Civil engineering site fire risk intelligent assessment system and method

Abstract

This invention relates to the field of civil engineering safety monitoring technology, specifically disclosing an intelligent fire risk assessment system and method for civil engineering sites. The system includes an image module, a smoke module, and an AI module. The image module includes a first camera and a second camera for acquiring image information of the construction site from different angles. The smoke module includes multiple smoke sensors for sensing smoke signals in specific areas of the construction site. The AI ​​module includes a location recognition submodule, a behavior recognition submodule, and a smoke feature analysis submodule, used to perform multi-dimensional fusion processing of the acquired images and smoke signals, and to intelligently identify and assess the smoking behavior of construction workers based on a set hazard scoring algorithm. The method combines sensor signal triggering, image behavior analysis, and multi-source information matching to construct a three-dimensional, real-time smoking identification and early warning mechanism, which can effectively improve the intelligent management level of fire safety at construction sites.

Description

Technical Field

[0001] This invention relates to the field of on-site fire protection technology, and in particular to an intelligent assessment system and method for on-site fire risk in civil engineering projects. Background Technology

[0002] Safety management at current civil engineering construction sites consistently faces significant challenges, especially given the increasingly complex fire-causing factors during construction and the hidden high-risk hazards posed by workers' improper operating procedures. Among these, the unauthorized smoking by construction workers on site has been listed in numerous engineering safety management reports as one of the most frequent, easily occurring, and difficult-to-identify major hazards.

[0003] Due to the temporary nature of construction site environments, the openness of structures, and the complexity of personnel movement, traditional monitoring methods, primarily relying on manual inspections and conventional smoke detectors, often struggle to detect smoking behaviors hidden in corners or remote areas in a timely manner. More seriously, most smoking behaviors are highly concealed: construction workers often choose areas with sparse foot traffic, blind spots in surveillance, or behind construction barriers to smoke, and their smoking actions are subtle, producing small amounts of smoke, making it difficult for existing video surveillance systems to accurately capture smoking actions or smoke characteristics.

[0004] Meanwhile, while some monitoring products on the market based on camera recognition or smoke sensing have been applied to some extent in static environments such as offices and building interiors, identifying smoking behavior in dynamic, open environments such as construction sites still faces the following technical challenges:

[0005] Smoke detectors cannot distinguish between cigarette smoke and interference sources such as construction dust and welding fumes, which can easily lead to false alarms or missed alarms.

[0006] Video cameras struggle to accurately identify individual smoking actions at long distances, high angles, or under complex lighting conditions, especially when the worker's face is not directly facing the camera or is obstructed, the recognition rate drops sharply.

[0007] A single sensing source (camera only or smoke sensor only) lacks a cross-validation mechanism, making it impossible to achieve high-confidence judgments;

[0008] There is a lack of a three-dimensional monitoring and intelligent analysis mechanism that integrates multi-angle image information, local smoke signals, personnel behavior trajectories and spatial positioning information.

[0009] Besides smoking, temporary electrical boxes and cable trays on construction sites also pose fire hazards. In particular, electrical components, when overloaded, experiencing poor contact, or aging, are prone to early abnormalities such as short-term arcing, electric sparks, and localized high temperatures. These electrical anomalies are mostly transient and non-continuous signals, making them difficult to detect and respond to in a timely manner using traditional overload protection or current fluctuation monitoring methods.

[0010] Therefore, it is evident that current monitoring systems lack a comprehensive approach that integrates image recognition, smoke detection, multi-source localization, and intelligent assessment to address fire hazards induced by minor behaviors such as smoking at construction sites and the risk of initial electrical malfunctions leading to fires. Consequently, there is an urgent need to develop a specialized intelligent fire and safety risk assessment system for civil engineering construction sites. This system should utilize multi-angle high-definition cameras, intelligent smoke detection, and artificial intelligence decision-making mechanisms to accurately identify, provide real-time warnings, and automatically intervene in hidden fire sources such as illegal smoking and electrical malfunctions, thereby improving the inherent safety level of construction sites. Summary of the Invention

[0011] This invention aims to address the problem of timely identification and intervention of illegal smoking by construction workers at civil engineering construction sites, especially in scenarios such as high-altitude, concealed areas, or nighttime construction. Traditional camera monitoring or smoke sensors suffer from low accuracy, delayed response, and high false alarm rates due to limited field of view and insufficient recognition capabilities. This invention constructs a multi-source perception fusion system and a hazard scoring model to achieve three-dimensional monitoring, intelligent assessment, and real-time early warning of smoking behavior at construction sites, thereby effectively preventing fire hazards and improving on-site fire safety.

