067] Finally, the system integrates four key data points from two regions and two types of sensors: physical load, trampling pressure, pushing trend, and movement speed. The integration method involves treating these data points collected at the same time point as a single data record unit. This data record unit is assigned a precise timestamp and then added to a time-series database. This continuously generated data stream, containing multi-dimensional quantitative indicators, serves as the final output, representing the physical feedback signal of the pedestrian flow and characterizing the effectiveness of the instruction execution. This provides direct data input for the next step of effectiveness verification.

[0068] The technical terminology used in the connecting steps has already been explained and does not need to be repeated here. The physical feedback signal of the crowd is a structured time-series data stream, which functions to describe the crowd's response to the field-effect guided environment in real time and quantitatively. Each record in this data stream contains a timestamp, along with the corresponding reduction in physical load, the gradient of trampling pressure, the easing of pushing trends, and the increase in movement speed. These data together constitute a comprehensive measure of the effectiveness of the guidance strategy.

[0069] S6. Generation of Closed-Loop Validity Verification Results: Input the physical feedback signal of the pedestrian flow to generate a closed-loop validity verification result of the guidance strategy to determine whether the current guidance strategy is successful.

[0070] In a specific embodiment of the present invention, the step of generating a closed-loop validity verification result of the guidance strategy to determine whether the current guidance strategy is successful includes: calculating the correlation between the degree of reduction of physical load in the visual deceleration field and the easing of the pushing trend, so as to evaluate the execution compliance of the deceleration command.

[0071] The computational vision-based acceleration of crowd movement speed within the field is compared with the achievement rate of a preset diversion efficiency target to evaluate the actual effect of diversion instructions.

[0072] Based on compliance and achievement rate, a closed-loop effectiveness verification result for the guidance strategy is generated.

[0073] It should be noted that this step is an automated evaluation process aimed at quantitatively assessing the effectiveness of the currently implemented guidance strategy. The process begins with the system receiving and parsing the physical feedback signal of the pedestrian flow generated in step five. This signal is a time series containing multi-dimensional real-time data.

[0074] First, the effectiveness of the deceleration command is evaluated. It extracts two key time-series data points from the physical feedback signals of the crowd: the degree of reduction in physical load within the visual deceleration field and the easing of the pushing trend. The system uses statistical methods to calculate the correlation between these two time series over a period of time, such as by calculating the Pearson correlation coefficient. If the physical load decreases while the pushing trend also eases, showing a strong positive correlation, this indicates that the crowd is actively responding to the deceleration guidance, and those behind are no longer blindly pushing forward. This calculated correlation coefficient is quantified as the compliance rate with the deceleration command.

[0075] Next, the system evaluates the actual effectiveness of the diversion command. It extracts the increase in the overall movement speed of the crowd within the visual acceleration zone from the physical feedback signals of the pedestrian flow. Simultaneously, the system reads a corresponding preset diversion efficiency target from the dynamic diversion strategy initiated in step three; this target is typically defined by the desired average movement speed. The system compares the actual monitored increase in movement speed with this preset target and calculates a percentage, i.e., the effectiveness achievement rate. This ratio directly reflects the actual contribution of the newly opened backup safety routes to crowd control.

[0076] Finally, the system makes a comprehensive judgment based on two core indicators calculated in the first two steps: compliance with the deceleration command and the effectiveness achievement rate. The system internally sets two thresholds: one for compliance and one for achievement rate. For example, compliance must be greater than 0.8, and the effectiveness achievement rate must be higher than 75%. Only when both indicators simultaneously meet or exceed their respective preset thresholds will the system determine that the current dynamic traffic diversion strategy is successful. Conversely, if either indicator fails to meet the standard, the strategy is deemed to have failed to effectively alleviate congestion. This final judgment is encapsulated into a clear data marker—the closed-loop effectiveness verification result of the guidance strategy—and output to the next processing stage.

