Fire safety evacuation management and control system for large-scale commercial complex

Abstract

The invention relates to the technical field of fire control, and particularly discloses a fire safety evacuation control system for a large-scale commercial complex, which is used for solving the problems of vertical penetration of multiple floors, rapid diffusion of flue gas, poison gas and high temperature, sudden reduction of visibility, high people flow density, channel capacity bottleneck and easy congestion and backflow in a fire disaster of the large-scale commercial complex. The system comprises an intelligent evacuation indicating lamp board, an intelligent illuminating lamp, an environment information acquisition device, an evacuation management and control platform and a management terminal. According to the invention, through fusion of toxic gas, temperature, visibility and video people flow density and speed, dynamic double-target constraint optimization is constructed, Pareto main and standby routes are output, intelligent evacuation indication boards and zoning light modulation shunting are linked, and self-adaptive guidance along with fire scene time-varying risks and congestion bottlenecks is realized. Backflow trampling and suffocation heat damage are reduced; and the evacuation time is shortened.

Description

Technical Field

[0001] This invention relates to the field of fire control technology, and in particular to a fire safety evacuation control system for large commercial complexes. Background Technology

[0002] A Chinese invention patent with publication number CN117547752A discloses an intelligent evacuation guidance system. This system involves deploying intelligent evacuation indicator lights and optional intelligent lighting along roadsides within the factory area. An environmental information collection device gathers parameters such as wind speed and direction, toxic and harmful gases, and upstream and downstream status of equipment involved in accidents. An evacuation control platform then assigns safety level values ​​to areas based on their distance from the accident zone. Combining the shortest path evacuation principle and the hazard avoidance evacuation principle, multiple evacuation routes are generated for the management terminal to select and confirm. Commands are then issued to control the direction and brightness of the indicator lights, thereby enabling dynamic response to dynamic disasters, ensuring adequate nighttime illumination, and providing precise route guidance in petrochemical, coal, and other explosion or fire hazard locations. While achieving higher evacuation efficiency and safety than traditional fixed signage, fires in large commercial complexes are characterized by multiple floors, vertically connected atriums, complex zoning, changing fire door status, high pedestrian density and bottlenecks in passageways, rapid spread of smoke and high temperatures along ventilation and air conditioning systems and vertical shafts, and a sharp drop in visibility. If the core planning basis is still based on accident distance zoning and the shortest path for a single source, distance cannot characterize the time-varying risks of smoke, toxic gases, temperature, visibility, and congestion speed. Furthermore, it lacks dynamic constraints for congestion diversion and the availability of vertical safety exits, which can easily lead people to smoke recirculation areas or bottleneck stairwells, causing backflow, shoving, trampling, suffocation, poisoning, and even mass casualties. Summary of the Invention

[0003] The technical problem to be solved by this invention is to provide a fire safety evacuation control system for large commercial complexes. By integrating toxic gas, temperature, visibility and video pedestrian density and speed, a dynamic dual-objective constraint optimization is constructed and Pareto primary and backup routes are output. This system is linked with intelligent evacuation signs and zoned dimming and diversion to achieve adaptive guidance based on the time-varying risks and congestion bottlenecks in the fire scene, reducing backflow, trampling and asphyxiation heat injury and shortening evacuation time.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A fire safety evacuation control system for large commercial complexes includes intelligent evacuation indicator lights, intelligent lighting, environmental information acquisition devices, an evacuation control platform, and a management terminal. The environmental information acquisition devices include toxic gas sensors, temperature sensors, visibility sensors, and cameras. The toxic gas sensors include, but are not limited to, carbon monoxide sensors, hydrogen cyanide sensors, hydrogen chloride sensors, nitrogen oxide sensors, and sulfur dioxide sensors. This system monitors toxic gases, temperature, visibility, and dynamic video information within the large commercial complex in real time. The evacuation control platform acquires images of each evacuation route from the cameras and extracts image features such as the number of human targets, occupancy rate, and flow direction. It characterizes the crowd size, congestion level, and direction of movement in the evacuation routes in real time, while providing input for crowd density and passage speed estimation and subsequent evacuation route diversion decisions. The evacuation control platform calculates crowd density and estimates passage speed simultaneously. Based on the multi-story evacuation topology, it constructs a dynamic dual-objective function within a preset update cycle, consisting of the channel risk cost and passage time generated by toxic gases, temperature, visibility, and passage speed. Under channel capacity constraints, it employs an improved dung beetle optimization algorithm that incorporates Kent chaotic initialization and dynamic back-learning to solve for the Pareto candidate evacuation route set, which is then sent to the management terminal. This enhances the global search and escape from local optima capabilities, thereby quickly outputting a Pareto candidate evacuation route set that balances minimum safety exposure with shortest evacuation time and is passable. This enables adaptive diversion guidance for the time-varying risks and congestion bottlenecks of fires in large commercial complexes. After the management terminal selects an evacuation route, the evacuation control platform controls the intelligent evacuation indicator lights according to the hysteresis switching threshold and controls the intelligent lighting to guide the diversion and evacuation of personnel.

[0006] As a further embodiment of the present invention, intelligent evacuation indicator lights are arranged along the evacuation route at the junctions, corners, entrance halls, connecting corridors, entrance halls of evacuation stairwells, entrances to refuge areas, and one side of safety exit doors, with their display faces facing the direction of the flow of people; intelligent lighting is arranged along the evacuation route and stair sections on the side walls.