[0012] To address the aforementioned technical problems, the present invention provides, in one aspect, an intelligent assessment system for fire safety risks at civil engineering sites, comprising:

[0013] The image acquisition module includes a first camera positioned high above the construction area and a second camera positioned near the ground within the construction area, used to acquire image information of construction personnel from different spatial angles, wherein:

[0014] The first camera is used to capture images of the overall behavior and spatial location of the construction workers.

[0015] The second camera is used to capture detailed images of the head and hands of construction workers.

[0016] A smoke detection module includes multiple smoke sensors deployed at different locations on the construction site. These sensors collect smoke information regarding the presence and intensity of smoke at corresponding spatial locations and output the spatial location information of the triggered smoke sensors.

[0017] An AI-powered intelligent assessment module, communicatively connected to the image acquisition module and the smoke detection module, is used to intelligently assess fire risks related to smoking by construction workers. The AI-powered intelligent assessment module includes:

[0018] The image behavior recognition submodule is used to extract smoking-related behavioral features of construction workers based on image information collected by the first and second cameras.

[0019] The location fusion submodule is used to spatially match the spatial coordinate information of construction workers identified from their images with the trigger positions of smoke sensors in a unified spatial coordinate system, and generate location matching degree parameters.

[0020] The smoke feature analysis submodule is used to analyze the smoke intensity, duration, and trend of change collected by the smoke sensor and generate smoke feature parameters.

[0021] The scoring calculation submodule is used to calculate the fire risk score D of the construction personnel based on the location matching degree parameter and the smoke characteristic parameter, according to a preset risk assessment algorithm; wherein

[0022] The fire risk score D satisfies the following relationship:

[0023]

[0024] in:

[0025] V A This represents the similarity score of smoking-related actions extracted from the first camera.

[0026] V B This represents the score extracted from the second camera, which includes hand and mouth movement features.

[0027] S represents the smoke concentration level score collected by the smoke sensor;

[0028] L match This involves spatially matching the spatial coordinates of construction workers identified from their images with the trigger positions of smoke sensors to generate position matching parameters.

[0029] T a The score indicates the duration of the suspected smoking action;

[0030] ΔS represents the smoke release stability score;

[0031] M id This represents the weighting factor for individual identification markers;

[0032] α, β, γ, δ, ε, ζ, η are weight coefficients;

[0033] K is the normalization coefficient.

[0034] Optionally, the first camera in the image acquisition module is positioned at a height of more than 5 meters above the ground at the construction site, for capturing an overall image of the construction site from above, and the second camera is positioned at a height of 1.5 to 2 meters above the ground, for capturing local behavioral images of the construction workers' heads and hands from a level perspective.

[0035] Optionally, the smoke sensor in the smoke detection module has the sensitivity to detect smoke particles in the PM2.5 particle size range and supports composite recognition based on a combination of infrared light scattering technology and a gas-sensitive chip, so as to improve the specific detection capability of cigarette smoke.

[0036] Optionally, the AI ​​intelligent assessment module further includes an image identity recognition submodule, used to bind the identity of construction workers based on the electronic name tag worn by the construction workers in the camera image, and the scoring calculation submodule dynamically adjusts the scoring factor Mi according to the personnel identity binding information.

[0037] Optionally, the image behavior recognition submodule uses a temporal neural network model or a multi-frame action fusion recognition model to perform temporal feature analysis on the hand and mouth / nose movements of construction workers in continuous image frames, so as to improve the recognition accuracy of smoking actions.

[0038] Optionally, the weight coefficients of each scoring factor in the scoring calculation submodule can be dynamically configured according to the construction environment, personnel category and risk level strategy, and adaptively adjusted based on the optimized parameter set trained from historical data.