[0077] To ensure the clarity and reproducibility of the correlation and achievement rate calculations, this step introduces the following formula: The calculation formula for the compliance of the deceleration command, C_compliance, is C_compliance=Cov(L,S) / (σ_L*σ_S), which is a Pearson correlation coefficient calculation formula, where C_compliance represents the compliance of the deceleration command. L is a time series representing the degree of reduction in physical load; S is a time series representing the easing of the pushing trend. Cov(L,S) is the covariance of the two time series. σ_L and σ_S are the standard deviations of the two time series, respectively. This formula is based on standard statistical theory and is used to measure the strength of the linear correlation between two variables. The calculation formula for the effect achievement rate, A_rate, is A_rate=(V_actual / V_target)*100%, where A_rate represents the effect achievement rate in percentage. V_actual is the value of the increase in crowd movement speed obtained from the physical feedback signal of the pedestrian flow, in meters per second. V_target is a preset target speed value for diversion efficiency read from the current dynamic diversion strategy, also in meters per second. Its setting is based on building safety evacuation regulations and historical exercise data. The formula uses normalization to convert the actual speed into a percentage of the target's completion.

[0078] The technical terminology used in the connecting steps has already been explained and does not need to be repeated here. The compliance rate of the deceleration instruction is a value between -1 and 1, used to quantify the consistency between crowd behavior and the intention to slow down. The closer the value is to 1, the better the compliance. The preset diversion efficiency target is a key performance indicator set when formulating a dynamic diversion strategy. It defines the ideal pedestrian flow speed that the backup safe path should achieve in a diversion scenario. The effectiveness achievement rate is a percentage value that intuitively represents how close the actual effectiveness of the diversion instruction is to the expected goal. The closed-loop effectiveness verification result of the guidance strategy is a Boolean or enumerated data output. Its function is to provide the system with a clear conclusion about whether the current strategy is successful, such as "success" or "failure," as a basis for subsequent adaptive adjustments.

[0079] S7. Adaptive Adjustment of Guiding Environment: Receives the closed-loop validity verification result of the guiding strategy. When the strategy is determined to fail, it performs dynamic, closed-loop adaptive adjustment of the field effect guiding environment.

[0080] In a specific embodiment of the present invention, the step of performing dynamic, closed-loop adaptive adjustment of the field effect guidance environment includes: if the closed-loop effectiveness verification result of the guidance strategy determines that the current strategy has failed to effectively alleviate congestion, then the warning intensity of the visual deceleration field is increased.

[0081] The operation to enhance the warning intensity is achieved by increasing the flashing frequency of the yellow light effect within the visual deceleration field and simultaneously increasing the volume of the directional voice reassurance instructions broadcast in that area.

[0082] It should be noted that this step is the closed-loop control and self-optimization stage of the entire guidance system. Its core is to dynamically adjust the on-site guidance measures based on the evaluation results of the previous stage to cope with complex and ever-changing crowd behavior. The starting point of the process is to receive and analyze the closed-loop effectiveness verification results of the guidance strategy output in step six.

[0083] The system first checks the result. If the result indicates that the current strategy is successful, the system will maintain the current field-effect guidance environment and continue the monitoring and evaluation cycle. However, if the result indicates that the current strategy has failed to effectively alleviate congestion, the system will immediately initiate an adaptive adjustment procedure.

[0084] As the first level of response to the adjustment, the system automatically enhances the warning intensity of the visual deceleration field. This operation is specific and quantifiable. The system sends an update command to the smart light strip controllers deployed upstream of the congestion bottleneck. In this command, the flashing frequency parameter of the yellow light effect will be increased, for example, from not flashing or flashing once per second to flashing three times per second. This more rapid flashing is visually more alert and can more strongly prompt people to notice the danger and slow down. At the same time, the system also sends an update command to the directional speakers in the area, increasing the broadcast volume parameter of the directional voice reassurance instructions, for example, from 65 decibels to 75 decibels, ensuring that the voice prompts can penetrate the noisy environment and be more clearly delivered to the crowd.