[0007] As a further embodiment of the present invention, the intelligent evacuation indicator sign includes a control module and a display module. The display module displays the main evacuation direction arrow, alternative direction indicators, current location indicators, target assembly point indicators, and no-entry warning indicators. The control module receives the main route, alternative routes, and their diversion weight parameters from the evacuation management platform, and dynamically switches the display direction of the display module according to changes in crowd density and passage speed within a preset refresh cycle. When the crowd density and passage speed of the corresponding channel of the main route are both within the threshold and the risk cost is the lowest, the main route arrow is displayed. When the crowd density of the main route exceeds the limit, the passage speed is lower than the threshold, or the risk cost increases and the alternative route meets the capacity margin, the display switches to the alternative route arrow and rotates between the main and alternative arrows or is zoned according to the diversion weight. When the main route or alternative route enters a prohibited area, the display direction is changed accordingly. When a traffic violation criterion (such as risk cost exceeding the upper limit, channel capacity severely exceeding the limit, or being blocked) is met, a detour arrow away from the prohibited direction is displayed with a prohibited traffic warning sign superimposed. A hysteresis threshold is set for switching to suppress backflow. The hysteresis threshold is set based on the density / speed change range of the main and backup routes, the risk cost difference, and the capacity margin, and is superimposed with the minimum hold time and the maximum switching frequency. The intelligent lighting and intelligent evacuation indicator lights adopt a linkage dimming strategy. The evacuation control platform includes a brightness control module. The brightness control module outputs brightness control commands to the intelligent lighting lights in the main route, backup route, and prohibited area according to the channel risk cost and crowd density. According to the evacuation topology relationship between the final evacuation route and the target area, the intelligent lighting lights are marked as main route lights, backup route lights, and prohibited area lights, and dimming control quantities are issued respectively.

[0008] As a further aspect of this invention, in the brightness control module, the dimming control quantity is either a PWM duty cycle setting value or a constant current drive current setting value. Specifically, the duty cycle of the main route lights is set to 70%–100%, and a uniformity constraint is applied to the brightness difference between adjacent lights to ensure the difference is no greater than 10%. This forms a continuous and uniform high-illuminance guide light band in the main evacuation channel, reducing the risk of misjudgment, stagnation, and tripping caused by dark areas and sudden changes in brightness. The PWM duty cycle of the backup route lights is set to 30%–60%, maintaining the identifiability and reachability of the backup channels without competing for attention with the main route, facilitating orderly traffic distribution according to the traffic distribution weight. With switching, the PWM duty cycle of the restricted area lights is set to 0-20% with a superposition frequency of 1-2Hz and a PWM duty cycle of 30%-70% for flashing control or intermittent lighting control according to a preset cycle. The low brightness and rhythmic flashing form a strong warning signal, suppressing people from accidentally entering the danger zone and reducing the glare interference of ineffective lighting on the smoky environment. The PWM duty cycle setting value is updated in a linear ramping manner within 1-3 seconds according to the changes in crowd density and channel risk cost, avoiding visual discomfort and indicator jitter caused by sudden brightness changes. At the same time, it can smoothly follow the changes in crowd density and risk, reducing backflow countermeasures and trampling causes.

[0009] As a further aspect of this invention, the evacuation control platform also includes an evacuation channel feature extraction module. After acquiring video images from cameras in each evacuation channel, this module performs distortion correction and adaptive defogging enhancement within a preset detection area. This eliminates scale and position errors caused by wide-angle / perspective distortion, ensuring consistent geometric references for subsequent people counting, occupancy calculation, and motion direction estimation. It also reduces false detections and misjudgments, improves target discernibility and boundary clarity in smoke, low-light, and contrast-reduced scenarios, and enhances the efficiency and continuity of human detection and tracking. This ensures stable extraction of human flow features even in a fire environment. Furthermore, it employs human target detection and multi-target tracking to perform inter-frame correlation on human targets to obtain stable IDs, avoiding duplicate counting and missed counts caused by frame-by-frame detection. This allows for continuous tracking of individuals and suppresses jitter caused by occlusion and short-term loss, making the number of human targets more accurate. The system approximates the actual number of people present. It then uses the statistical value of valid IDs within the waiting area as the number of human targets, improving the number of people estimation from an instantaneous frame count to a time-consistent number of valid targets. This reduces counting fluctuations during backflow, convergence, and congestion, providing reliable input for density and velocity estimation. The occupancy rate is calculated based on the overlap area ratio between the instance mask and the waiting area, using area occupancy to characterize the degree of congestion. This provides a compensation signal for cases where severe occlusion leads to an underestimation of the number of people, improving sensitivity and robustness to high-density crowd compression and accumulation. Based on the direction histogram of human target displacement vectors obtained from multi-target tracking, the system extracts the main channel direction and backflow identifiers, obtaining the number of human targets, occupancy rate, and flow characteristics. It quantifies the crowd movement trend as the main direction and opposing / backflow status, enabling timely identification of collisions, reverse flows, and stagnant backflows. This provides a discriminative basis for diversion control and direction switching, reducing misguidance and the risk of stampedes.