[0039] Optionally, the location fusion submodule, based on a unified three-dimensional coordinate system of the construction site, performs spatial matching between the coordinates of the target detection box in the camera image and the location coordinates of the smoke sensor, and generates a location matching score by calculating Euclidean distance. .

[0040] Optionally, the smoke feature analysis submodule extracts the time series change trend of the smoke intensity signal output by the smoke sensor within a set time period, and the parameter score in the fire risk score D is assigned a value based on the stability and persistence of the intensity change.

[0041] Optionally, the AI ​​intelligent assessment module compares the hazard score D with the set risk judgment threshold YD. When D≥YD, a risk event label is generated and the subsequent alarm processing procedure is triggered.

[0042] Another aspect of the present invention provides an intelligent identification method for illegal smoking by construction workers at civil engineering sites, applied to an identification system including an image acquisition module, a smoke detection module, and an AI intelligent evaluation module, wherein the method includes:

[0043] Step 1: A first camera positioned high above the construction site captures an overhead view of the construction area, generating first image information containing the outlines of the construction workers' actions;

[0044] Step 2: A second camera positioned at a low position on the construction site simultaneously captures a horizontal view of the construction workers, generating a second image information that includes hand and mouth movements;

[0045] Step 3: The smoke sensor detects whether there are smoke particles with abnormal particle size ranges in the air within the set area, and outputs the corresponding smoke intensity signal;

[0046] Step 4: Input the first image information and the second image information into the AI ​​intelligent evaluation module, and extract the behavioral characteristics and identity tags of the construction personnel through the image behavior recognition submodule and the image identity recognition submodule, respectively;

[0047] Step 5: Spatial matching is performed between the spatial coordinates of the person identified in the image and the trigger position of the smoke sensor, and the spatial matching degree Lm is calculated by the position fusion submodule;

[0048] Step 6: Input the smoke intensity change signal into the smoke feature analysis submodule and calculate the smoke feature score ΔS;

[0049] Step 7: Quantitatively assess the smoking behavior of construction workers based on the following hazard scoring formula:

[0050]

[0051] in:

[0052] V A This represents the similarity score of smoking-related actions extracted from the first camera.

[0053] V B This represents the score extracted from the second camera, which includes hand and mouth movement features.

[0054] S represents the smoke concentration level score collected by the smoke sensor;

[0055] L match This involves spatially matching the spatial coordinates of construction workers identified from their images with the trigger positions of smoke sensors to generate position matching parameters.

[0056] T a The score indicates the duration of the suspected smoking action;

[0057] ΔS represents the smoke release stability score;

[0058] M id This represents the weighting factor for individual identification markers;

[0059] α, β, γ, δ, ε, ζ, η are weight coefficients;

[0060] K is the normalization coefficient;

[0061] Step 8: Compare the score D with the preset risk judgment threshold YD. If D≥YD, it is determined that there is a violation of smoking regulations, and an alarm signal is sent to the name tag worn by the target construction worker, while triggering the alarm process of the central platform.

[0062] The beneficial effects of the technical solution of this invention are:

[0063] The risk intelligent assessment system of the present invention achieves comprehensive observation and three-dimensional cross-verification of suspicious smoking behavior of construction workers by setting up a high-position camera (first image acquisition device) and a low-position camera (second image acquisition device) for multi-angle linkage acquisition, combined with the synchronous detection of smoke sensors. This effectively solves the problem of missed detection or misjudgment caused by traditional single-point cameras or smoke sensors due to angle obstruction and monitoring blind spots.

[0064] The risk intelligent assessment system of the present invention adopts a comprehensive hazard scoring model, which integrates multiple feature factors such as smoking action similarity score, hand and mouth feature matching degree, smoke concentration level, sensor image coordinate matching degree, action duration, and smoke release stability to establish a mathematical scoring model for intelligent assessment, avoiding false alarms caused by a single factor and significantly improving the overall recognition accuracy.

[0065] The risk intelligent assessment system of the present invention can trigger an immediate sound and light reminder through an alarm module bound to the construction worker's badge or wearing device when the system's intelligent score is higher than a preset danger threshold. This effectively stops the continued development of illegal smoking behavior and avoids serious consequences such as fires caused by concealed smoking.