[0085] After the first level of adjustment, the system does not stop immediately but re-enters an observation and evaluation cycle. It will repeat steps five and six, collecting new physical feedback signals of pedestrian flow and generating new verification results for the closed-loop effectiveness of the guidance strategy. If, after a preset period of time, the new verification results show that the strategy remains ineffective—meaning the enhanced guidance measures have still failed to achieve the expected results—the system will initiate a second-level response.

[0086] In the second-level response, the system takes more decisive action. Within its internal electronic map and route planning module, it marks the path leading to the current congestion bottleneck as temporarily blocked. This marking means that this path will be temporarily excluded from any subsequent route planning calculations and will no longer be considered a viable option. Immediately following, the system triggers a completely new route search calculation, replanning a new, safe alternative route based on real-time pedestrian flow distribution and available passage information. Once the new route is determined, the system generates a new set of spatially differentiated collaborative guidance instructions and sends them to the fourth step, thereby driving a new visual acceleration field and corresponding voice guidance, achieving dynamic, closed-loop adaptive adjustment of the field-effect guidance environment. This process ensures that the guidance system does not adhere to ineffective strategies but can flexibly change and optimize its guidance scheme according to the actual situation.

[0087] The technical terms used in the connection steps have already been explained and do not need to be repeated here. Adaptive adjustment refers to the process by which a system can automatically modify its behavior based on external feedback to achieve better performance. Temporary blocking is an internal system status flag used in path planning algorithms to temporarily disable a path, preventing pedestrian flow from being further guided to areas that have been confirmed as uncontrollable.

[0088] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A safety intelligent guidance method based on scenario analysis and emergency plan triggering, characterized in that, Includes the following steps: S1. Situational Awareness Data Stream Generation: Generate a multi-source fusion situational awareness data stream that represents the on-site physical environment and the state of crowd behavior. S2. Construction of a composite crisis state model: Based on multi-source fusion of situational awareness data streams, a composite crisis state model is generated that includes the location of the danger source, the precise coordinates of the congestion bottleneck, and the risk level of panic spread. S3. Cooperative Guidance Command Generation: Based on the composite crisis state model, generate a set of spatially differentiated cooperative guidance commands that include light effect patterns, flow speed, and audio content. S4. Field Effect Guiding Environment Deployment: Receive spatially differentiated collaborative guidance instructions to deploy an acoustic-optical linked field effect guiding environment; S5. Physical Feedback Signal Generation: During the field effect-guided environmental action, a physical feedback signal representing the effect of command execution is generated by monitoring the physical response of the crowd. S6. Generation of closed-loop validity verification results: Input the physical feedback signal of the pedestrian flow to generate a closed-loop validity verification result of the guidance strategy to determine whether the current guidance strategy is successful. S7. Adaptive Adjustment of Guiding Environment: Receives the closed-loop validity verification result of the guiding strategy. When the strategy is determined to fail, it performs dynamic, closed-loop adaptive adjustment of the field effect guiding environment.

2. The safety intelligent guidance method based on scenario analysis and emergency plan triggering according to claim 1, characterized in that: The steps for generating a multi-source fused situational awareness data stream characterizing the on-site physical environment and the state of crowd behavior include: By scanning the crowd with millimeter-wave radar, a dynamic mechanical signal of the group can be generated. A high-precision local pressure distribution map is generated by collecting data on the trampling load of people through a pressure sensor array. By fusing group dynamic mechanical signals with local pressure distribution maps and aligning timestamps, a multi-source fused situational awareness data stream is generated.

3. The safety intelligent guidance method based on scenario analysis and emergency plan triggering according to claim 2, characterized in that: The steps for generating a composite crisis state model that includes the location of the hazard source, the precise coordinates of the congestion bottleneck, and the risk level of panic spread include: By cross-validating the stagnant regions in the dynamic mechanical signals of the population with the high-load hotspot regions in the local pressure distribution map, congestion bottlenecks can be identified. By analyzing the dynamic mechanical signals of the population in the upstream area of ​​the congestion bottleneck and combining them with the abnormal acoustic features in the audio signals collected by the spatial microphone, the worsening trend of congestion and the level of panic among the population are quantitatively assessed. Based on the quantitative assessment results, a composite crisis state model is generated.