[0010] As a further aspect of this invention, the evacuation channel feature extraction module is connected to a traffic speed estimation module. The traffic speed estimation module performs geometric calibration of the candidate area from pixels to metric coordinates and obtains the actual area of ​​the candidate area, converting the number of pixels in the video into a real spatial scale. This ensures that density and speed have consistent physical dimensions and are comparable, avoiding cross-channel errors caused by differences in lens perspective and installation height. Based on the number of human targets and the actual area, the crowd density is calculated, converting the number of people into a density index that directly reflects the degree of congestion. This quantifies the channel congestion status and provides a definite input for capacity constraints, risk costs, and diversion decisions. When the occlusion rate or tracking loss rate exceeds a threshold, occupancy rate and density map integrals are introduced to fuse and correct the crowd density. When high-density occlusion or smoke interference leads to missed detections or tracking, area occupancy and density map counts are used to compensate for the number of people estimated, improving the robustness and continuity of density estimation and reducing the erroneous release of bottleneck channels due to underestimated density. Finally, based on the inter-frame displacement of human targets in the metric coordinates and after sliding window filtering, an individual speed set is obtained. Walking speed is inverted from actual displacement and time smoothing is used to suppress jitter and instantaneous errors, making the speed estimation stable and noise-resistant, avoiding speed spikes caused by tracking jumps. The median of the individual speed set is taken as the channel passage speed. The median is used to resist outliers and interference from a few fast / stationary individuals, resulting in a channel speed that better represents the actual passage status of most people, thus more accurately assessing passage time and congestion level. When the image quality index is lower than a preset threshold, the channel passage speed obtained from human target displacement and the model speed obtained from the density-velocity constraint model input by crowd density are calculated separately. The confidence level generated by the image quality index is used as a weight to weight and fuse the channel passage speed obtained from human target displacement and the model speed obtained from the density-velocity constraint model input by crowd density. When smoke obscuration and low illumination cause the reliability of displacement speed to decrease, the model speed obtained from density constraint is introduced as compensation, and adaptive fusion is performed according to quality confidence level to ensure that the passage speed can still be output without distortion under poor imaging conditions, thus obtaining the final channel passage speed estimate.

[0011] As a further aspect of this invention, the passage speed estimation module is connected to an evacuation path solving module. This module constructs the multi-story evacuation space of the commercial complex into a directed evacuation topology graph composed of nodes and passage edges, structuring the complex building space into a computable road network model. It explicitly expresses the connectivity between floors and the directionality of passageways, and constructs a dynamic bi-objective function for any candidate evacuation route within a preset update cycle. The first objective is to minimize evacuation passage time, and the second objective is to minimize path risk exposure. This ensures that route planning is updated in real-time with time-varying information such as toxic gases, temperature, visibility, and passage speed, avoiding the lag risk of using static routes that lead to smoke diffusion and subsequent stagnation along the original path. This improves the adaptability to changes in the fire situation, suppresses the concentration of crowds in bottleneck passages, reduces the probability of queuing backflow, collisions, and stampedes, and ensures that the routes are physically... The algorithm is feasible in terms of traffic capacity. The channel capacity constraint is that the expected pedestrian flow on any channel side passed by the candidate route during the update period is not greater than the maximum throughput capacity determined by the effective net width and the unit width throughput capacity, or the crowd density on the channel side does not exceed the preset density limit. Under the channel capacity constraint, an improved dung beetle optimization algorithm is adopted to generate an initial candidate route population by introducing Kent chaotic mapping and to generate reverse candidate routes by using dynamic back learning. This algorithm performs non-dominated sorting to output a Pareto candidate evacuation route set, which improves the coverage and diversity of the initial search. This allows the algorithm to explore extensively in combinations of multiple exits, multiple stairwells, and multiple corridors, reducing the risk of early convergence to local shortest paths and ignoring safer diversion routes. It can quickly expand the search range when there is a local optimum or a sudden change in the fire situation (blocking / smoke spread / sudden increase in congestion), enhance the ability to escape local optima and quickly find alternative routes, and improve the stability and resistance to disturbances of the planning.

[0012] As a further aspect of this invention, the management terminal receives a set of Pareto candidate evacuation routes and visualizes them using evacuation time and risk exposure indicators for each candidate route as evaluation dimensions. It also displays the floor sequence, key node sequence, bottleneck channel markers, and channel capacity margin of each candidate route. A primary route and at least one backup route are selected from the candidate routes. Restricted areas or priority channels are set using a checkmark constraint. Diversion weight parameters are set using a sliding mechanism to limit the carrying capacity of each route. Finally, a final evacuation route instruction is generated and sent to the evacuation control platform. By visualizing the Pareto candidate routes with both evacuation time and risk exposure indicators and simultaneously presenting floor paths, key nodes, bottlenecks, and capacity margins, efficiency and safety can be quickly weighed and primary / backup routes determined when the fire situation is uncertain. Furthermore, the restricted / priority constraints and diversion weights finely control crowd guidance and carrying capacity, reducing bottleneck congestion and the risk of backflow and stampedes, and improving the executability of evacuation instructions and the reliability of emergency response.

[0013] As a further aspect of this invention, in the management terminal, the evacuation time index is obtained by the evacuation control platform by accumulating the passageway sequence of each candidate route in the Pareto candidate evacuation route set segment by segment. The passageway time is determined by the length of the passageway and the passage speed. The passage speed is taken from the output value of the passage speed estimation module. When the expected flow of people at the passageway exceeds the maximum throughput capacity of the current passageway determined by the effective net width and the unit width throughput capacity, the over-limit ratio is mapped to a queuing delay, which is added to the travel time obtained by dividing the length by the passage speed to form the segment passage time of the current passageway. The accumulated evacuation time index of the candidate route is obtained and updated on a rolling basis according to a preset update cycle. By calculating the evacuation time of the candidate route segment by the passageway length and the real-time passage speed, and superimposing the queuing delay mapped by the over-limit ratio when the capacity exceeds the limit, the time penalty for congestion bottlenecks and dynamic throughput capacity is quantified, enabling the management terminal to obtain an evacuation time index that is closer to the actual evacuation time at the fire scene and can be updated on a rolling basis according to the situation.

[0014] As a further aspect of this invention, in the management terminal, the risk exposure index is obtained by cumulatively summing the values ​​of each candidate route in the Pareto candidate evacuation route set along the sequence of passage edges it passes through. Specifically, the concentration of toxic gases, temperature, visibility, and travel speed at each passage edge are normalized to dimensionless risk quantities according to preset upper and lower thresholds, and then weighted according to preset weights to obtain the risk intensity of the current passage edge. The risk intensity is then multiplied by the travel time of each segment of the current passage edge to obtain the exposure amount of the current passage edge. Finally, the exposure amounts of all passage edges of the candidate routes are summed to obtain the risk exposure index of the candidate route, which is updated on a rolling basis with a preset update cycle. By uniformly normalizing and weighting toxic gases, temperature, visibility, and travel speed to form the passage risk intensity, and then multiplying it by the travel time of each segment to obtain the exposure amount and accumulating it along the route, a quantitative assessment of the increased risk of staying in high-risk areas for longer periods is achieved. This allows for the rolling acquisition of comparable and rankable route safety indicators with each update cycle, thereby prioritizing the selection of low-exposure evacuation routes and reducing the risk of asphyxiation, poisoning, and heat injury.