[0066] The risk intelligent assessment system of this invention supports the flexible deployment of cameras and smoke detectors in construction worker activity areas such as high-altitude steel frames, floor structures, and scaffold edges. It can be seamlessly integrated with existing safety monitoring systems and has good engineering practicality and scalability.

[0067] The risk intelligent assessment system of this invention, by introducing an AI intelligent recognition and quantitative scoring system, transforms the smoking supervision problem, which previously relied on manual inspections and inefficient management, into a closed-loop system that can perceive in real time, quantitatively assess, and automatically warn. This significantly improves the efficiency and technical level of on-site fire safety management and has significant value for promotion and application. Attached Figure Description

[0068] Figure 1 This is a structural diagram of the intelligent evaluation system in an embodiment of the present invention;

[0069] Figure 2 This is a flowchart illustrating the method steps of the intelligent evaluation system in an embodiment of the present invention. Detailed Implementation

[0070] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0071] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0072] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0073] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0074] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature being directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature being directly above or diagonally above the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0075] like Figure 1 As shown, the risk intelligent assessment system in this embodiment consists of the following modules:

[0076] Image Acquisition Module: This module consists of a first camera positioned high above the construction area and a second camera positioned near the ground. The first camera, located at least 5 meters above the ground, is responsible for capturing an overall image of the construction site from a top-down perspective, thereby obtaining images of the overall behavior and spatial position of the construction workers. The second camera, positioned between 1.5 and 2 meters above the ground, captures detailed images of the workers' heads and hands from a level viewpoint, obtaining detailed images of their head, hands, and other specific behaviors. Through the cooperation of the high and low-position cameras, image information of the construction workers is acquired from different spatial angles, providing comprehensive data support for subsequent behavior recognition.

[0077] Smoke Detection Module: This module comprises multiple smoke sensors deployed at various locations within the construction site. These sensors are sensitive to PM2.5 particles and support composite identification based on a combination of infrared light scattering technology and a gas-sensitive chip, thereby enhancing the specific detection capability of cigarette smoke. The smoke detection module collects information on the presence and intensity of smoke at corresponding spatial locations and outputs the spatial location information of the triggered smoke sensors, providing a basis for determining the source of the smoke.

[0078] The AI-powered intelligent assessment module communicates with the image acquisition and smoke detection modules to intelligently assess fire risks related to smoking by construction workers. The AI-powered intelligent assessment module also includes the following sub-modules:

[0079] Image Behavior Recognition Submodule: Based on image information acquired by the first and second cameras, a temporal neural network model or a multi-frame action fusion recognition model is used to perform temporal feature analysis on the hand and mouth / nose movements of construction workers in continuous image frames, thereby extracting smoking-related behavioral features of construction workers and improving the accuracy of smoking action recognition.

[0080] Location fusion submodule: Based on a unified 3D coordinate system of the construction site, it performs spatial matching of the target detection box coordinates in the camera image with the location coordinates of the smoke sensor, and generates a location matching score Lm by calculating Euclidean distance. Under the unified spatial coordinate system, it performs spatial matching of the spatial coordinate information of the construction personnel identified from the image with the trigger position of the smoke sensor to generate location matching parameters.

[0081] Smoke Feature Analysis Submodule: Analyzes the smoke intensity, duration, and trend of change collected by the smoke sensor. Specifically, it extracts the time series trend of the smoke intensity signal output by the smoke sensor within a set time period, assigns a value to the ΔS score based on the stability and persistence of the intensity change, and generates smoke feature parameters.

[0082] The scoring calculation submodule dynamically configures the weight coefficients (α, β, γ, δ, ε, ζ, η) of each scoring factor based on the construction environment, personnel category, and risk level strategy, and adaptively adjusts them based on an optimized parameter set trained from historical data. This submodule calculates the fire risk score D for construction personnel according to a preset risk assessment algorithm, based on location matching parameters and smoke characteristic parameters. The fire risk score D satisfies the following relationship:

[0083]

[0084] in:

[0085] V A This represents the similarity score of smoking-related actions extracted from the first camera.