4. The safety intelligent guidance method based on scenario analysis and emergency plan triggering according to claim 3, characterized in that: The specific method for quantitatively assessing the worsening trend of congestion and the level of panic among the public is as follows: The congestion worsening trend score is calculated by weighting the rate of change of the average movement speed of people in the upstream area of ​​the congestion bottleneck with the pushing trend index. The panic level score of the crowd is calculated by weighting the maximum sound decibel value collected by the spatial microphone with the number of identified screaming events.

5. The safety intelligent guidance method based on scenario analysis and emergency plan triggering according to claim 1, characterized in that: The steps for generating a set of spatially differentiated collaborative guidance instructions that include light effect patterns, flow speed, and audio content include: Based on the coordinates of congestion bottlenecks and the risk level of panic spread in the composite crisis model, a dynamic diversion strategy is matched and activated from a pre-set contingency plan library. Based on the dynamic diversion strategy, visual deceleration guidance instructions are generated for the upstream area of ​​the congestion bottleneck, and visual acceleration guidance instructions are generated for a preset backup safety path. Based on the dynamic diversion strategy, targeted voice reassurance instructions with different content are generated for different areas, and together with visual deceleration guidance instructions and visual acceleration guidance instructions, they form spatially differentiated collaborative guidance instructions.

6. The safety intelligent guidance method based on scenario analysis and emergency plan triggering according to claim 5, characterized in that: The steps for deploying an acoustically-optically-linked field-effect guided environment include: The intelligent light strip in the upstream area of ​​the congestion bottleneck point executes visual deceleration guidance commands to create a visual deceleration field; Drive the intelligent light strip on the backup safety path to execute visual acceleration guidance commands to construct a visual acceleration field; Activate directional speakers to accurately deliver directional voice reassurance commands, which, together with visual deceleration and visual acceleration fields, form a sound-light-interactive field effect guidance environment.

7. A safety intelligent guidance method based on scenario analysis and emergency plan triggering according to claim 6, characterized in that: The step of generating a physical feedback signal of the flow of people characterizing the effect of instruction execution includes: Using a pressure sensor array, the degree of reduction in the physical load of the crowd in the visual deceleration field and the gradient of the trampling pressure of the crowd in the visual acceleration field are monitored. Using millimeter-wave radar, we can track the easing of the crowd pushing trend in the visually decelerating field, as well as the increase in the overall movement speed of the crowd in the visually accelerating field. The system integrates the degree of reduction in physical load, the gradient of changes in trampling pressure, the easing of pushing trends, and the increase in movement speed to generate physical feedback signals of pedestrian flow.

8. The safety intelligent guidance method based on scenario analysis and emergency plan triggering according to claim 7, characterized in that: The step of generating a closed-loop validity verification result for determining whether the current boot strategy is successful includes: The correlation between the degree of reduction in physical load within the visual deceleration field and the easing of the pushing trend is calculated to assess the compliance with deceleration commands. The computational vision-accelerated crowd movement speed improvement value and the achievement rate of a preset diversion efficiency target are compared to evaluate the actual effect of the diversion command. Based on compliance and achievement rate, a closed-loop effectiveness verification result for the guidance strategy is generated.

9. A safety intelligent guidance method based on scenario analysis and emergency plan triggering according to claim 1, characterized in that: The steps for performing dynamic, closed-loop adaptive adjustments to the field-effect guided environment include: If the closed-loop effectiveness verification result of the guidance strategy determines that the current strategy has failed to effectively alleviate congestion, then the warning intensity of the visual deceleration field will be increased. The operation to enhance the warning intensity is achieved by increasing the flashing frequency of the yellow light effect within the visual deceleration field and simultaneously increasing the volume of the directional voice reassurance instructions broadcast in that area.

Images