[0015] The technical advantages of the fire safety evacuation control system for large commercial complexes according to the present invention are as follows:

[0016] This invention achieves multi-source fusion perception of toxic gases, temperature, visibility, and video pedestrian flow. It dynamically constructs a dual-objective function of passage time and risk exposure on a multi-story evacuation topology, superimposed with channel capacity constraints. An improved dung beetle optimization algorithm incorporating Kent chaos initialization and dynamic back-learning outputs a set of passable Pareto primary and backup evacuation routes. This allows evacuation decisions to be adaptively updated in real-time according to smoke diffusion, sudden drops in visibility, and congestion bottlenecks, avoiding the misleading effect of single-source shortest paths in vertically connected and bottleneck stairwell scenarios. Simultaneously, based on human target detection, multi-target tracking, occupancy rate and density map correction, and density-velocity constraint fusion, a passage speed estimation mechanism is implemented in smoke... Even under obstructed and low-light conditions, it can stably output crowd density and passage speed, reducing erroneous diversions caused by missed detections and misjudgments. Furthermore, the intelligent evacuation indicator lights dynamically switch between primary, backup, and prohibited directions, and set hysteresis thresholds to suppress backflow. The intelligent lighting provides directional-illuminance consistent guidance based on primary routes, backup routes, and prohibited areas, reducing the risks of backflow, stagnation, and stampedes. The management terminal presents evacuation time and risk exposure indicators in a visual manner and supports the configuration of prohibited, priority constraints, and diversion weights, improving the interpretability and enforceability of command decisions. This enhances the efficiency and safety of personnel evacuation in large commercial complex fires and reduces the probability of asphyxiation, poisoning, and heat injury. Attached Figure Description

[0017] Figure 1 This is a system diagram of the present invention;

[0018] Figure 2 This is an industrial design interface diagram of the intelligent evacuation indicator sign of the present invention;

[0019] Figure 3 This is a schematic diagram of the fire analysis of the invention as shown in the embodiment of the present invention. Detailed Implementation

[0020] 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.

[0021] Example

[0022] like Figure 1As shown, this invention proposes a fire safety evacuation control system for large commercial complexes, comprising intelligent evacuation indicator lights, intelligent lighting, environmental information acquisition devices, an evacuation control platform, and a management terminal. The intelligent evacuation indicator lights, intelligent lighting, and environmental information acquisition devices are connected to the evacuation control platform, which is connected to the management terminal. The environmental information acquisition devices include toxic gas sensors, temperature sensors, visibility sensors, and cameras. The toxic gas sensors include, but are not limited to, carbon monoxide sensors, hydrogen cyanide sensors, hydrogen chloride sensors, nitrogen oxide sensors, and sulfur dioxide sensors. The intelligent evacuation indicator lights include a control module and a display module. The control module is connected to both the evacuation control platform and the display module. The evacuation control platform includes a brightness control module, an evacuation route feature extraction module, a passage speed estimation module, and an evacuation path solving module. The evacuation path solving module is connected to the management terminal.

[0023] like Figure 2 As shown, a feasible implementation method of the intelligent evacuation indicator signs in the system proposed in this invention is realized in the form of an industrial large screen. As illustrated, the layout of the intelligent evacuation indicator signs in this invention follows the hierarchical deployment principle of evacuation topology nodes-channel segments-key functional points: First, main route indicator signs (green main route blocks on the interface) are continuously deployed along the main evacuation channel selected by the managed terminal, forming a continuous directional information chain at the personnel's line of sight, enabling personnel to maintain directional consistency even in smoke and low visibility conditions; second, node indicator signs with dynamically switchable arrows (gray direction blocks on the interface) are deployed at decision nodes such as forks, corners, and connecting corridors. The evacuation control platform switches the direction of indication in real time according to changes in crowd density and passage speed. Traffic flow is weighted and allocated to primary and backup routes to avoid congestion and backflow. Simultaneously, backup route signs are grouped and placed along backup routes (yellow backup route blocks on the interface) to ensure stable and identifiable guidance information and rapid switching with the primary route. For entrances to routes deemed high-risk or under lockdown, no-entry signs (red no-entry zone X on the interface) are placed at upstream key nodes and entrances, using no-entry symbols and flashing warnings to prevent accidental entry and guide detours. Staircase directional signs (“Evacuation Staircase” area on the interface) are placed in the entrance lobbies, stairwells, and beside safety exit doors of evacuation stairwells to clearly indicate the vertical evacuation direction. Assembly point signs (blue flags on the interface) are placed at safety assembly points to achieve closed-loop guidance from the fire area to the stairwell / exit and then to the assembly point. These signs are grouped and linked for addressing and control according to primary route groups, backup route groups, and no-entry groups, ensuring continuous guidance and rapid traffic diversion and lockdown in case of changing circumstances.