[0086] VB represents the score extracted from the second camera, which includes hand and mouth movement features.

[0087] S represents the smoke concentration level score collected by the smoke sensor;

[0088] L match This involves spatially matching the spatial coordinates of construction workers identified from their images with the trigger positions of smoke sensors to generate position matching parameters.

[0089] Ta indicates the duration score of the suspected smoking action;

[0090] ΔS represents the smoke release stability score;

[0091] Mid represents the weighting factor for individual identification markers;

[0092] α, β, γ, δ, ε, ζ, η are weight coefficients;

[0093] K is the normalization coefficient;

[0094] Image identification submodule: Based on the electronic name tags worn by construction workers in the camera image, the submodule identifies the construction workers. The scoring calculation submodule dynamically adjusts the scoring factor Mi according to the personnel identification information. The AI ​​intelligent assessment module compares the hazard score D with the set risk judgment threshold YD. When D≥YD, a risk event label is generated and the subsequent alarm processing procedure is triggered.

[0095] like Figure 2 As shown, the specific assessment method steps of the risk intelligent assessment system in this embodiment are as follows:

[0096] A smart identification method for illegal smoking by construction workers at civil engineering sites is applied to the aforementioned identification system, which includes an image acquisition module, a smoke detection module, and an AI intelligent evaluation module. The specific steps are as follows:

[0097] Step 1: A first camera positioned high above the construction site captures an overhead view of the construction area, generating first image information containing the outlines of the construction workers' behavior, providing data for overall behavior and location analysis.

[0098] Step 2: A second camera positioned at a low position on the construction site simultaneously captures a horizontal view of the construction workers, generating a second image that includes hand and mouth movements, focusing on key local movements related to smoking.

[0099] Step 3: The smoke sensor detects whether there are smoke particles with abnormal particle size ranges in the air within the set area, and outputs the corresponding smoke intensity signal to determine whether smoke exists.

[0100] Step 4: Input the first and second image information into the AI ​​intelligent assessment module. The image behavior recognition submodule and image identity recognition submodule will extract the behavioral characteristics and identity tags of the construction personnel to prepare for subsequent risk assessment and targeted processing.

[0101] Step 5: Spatial matching is performed between the spatial coordinates of the person identified in the image and the trigger position of the smoke sensor. The spatial matching degree Lm is calculated by the position fusion submodule to help determine the correlation between smoke and person behavior.

[0102] Step 6: Input the smoke intensity change signal into the smoke feature analysis submodule, calculate the smoke feature score ΔS, and further analyze the smoke features.

[0103] Step 7: Quantitatively assess the smoking behavior of construction workers based on the following hazard scoring formula:

[0104]

[0105] The meanings of each parameter are consistent with the scoring formula in the system. This formula comprehensively considers multiple factors to obtain an accurate risk score.

[0106] Step 8: Compare the score D with the preset risk judgment threshold YD. If D≥YD, it is determined that there is a violation of smoking regulations. An alarm signal is sent to the name tag worn by the target construction worker, and the alarm process of the central platform is triggered at the same time to achieve timely warning and handling.

[0107] The following is a detailed description of the structure and method of the risk intelligent assessment system in this embodiment:

[0108] First, the system deployment method and structure of the risk intelligence assessment system include:

[0109] Image acquisition module: A first camera is installed at a height of 5 meters or more above the ground, such as on a tower crane at the construction site, to ensure that it can capture a large-scale overall image of the construction site from above, obtaining information on the overall behavior and spatial position of the construction workers. A second camera is installed near the area where the construction workers are active, at a height of 1.5 to 2 meters above the ground, to capture detailed images of the construction workers' heads and hands from a level perspective.

[0110] Smoke detection module: Multiple smoke sensors are strategically placed at key locations throughout the construction site, such as material storage areas, worker rest areas, and near electrical equipment. These smoke sensors possess high detection sensitivity for smoke particles within the PM2.5 particle size range, and through a combination of infrared light scattering technology and a gas-sensitive chip, they effectively enhance the specific detection capability for cigarette smoke.

[0111] AI Intelligent Assessment Module: The AI ​​intelligent assessment module is deployed on the monitoring center server at the construction site, ensuring stable communication with the image acquisition module and smoke detection module. Initial settings are performed on each submodule of the AI ​​intelligent assessment module. Based on the specific conditions of the construction site, such as environmental characteristics and personnel composition, initial weight coefficients (α, β, γ, δ, ε, ζ, η) are set for each scoring factor in the scoring calculation submodule, and the risk judgment threshold YD is set to an empirical value. Simultaneously, the image identification submodule is associated with the electronic name tags worn by construction personnel.

[0112] Next is the intelligent identification process of the risk intelligent assessment system of this invention:

[0113] Image Acquisition: The first camera continuously acquires overhead images of the construction area, generating first image information containing the outlines of construction workers' behavior, such as capturing their walking routes and gathering patterns. The second camera simultaneously acquires overhead images of the construction workers, generating second image information including hand and mouth movements, such as whether there are hand movements near the mouth.

[0114] Smoke detection: Smoke sensors detect the presence of smoke particles with abnormal particle sizes in the air within a designated area in real time and output the corresponding smoke intensity signal. For example, if a smoke sensor located near a material storage area detects a change in smoke intensity at a certain moment, it indicates that smoke may be present in that area.

[0115] After image acquisition and smoke detection are completed, the following intelligent evaluation steps are performed:

[0116] Feature Extraction: The first and second image information are input into the AI ​​intelligent evaluation module. The image behavior recognition submodule uses a temporal neural network model to perform temporal feature analysis on the hand and mouth / nose movements of construction workers in consecutive image frames, extracting smoking-related behavioral features, and obtaining a similarity score V for smoking-related actions extracted by the first camera. A The second camera extracts a score (VB) based on hand and mouth movement features. The image identification submodule binds the construction workers' identities to the electronic badges worn by them in the camera images and transmits the identity information to the scoring calculation submodule.

[0117] Location matching: The location fusion submodule, based on a unified 3D coordinate system of the construction site, spatially matches the spatial coordinates of personnel identified from the image with the trigger positions of the smoke sensors, and generates a location matching score Lm by calculating Euclidean distance. For example, if the location of the smoke detected by the smoke sensor is close to the location of a construction worker in the image, the location matching score is high.

[0118] Smoke Feature Analysis: The smoke feature analysis submodule extracts the time-series variation trend of the smoke intensity signal output by the smoke sensor within a set time period, and assigns a smoke release stability score ΔS based on the stability and persistence of the intensity change. If the smoke intensity continues to rise steadily, the ΔS score is relatively high.

[0119] Scoring Calculation: The scoring calculation submodule calculates the score based on the above parameters and according to the hazard scoring formula.

[0120]

[0121] Calculate the fire risk score D for construction workers, for example, if V A =0.8 (indicating a high degree of similarity to the act of smoking), V B =0.7, S=0.6 (smoke concentration level rating), L m =0.8 (high positional matching), T a =0.7 (high score for suspected smoking action duration), ΔS=0.7, M id =0.5 (assuming a weighted factor for the individual identification marker of the construction worker), with weight coefficients α=0.2, β=0.2, γ=0.2, δ=0.1, ε=0.1, ζ=0.1, η=0.1, and K=1 (normalization coefficient), then the calculated D=1 / 1×[0.2×0.8+0.2×0.7+0.2×0.6+0.1×0.8+0.1×0.7+0.1×0.7]×(1+0.1×0.5)≈0.68.

[0122] Finally, risk assessment and alarm are performed: the score D is compared with the preset risk assessment threshold YD (assuming YD=0.6). Since D=0.68≥YD, it is determined that there is a violation of smoking regulations. The system immediately sends an alarm signal to the name tag worn by the target construction worker, reminding the worker to stop the violation, and simultaneously triggers the alarm process on the central platform to notify the on-site safety management personnel to handle the situation promptly.

[0123] The other scoring parameters are as follows:

[0124] V A The similarity score of smoking-related actions extracted from the first camera can be obtained by the following formula:

[0125]

[0126] in:

[0127] M hand-mouth A matching score is given for whether the hand is close to the mouth and nose area (e.g., based on human key point recognition).

[0128] T hover Rate the time the hand stays near the mouth and nose area (based on video frame rate).