[0024] Taking a commercial complex in a certain city as an example, the complex has five floors above ground and two floors below ground, with a total construction area of ​​120,000 square meters. The B1 to 5F floors are connected by a central atrium, forming a circular corridor on each floor. There are two enclosed evacuation stairwells, S1 (east side) and S2 (west side). There are outdoor assembly points in the east and west plazas on the 1F floor. The 4F floor houses a cinema and a catering area, resulting in a high density of people gathering. The corridor has a net width of 2.5 to 3.2 meters, and the stairwells have net widths of 2.2 meters for S1 and 3.0 meters for S2. There are two fire doors, F1 and F2, on the 2F corridor for zoning. Intelligent evacuation indicator lights and intelligent lighting are installed at key nodes on the 2F floor. Environmental information collection devices include toxic gas sensors (CO, HCN, HCl, NOx, SO2), temperature sensors, visibility sensors, and cameras. The update cycle of the evacuation control platform is 2 seconds, and the refresh cycle of the intelligent evacuation indicator lights is 1 second.

[0025] On a certain day, the fire started at 19:22:10 in shop B217 on the north side of the 2nd floor, near the north corridor of the atrium. An electrical fire occurred and a large amount of smoke was produced. The corridor outside the shop was one of the main evacuation routes. After the fire, the flow of people on the 2nd floor quickly converged and reversed.

[0026] (1) First, the fire safety evacuation control system for large commercial complexes proposed in this invention collects real-time data at 19:22:20 through an environmental information collection device:

[0027] The 2F connecting corridor is divided into several passageway sections (the waiting area is bound to the sensor point one by one), and four key sections are selected:

[0028] e A North-facing corridor outside Gate B-217 (near the atrium)

[0029] CO = 480ppm, HCN = 5.2ppm, HCl = 1.5ppm, NOx = 2.0ppm, SO2 = 0.8ppm, temperature T = 62℃, visibility Vis = 12m, video acquisition occlusion rate 0.58, tracking loss rate 0.41 (smoke + congestion);

[0030] e B The east-facing corridor leads to the S1 anteroom.

[0031] CO = 180ppm, HCN = 1.2ppm, HCl = 0.4ppm, NOx = 0.8ppm, SO2 = 0.3ppm, T = 38℃, Vis = 25m, video acquisition occlusion rate 0.32, loss rate 0.18;

[0032] e C The west-facing corridor leads to the S2 vestibule.

[0033] CO = 60ppm, HCN = 0.3ppm, HCl = 0.1ppm, NOx = 0.2ppm, SO2 = 0.1ppm, T = 30℃, Vis = 40m, video acquisition occlusion rate 0.22, loss rate 0.12;

[0034] e D S1 anteroom entrance (bottleneck point)

[0035] CO = 210ppm, HCN = 1.5ppm, T = 40℃, Vis = 22m, and video footage shows a clear crowd gathering.

[0036] Meanwhile, the platform reads the status of the fire doors: F1 (north side partition door) = closed, F2 (east side partition door) = open, which means that smoke is more likely to spread eastward along the atrium side, and the risk in the S1 direction is on the rise.

[0037] (2) Perform video pedestrian flow feature extraction and density and speed estimation.

[0038] 1) with e B Taking the test area as an example, e is obtained by calibration respectively. A e B e C The actual areas are A A A B A C The specific values ​​are 45 square meters, 54 square meters, and 60 square meters, respectively.

[0039] 2) Extract the number, occupancy rate, and flow characteristics of human targets at 19:22:20.

[0040] e B Human detection + multi-target tracking yields the number of valid IDs N. B =52, instance mask occupies area S mask =12.4m 2 Occupancy rate O B =12.4 / 54=0.23; The main direction of the displacement vector points to "East→S1", and the return flow indicator =0;

[0041] e C : Number of valid IDs N C =41, occupancy rate 0 C =0.18; Main direction points to "West → S2", return flow indicator =0;

[0042] e A Since the occlusion rate of 0.58 and the loss rate of 0.41 exceed the threshold (such as 0.55 and 0.35), the "density map counting + occupancy rate fusion correction" mode is triggered. The detection ID is for reference only.

[0043] 3) Calculate population density

[0044] e B Basic density: ρ B =N B / A B =52 / 54=0.96 people / m 2 ;

[0045] e C Basic density: ρ C =41 / 60=0.68 people / m 2 ;

[0046] e A Fusion density: The equivalent number of people is obtained by integrating the density map. Occupancy rate measured O A =0.38, low image quality results in a confidence weight α = 0.35, then The number of people that can be tracked is determined by analyzing the time-frequency data from the cameras. When ρ is 40, A = 1.44 people / m 2 ;

[0047] 4) Estimate the passage speed

[0048] e B : Median of the set of individual velocities v B =0.92m / s;

[0049] e C : median v C =1.08m / s;

[0050] e A (Low quality): Tracking speed v trk =0.55m / s, from the density-velocity constraint model (example: v) model =1.30exp(-0.70ρ) gives v model ≈0.47m / s, image quality confidence γ=0.40, fused to obtain v A =γv trk +(1-γ) vmodel =0.40×0.55+0.60×0.47≈0.50m / s;

[0051] (3) Construct a dual objective function of channel risk cost and travel time.

[0052] 1) Normalization and weighting of toxic gas risk levels:

[0053] Set upper and lower threshold values ​​(which can be given by the platform configuration table):

[0054] CO: 0-1000ppm, HCN: 0-20ppm, HCl: 0-10ppm, NOx: 0-10ppm, SO 2: 0-10ppm; temperature 20–120℃; visibility 5–50m; speed 0.2–1.5m / s.