[0129] R cycle A repetitive rating for whether there are periodic repetitive actions (such as taking a few puffs of a cigarette);

[0130] α1, α2, and α3 are weighting coefficients;

[0131] K A This is a normalization factor that ensures the score falls within the 0-1 range;

[0132] V B The score representing the hand and mouth movement features extracted by the second camera can be obtained using the following formula:

[0133]

[0134] in:

[0135] F hand-cig The confidence level for whether the hand is holding a cigarette-like object (e.g., by shape or color identification);

[0136] F mouth-exhale To identify whether there is a similar smoking action on the face (based on the opening and closing of key facial areas and micro-expression analysis).

[0137] F smoke-trail To determine whether there is a continuously rising smoke trail above the face (identification of the movement path of gray-white objects in video frames).

[0138] β1, β2, and β3 are weighting coefficients;

[0139] K B This is the normalization factor.

[0140] S represents the smoke concentration level score collected by the smoke sensor, which can be obtained by the following formula:

[0141]

[0142] in:

[0143] C PM2.5 This refers to the concentration of fine particulate matter.

[0144] C CO This refers to the carbon monoxide concentration.

[0145] V variation This represents the rate of increase in smoke signal over a short period of time.

[0146] γ1, γ2, and γ3 are weighting coefficients;

[0147] K s This is the normalization factor.

[0148] In summary, this invention addresses the problem of accurately and efficiently identifying smoking behavior at construction sites by proposing an intelligent assessment system that integrates multi-angle image acquisition and smoke sensing. This system utilizes multiple high- and low-angle cameras to collaboratively identify worker behavior characteristics and combines this with smoke signals acquired from smoke sensors. An AI assessment model is then used to comprehensively determine and score the hazard of smoking behavior, ultimately enabling rapid identification and immediate warning of illegal smoking, significantly improving fire safety management capabilities and response efficiency at construction sites.

[0149] The above are merely preferred embodiments of the present invention and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should recognize that any equivalent substitutions and obvious changes made based on the description and illustrations of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart assessment system for fire safety risks at civil engineering sites, characterized in that, include: The image acquisition module includes a first camera positioned high above the construction area and a second camera positioned near the ground within the construction area, used to acquire image information of construction personnel from different spatial angles, wherein: The first camera is used to capture images of the overall behavior and spatial location of the construction workers. The second camera is used to capture detailed images of the head and hands of construction workers. A smoke detection module includes multiple smoke sensors deployed at different locations on the construction site. These sensors collect smoke information regarding the presence and intensity of smoke at corresponding spatial locations and output the spatial location information of the triggered smoke sensors. An AI-powered intelligent assessment module, communicatively connected to the image acquisition module and the smoke detection module, is used to intelligently assess fire risks related to smoking by construction workers. The AI-powered intelligent assessment module includes: The image behavior recognition submodule is used to extract smoking-related behavioral features of construction workers based on image information collected by the first and second cameras. The location fusion submodule is used to spatially match the spatial coordinate information of construction workers identified from their images with the trigger positions of smoke sensors in a unified spatial coordinate system, and generate location matching degree parameters. The smoke feature analysis submodule is used to analyze the smoke intensity, duration, and trend of change collected by the smoke sensor and generate smoke feature parameters. The scoring calculation submodule is used to calculate the fire risk score D of the construction personnel based on the location matching degree parameter and the smoke characteristic parameter, according to a preset risk assessment algorithm; wherein The fire risk score D satisfies the following relationship: ； in: V A represents a smoking-related action similarity score extracted by the first camera; V B This represents the score extracted from the second camera, which includes hand and mouth movement features. S represents the smoke concentration level score collected by the smoke sensor; L match This involves spatially matching the spatial coordinates of construction workers identified from their images with the trigger positions of smoke sensors to generate position matching parameters. T a The score indicates the duration of the suspected smoking action; ΔS represents the smoke release stability score; M id This represents the weighting factor for individual identification markers; α, β, γ, δ, ε, ζ, η are weight coefficients; K is the normalization coefficient.