[0055] With e B Normalization for example: The overall normalized value of toxic gases The normalized value of temperature is Normalized value of visibility risk normalized value of velocity Risk intensity

[0056] Similarly, R can be calculated. C =0.11; R A =0.71;

[0057] 2) Segment-by-segment passage time and queuing delay

[0058] Travel time τ e =L e / v e +Δτ e ,e B and e C The section lengths are 40m and 45m respectively, and the stair section lengths S1 and S2 are 60m and 65m respectively. The calculated travel times are 40 / 0.92 = 43.5s and 45 / 1.08 = 41.7s respectively. Taking the unit width passage capacity q... max =1.3 people / s / m, then the capacity C of S1 S1 =2.2 * 1.3 = 2.86 people / s, the capacity C of S2 S2 =3.0 * 1.3 = 3.90 people / s. The evacuation control platform estimates the expected traffic based on the upstream population arrival rate and diversion weight. If no diversion is implemented, the estimated video feed into S1 will be F. S1 The rate is approximately 46 people / s (exceeding the limit), while in S2 it is only 1.8 people / s, representing an over-limit ratio. The evacuation control platform maps the over-limit ratio to queuing delays. Specific feasible methods include:

[0059] The queuing delay Δτ = 120s*(F / C-1), therefore Δτ S1 =73s, therefore the travel time for stair section S1 is 60 / 0.6 + 73 = 173s (the stair speed decreases due to congestion, for example, to 0.6m / s), while S2 satisfies the capacity Δτ S2 =0;

[0060] (4) Improved Dung Beetle Algorithm for Solving

[0061] 1) Construct a dual objective function

[0062] For any route P, the primary objective is the time objective: J time (P)=Στ e The second objective is the risk exposure objective: J risk (P)=Σ(R e ·τ e );

[0063] 2) Set capacity constraints

[0064] For any evacuation route or staircase along the path: F e ≤C e or ρ e ≤ρ max (In this embodiment, ρ is set) max =2.5 people / m 2 );

[0065] 3) Output Pareto candidate paths

[0066] The evacuation route solution module generates initial candidate routes covering the east staircase, west staircase, and cross-zone detours through Kent chaos initialization, and constructs structurally complementary routes using dynamic back learning. Representative candidate routes are obtained through non-dominated sorting.

[0067] P1 (shortest route but congested and with moderate risk of toxic gas): B217 external corridor → eB → S1 → 1F East Plaza, where J time =44 + 173 = 217s, J risk =0.11*51 + 0.09*76 = 12.4;

[0068] P2 (lower exposure backup route): via fire door F1 isolation zone → refuge corridor → S2. The time increase is not significant, but the exposure risk is lower, so it is used as a backup route.

[0069] Specifically, within each update cycle, the evacuation route solution module maps the toxic gas, temperature, visibility, crowd density, and passage speed data obtained from sensor and video fusion to the edges of each passage in a multi-story directed evacuation topology map, forming a time-varying set of edge attributes. A dynamic bi-objective function (minimum passage time, minimum risk exposure) is established on this directed graph, and capacity / density constraints are superimposed, making the problem a time-updated bi-objective constrained shortest path / path planning model. Treating each path as an individual, a path encoding-decoding mechanism is used to transform the discrete route selection problem into a searchable solution space. Under this bi-objective constrained model, an improved dung beetle optimization algorithm incorporating Kent chaotic initialization and dynamic back-learning is called for swarm search, and a Pareto candidate path set feasible to the constraints is output using non-dominated sorting (with feasibility priority / penalty function to handle constraints when necessary). After selecting the primary and backup routes and diversion weights from the Pareto set, only the indication / lighting control for the current cycle is executed, and the graph is updated and the solution is re-solved in the next cycle, realizing online adaptive closed-loop optimization of smoke diffusion, passage congestion, and fire door status changes.

[0070] (5) Management terminal line selection and traffic distribution weight settings

[0071] The management terminal uses intelligent evacuation indicator lights to visually display the evacuation time and risk exposure value for routes P1, P2, and P3, and marks S1 as exceeding the bottleneck and S2 as having sufficient margin. The management terminal selects P2 as the primary route and P3 as the backup route, with diversion weights of 0.75 for P2 and 0.25 for P3. The restricted area is designated as follows: A (The north-facing corridor outside Gate B217, near the atrium) is marked as restricted, and the management terminal recalculates the flow rate F. S2 = (0.75 + 0.25) * 3.8 = 3.8 people / s is less than C S2 = 3.9 people / s, satisfying the condition that when F S1 Reduce to 0, eliminating the bottleneck backflow warning risk markers from the planned route for S1;

[0072] (6) Intelligent evacuation indicator signs and intelligent lighting lights work together to perform actions.

[0073] 1) 2F fork in the road sign receives primary / backup routes and traffic splitting weights:

[0074] The default display shows the main arrow pointing to S2;

[0075] At the bifurcation point related to P3, alternative directions are displayed in rotation with a splitting weight of 0.25;

[0076] If the S2 inlet density is detected to rise to the threshold (e.g., ρ > 2.0 and lasts for 10 s) or the velocity is detected to decrease (e.g., v < 0.5 m / s), then... sAnd it lasts for 10 seconds before triggering a switch to increase the P3 weight, and it remains for at least 20 seconds after the switch to suppress frequent back-switching that causes backflow;

[0077] 2) The brightness control module marks the smart lights as: main route lights, backup route lights, and no-entry zone lights, and sends PWM signals accordingly:

[0078] Main street lights: PWM = 90% (difference between adjacent lights ≤ 10%);

[0079] Backup streetlights: PWM = 50%;

[0080] No-entry zone lights: PWM = 10% + 1.5Hz flashing (duty cycle varies between 30% and 70%);

[0081] PWM updates use a 1-3s linear ramp-up to avoid sudden brightness changes that could cause glare and indicator jitter.

[0082] The lighting unit has a built-in dimmable constant current LED driver, a lighting controller (MCU + PWM / current setting circuit), and a communication transceiver. By setting the sent group and PWM duty cycle / constant current, it can achieve high brightness for the main route, medium brightness for the backup route, low brightness flashing in the restricted area, and 1-3s linear ramp-up update. It can also be configured with an illuminance sensor or calibration coefficient to achieve uniform brightness of adjacent lights. If necessary, the color of each route indicator light can be changed. Specifically, the main route light can be set to green, the backup route light to blue, and the restricted area light can be set to low brightness flashing warning and set to red, which can be achieved through the adjustable RGB module.