2. The system according to claim 1, characterized in that, The first camera in the image acquisition module is positioned at a height of more than 5 meters above the ground at the construction site, and is used to capture an overall image of the construction site from above. The second camera is positioned at a height of 1.5 to 2 meters above the ground, and is used to capture images of the head and hands of construction workers from a level perspective.

3. The system according to claim 1, characterized in that, The smoke sensor in the smoke detection module has the sensitivity to detect smoke particles in the PM2.5 particle size range, and supports composite recognition based on infrared light scattering technology and gas-sensitive chip combination to improve the specific detection capability of cigarette smoke.

4. The system according to claim 1, characterized in that, The AI ​​intelligent assessment module further includes an image identity recognition submodule, which is used to bind the identity of construction workers based on the electronic name tags worn by the construction workers in the camera image. The scoring calculation submodule dynamically adjusts the scoring factor Mi according to the personnel identity binding information.

5. The system according to claim 1, characterized in that, The image behavior recognition submodule uses a temporal neural network model or a multi-frame action fusion recognition model to perform temporal feature analysis on the hand and mouth / nose movements of construction workers in continuous image frames, so as to improve the recognition accuracy of smoking actions.

6. The system according to claim 1, characterized in that, The weight coefficients of each scoring factor in the scoring calculation submodule can be dynamically configured according to the construction environment, personnel category and risk level strategy, and adaptively adjusted based on the optimized parameter set trained from historical data.

7. The system according to claim 1, characterized in that, The location fusion submodule, based on a unified three-dimensional coordinate system of the construction site, performs spatial matching between the coordinates of the target detection box in the camera image and the location coordinates of the smoke sensor, and generates a location matching score by calculating Euclidean distance. .

8. The system according to claim 1, characterized in that, The smoke feature analysis submodule extracts the time series change trend of the smoke intensity signal output by the smoke sensor within a set time period, and the parameter ΔS in the fire risk score D is assigned a value based on the stability and persistence of the intensity change.

9. The system according to claim 1, characterized in that, The AI ​​intelligent assessment module compares the hazard score D with the set risk judgment threshold YD. When D≥YD, a risk event label is generated and the subsequent alarm processing procedure is triggered.

10. A method for intelligently identifying illegal smoking behavior by construction workers at civil engineering sites, applied to an identification system including an image acquisition module, a smoke detection module, and an AI intelligent evaluation module, characterized in that... The method includes: Step 1: A first camera positioned high above the construction site captures an overhead view of the construction area, generating first image information containing the outlines of the construction workers' actions; Step 2: A second camera positioned at a low position on the construction site simultaneously captures a horizontal view of the construction workers, generating a second image information that includes hand and mouth movements; Step 3: The smoke sensor detects whether there are smoke particles with abnormal particle size ranges in the air within the set area, and outputs the corresponding smoke intensity signal; Step 4: Input the first image information and the second image information into the AI ​​intelligent evaluation module, and extract the behavioral characteristics and identity tags of the construction personnel through the image behavior recognition submodule and the image identity recognition submodule, respectively; Step 5: Spatial matching is performed between the spatial coordinates of the person identified in the image and the trigger position of the smoke sensor. The spatial matching degree is calculated by the location fusion submodule. ; Step 6: Input the smoke intensity change signal into the smoke feature analysis submodule and calculate the smoke feature score ΔS; Step 7: Quantitatively assess the smoking behavior of construction workers based on the following hazard scoring formula: ； in: V A This represents the similarity score of smoking-related actions extracted from the first camera. V B This represents the score extracted from the second camera, which includes hand and mouth movement features. S represents the smoke concentration level score collected by the smoke sensor; L match This involves spatially matching the spatial coordinates of construction workers identified from their images with the trigger positions of smoke sensors to generate position matching parameters. T a The score indicates the duration of the suspected smoking action; ΔS represents the smoke release stability score; M id This represents the weighting factor for individual identification markers; α, β, γ, δ, ε, ζ, η are weight coefficients; K is the normalization coefficient; Step 8: Compare the score D with the preset risk judgment threshold YD. If D≥YD, it is determined that there is a violation of smoking regulations, and an alarm signal is sent to the name tag worn by the target construction worker, while triggering the alarm process of the central platform.

Images