[0083] like Figure 3 The diagram shows a schematic representation of the system provided by this invention on a circular floor plan of an example commercial complex on the 2nd floor. Point B217 (marked in red) indicates the location of the fire source in a shop on the north side of the 2nd floor. The area filled with red diagonal lines represents the fire's impact / high-risk impact zone (a zone affected by a combination of smoke, toxic gases, increased temperature, and decreased visibility). AHigh-risk section (red box): A critical passageway located outside the fire ignition point and connected to the inner ring passage. This indicates that the risk cost of this passageway is significantly increased during the current update cycle (e.g., smoke backflow, increased toxic gas concentration, low visibility). Therefore, routes passing through this section will be subject to a higher risk exposure cost during path solving. N0 (Shop Exit): The entrance point for evacuating from the burning shop / adjacent area into the public passageway. N1 (Fork): A critical branching point after entering the inner ring passageway. The system will decide whether to take the main route or the backup route here. It is also one of the key locations for intelligent evacuation signs. Fire doors F1 / F2 (yellow dots): Indicates door control points that can be used to isolate smoke or switch to different fire compartments. In the embodiment, F1 is used to enter the refuge corridor / isolation area to form a safer detour route. Main route P2 (green arrow): After N0→N1, prioritize guiding people to evacuate towards the west staircase of S2. Green indicates the priority / dominant route, and the corresponding lighting indicators and intelligent evacuation signs will use brighter and more conspicuous directional arrows. Alternative Route P1 (blue arrow): When the main route becomes congested (density exceeds limits, speed decreases) or the risk increases, it guides some people to take the S1 East Staircase. Blue indicates a passable alternative route, with lower brightness than the main route but still identifiable. Alternative Route P3 (cyan dashed line): Indicates a detour route from F1 to the isolation zone / refuge corridor and then to S2. It is used in certain situations (e.g., the inner ring is partially blocked by smoke, requiring avoidance of the high-risk eA section) to provide a lower exposure alternative path. The S1 East Staircase and S2 West Staircase are now marked in the center of the stair tread symbols, indicating that the route's endpoint is the actual location entering the stairwell, facilitating the direct binding of stair capacity constraints, congestion assessments, and availability status (whether closed / smoke-infiltrated) to this stair node / edge. A restricted area is a zone that the system determines to be restricted when it meets the criteria for prohibition (excessive risk or cost, blocked passage, severely overloaded capacity, etc.). The system will then display a warning sign indicating that the person should stay away and will also use low-brightness lighting with flashing lights to indicate this.

[0084] If the existing technical solution guides the passenger to S1 based on the shortest distance, the capacity of S1 will exceed the limit, causing queuing delays and backflow at the stairwell, resulting in significant pushing and shoving. This invention, through a dynamic dual-objective approach combining the risks of toxic gas, temperature, visibility, and speed with travel time constraints and capacity limitations, prioritizes P2 and diverts the passenger to P3, allowing the 2F crowd to avoid the area above e. A In high-smoke areas and the S1 bottleneck, the estimated evacuation time along the main path was reduced from approximately 217 seconds to approximately 127 seconds in the example data. At the same time, the risk exposure index was reduced from approximately 72.8 to approximately 12.4, reducing the risk of asphyxiation and stampede and improving the feasibility and stability of evacuation instructions.

[0085] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0086] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A fire safety evacuation control system for large commercial complexes, characterized in that, The system includes intelligent evacuation indicator lights, intelligent lighting, environmental information collection devices, an evacuation control platform, and a management terminal. The environmental information collection devices include toxic gas sensors, temperature sensors, visibility sensors, and cameras. The evacuation control platform acquires images of each evacuation route from the cameras and extracts image features such as the number of human targets, occupancy rate, and flow direction. It calculates crowd density and estimates passage speed. Based on the multi-floor evacuation topology, the evacuation control platform constructs a dynamic bi-objective function within a preset update cycle, consisting of the channel risk cost and passage time generated by toxic gas, temperature, visibility, and passage speed. Under the constraint of channel capacity, it uses an improved dung beetle optimization algorithm that introduces Kent chaotic initialization and dynamic back learning to solve for the Pareto candidate evacuation route set, which is then sent to the management terminal. After the management terminal selects an evacuation route, the evacuation control platform controls the intelligent evacuation indicator lights according to the hysteresis switching threshold and controls the intelligent lighting to guide the diversion and evacuation of personnel.

2. The fire safety evacuation control system for large commercial complexes according to claim 1, characterized in that, Intelligent evacuation indicator lights are installed along the evacuation routes at intersections, corners, lobby exits, connecting corridors, entrance lobbies of evacuation stairwells, entrances to refuge areas, and on one side of safety exit doors, with their display faces facing the direction of the incoming flow of people; intelligent lighting is installed along the evacuation routes and stair sections on the side walls.

3. The fire safety evacuation control system for large commercial complexes according to claim 1, characterized in that, The intelligent evacuation indicator sign includes a control module and a display module. The display module shows the main evacuation direction arrow, alternative direction indicators, current location indicator, target assembly point indicator, and no-entry warning indicator. The control module receives the main route, alternative routes, and their diversion weight parameters from the evacuation management platform, and dynamically switches the display direction of the display module according to changes in crowd density and passage speed within a preset refresh cycle. It also sets a hysteresis threshold for switching to suppress backflow. The intelligent lighting and intelligent evacuation indicator sign adopt a linked dimming strategy. The evacuation management platform includes a brightness control module. The brightness control module outputs brightness control commands to the intelligent lighting in the main route, alternative routes, and no-entry areas according to the channel risk cost and crowd density. Based on the evacuation topology relationship between the final evacuation route and the target area, the intelligent lighting is marked as the main route light, alternative route light, and no-entry area light, and dimming control values ​​are issued to each.

4. A fire safety evacuation control system for large commercial complexes according to claim 3, characterized in that, In the brightness control module, the dimming control quantity is either the PWM duty cycle setting value or the constant current drive current setting value. Specifically, the duty cycle of the main street lights is set to 70% to 100%, and the brightness difference between adjacent lights is uniformized to ensure that the difference is no more than 10%. The PWM duty cycle of the backup street lights is set to 30% to 60%, and the PWM duty cycle of the restricted area lights is set to 0% to 20%, with a superimposed frequency of 1 to 2 Hz and a PWM duty cycle of 30% to 70% for flashing control or intermittent lighting control according to a preset cycle. The PWM duty cycle setting value is updated in a linear ramping manner within 1 to 3 seconds according to the changes in crowd density and channel risk cost.

5. A fire safety evacuation control system for large commercial complexes according to claim 1, characterized in that, The evacuation control platform also includes an evacuation channel feature extraction module. After acquiring video images from cameras in each evacuation channel, the evacuation channel feature extraction module performs distortion correction and adaptive defogging enhancement in a preset waiting area, and uses human target detection and multi-target tracking to perform inter-frame correlation on human targets to obtain stable IDs. Then, the statistical value of the effective IDs in the waiting area is used as the number of human targets. The occupancy rate is calculated based on the overlap area ratio between the instance mask and the waiting area. Based on the direction histogram of the human target displacement vector obtained by multi-target tracking, the main flow direction and return flow identifier of the channel are extracted to obtain the number of human targets, occupancy rate and flow direction features.

6. A fire safety evacuation control system for large commercial complexes according to claim 5, characterized in that, The evacuation channel feature extraction module is connected to a passage speed estimation module. The passage speed estimation module performs geometric calibration of the candidate area from pixels to metric coordinates and obtains the actual area of ​​the candidate area. It calculates the crowd density based on the number of human targets and the actual area. When the occlusion rate or tracking loss rate exceeds the threshold, it introduces occupancy rate and density map integral to fuse and correct the crowd density. Then, based on the inter-frame displacement of human targets in the metric coordinates and through sliding window filtering, it obtains the individual velocity set. The median of the individual velocity set is taken as the passage speed. When the image quality index is lower than the preset threshold, it calculates the passage speed obtained from the human target displacement and the model speed obtained from the crowd density input density-velocity constraint model, respectively. It uses the confidence level generated by the image quality index as the weight to weight and fuse the passage speed obtained from the human target displacement and the model speed obtained from the crowd density input density-velocity constraint model to obtain the final estimated value of the passage speed.

7. A fire safety evacuation control system for large commercial complexes according to claim 6, characterized in that, The traffic speed estimation module is connected to the evacuation path solving module. The evacuation path solving module constructs the multi-story evacuation space of the commercial complex into a directed evacuation topology graph composed of nodes and passage edges. Within a preset update period, it constructs a dynamic bi-objective function for any candidate evacuation route. The first objective is to minimize the evacuation travel time, and the second objective is to minimize the path risk exposure. The passage capacity constraint is that the expected pedestrian flow of any passage edge through which the candidate route passes within the update period is not greater than the maximum throughput capacity determined by the current effective net width and unit width throughput capacity, or the crowd density of the passage edge does not exceed the preset density limit. Under the passage capacity constraint, an improved dung beetle optimization algorithm is used to generate an initial candidate route population by introducing Kent chaotic mapping and to generate reverse candidate routes by dynamic back learning. Non-dominated sorting is performed to output a Pareto candidate evacuation route set.

8. A fire safety evacuation control system for large commercial complexes according to claim 1, characterized in that, The management terminal receives the Pareto candidate evacuation route set and visualizes it using the evacuation time index and risk exposure index of each candidate route as evaluation dimensions. It also displays the floor sequence, key node sequence, bottleneck channel identification, and channel capacity margin of the candidate routes. The terminal selects the main route and at least one backup route from the candidate routes, sets check-based constraints on restricted areas or priority channels, slides the diversion weight parameters to limit the carrying ratio of each route, and generates the final evacuation route instruction to be sent to the evacuation control platform.

9. A fire safety evacuation control system for large commercial complexes according to claim 8, characterized in that, In the management terminal, the evacuation time index is obtained by the evacuation control platform by accumulating the passageway sequence of each candidate route in the Pareto candidate evacuation route set segment by segment. The passageway time is determined by the length of the passageway and the passage speed. The passage speed is taken from the output value of the passage speed estimation module. When the expected flow of people on the passageway exceeds the maximum throughput capacity of the current passageway determined by the effective net width and the unit width throughput capacity, the excess ratio is mapped to queuing delay and added to the travel time obtained by dividing the length by the passage speed to form the passageway time segment by segment of the current passageway. The evacuation time index of the candidate route is accumulated and updated on a rolling basis with the preset update cycle.

10. A fire safety evacuation control system for large commercial complexes according to claim 9, characterized in that, In the management terminal, the risk exposure index is obtained by accumulating the sequence of passage edges traversed by each candidate route in the Pareto candidate evacuation route set. Specifically, the concentration of toxic gases, temperature, visibility, and travel speed at each passage edge are normalized into dimensionless risk quantities according to preset upper and lower thresholds, and then weighted according to preset weights to obtain the risk intensity of the current passage edge. The risk intensity is then multiplied by the travel time of each segment of the current passage edge to obtain the exposure amount of the current passage edge. Finally, the exposure amounts of all passage edges of the candidate route are summed to obtain the risk exposure index of the candidate route, which is updated on a rolling basis with a preset update cycle.

Images