Automatic identification system for smoke control area

Abstract

The invention relates to the technical field of smoke control area automatic identification, and discloses a smoke control area automatic identification system, which comprises a behavior data acquisition module, a behavior analysis module, a behavior risk judgment module, a response generation module, a co-processing unit and the like. The behavior data acquisition module acquires real-time sensing data streams and divides the real-time sensing data streams into far infrared thermal sensing and visible light picture data streams; the behavior analysis module generates abnormal feature maps respectively; the behavior risk judgment module fuses the feature map based on a preset strategy matrix and determines a risk level; the response generation module matches a response strategy; and the co-processing unit generates a processing instruction after cross validation. The system further comprises a network communication gateway, a security linkage module and the like, and the functions of data protocol adaptation, processing queue optimization, event archiving and the like can be achieved. According to the invention, through multi-source data fusion and intelligent analysis, the accuracy and response efficiency of smoke control monitoring are improved, and the method is suitable for automatic identification and management of smoking behaviors in public places.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic identification of tobacco control areas, and specifically to an automatic identification system for tobacco control areas. Background Art

[0002] With the improvement of public health awareness and the popularization of tobacco control regulations in public places, how to efficiently and accurately achieve automatic monitoring and management of tobacco control areas has become a technical problem to be solved urgently. The traditional manual inspection method has defects such as poor real-time performance, high labor cost, and limited supervision scope, and it is difficult to meet the tobacco control requirements in large-scale and complex scenarios.

[0003] In the prior art, monitoring systems based on a single sensor (such as relying only on cameras or thermal detectors) have obvious limitations. When collecting picture data only through visible light cameras, it is easily interfered by factors such as environmental light changes and obstacles, resulting in a decrease in the accuracy of smoke recognition; while simply relying on far-infrared thermal sensing devices to detect heat sources, it is impossible to accurately obtain the morphological characteristics of smoking behaviors (such as smoke diffusion trajectories, human motion patterns, etc.), and false alarms or missed alarms are likely to occur. In addition, most of the existing systems lack a multi-dimensional data fusion and analysis mechanism, it is difficult to construct a complete smoking behavior feature model, and they lack dynamic adaptability in aspects such as risk level determination, disposal instruction generation, and collaborative response process optimization, and cannot flexibly adjust monitoring strategies and disposal priorities according to the actual scenario.

[0004] At the same time, traditional systems have compatibility problems in data transmission and protocol adaptation, and it is difficult to achieve efficient docking with existing security IoT networks, resulting in poor data circulation and high system integration costs. At the level of event traceability and management, existing solutions generally lack a complete event disposal record chain generation mechanism, which is not conducive to subsequent supervision effect evaluation and system optimization and upgrading. Therefore, there is an urgent need for an automatic identification system for tobacco control areas that can integrate multi-source perception data and has dynamic analysis and intelligent response capabilities to solve the problems of low detection accuracy, response lag, poor system compatibility, and insufficient management efficiency existing in the prior art. Summary of the Invention

[0005] The purpose of the present invention is to provide an automatic identification system for tobacco control areas to solve the problems raised in the above background art.

[0006] To achieve the above purpose, the present invention provides the following technical solution: An automatic identification system for tobacco control areas, the system includes: A behavior data acquisition module, which is used to obtain the real-time perception data stream of the monitoring area and divide the perception data stream into a far-infrared thermal sensing data stream and a visible light picture data stream based on the smoking behavior characteristics; A behavior analysis module for dynamically tracking the temperature of the far-infrared thermal data stream to generate a first anomaly feature map, and performing morphological analysis on the visible light image data stream to generate a second anomaly feature map; A behavior risk determination module for spatio-temporally fusing the first anomaly feature map and the second anomaly feature map according to a preset recognition strategy matrix, associating them with corresponding risk levels, and using the risk levels as the trigger thresholds for the current scene; A response generation module for matching a target response strategy in a preset behavior library based on the risk level and using the target response strategy as the execution logic for the current scene; A collaborative disposal unit for cross-verifying the trigger threshold and the execution logic to generate a collaborative disposal instruction.

[0007] Preferably, performing morphological analysis on the visible light image data stream includes: Dividing the dynamic change sequence in the visible light image data stream into a stable image group and an abnormal deformation group, and calculating contour features of the abnormal deformation group based on a preset smoke analysis model to generate a morphological feature set; Performing regional segmentation processing on the brightness gradient data in the visible light image data stream, extracting the texture change features of each region and constructing an abnormal atlas; Performing multi-dimensional matching on the morphological feature set and the abnormal atlas to generate the second anomaly feature map.

[0008] Preferably, the smoking behavior characteristics include a core detection dimension and an auxiliary detection dimension; the core detection dimension includes a heat source distribution identifier, a smoke diffusion identifier, and an action trajectory identifier; the auxiliary detection dimension includes an environmental interference identifier and a light fluctuation identifier, and each identifier corresponds to an independent data processing channel.

[0009] Preferably, the system further includes a network communication gateway for realizing protocol adaptation between the behavior data acquisition module, the behavior analysis module, the behavior risk determination module, and the response generation module and the security Internet of Things network respectively; The behavior data acquisition module divides the perception data stream based on the smoking behavior characteristics as follows: Receiving a mixed signal frame from the security Internet of Things network through the network communication gateway, and matching the protocol header of the mixed signal frame according to the identifier in the core detection dimension to separate the core signal segment; Traversing the extended header of the mixed signal frame according to the identifier in the auxiliary detection dimension to extract the auxiliary signal segment; Synchronizing the core signal segment and the auxiliary signal segment by timestamp and writing them into the thermal data storage area and the visible light data buffer respectively.

[0010] Preferably, when the preset recognition strategy matrix adopts a static association model, the risk level is the threshold mapping result of the fusion weight of the first abnormal feature map and the second abnormal feature map; When the preset recognition strategy matrix adopts a dynamic association model, the risk level is a collaborative determination set obtained by calibrating the association parameters of the first abnormal feature map and the second abnormal feature map in real time through an adaptive algorithm.

[0011] Preferably, the system further includes a security linkage module connected to the collaborative disposal unit, and the security linkage module is connected to the disposal terminal database through the network communication gateway; The security linkage module is configured to screen an adapted disposal queue from the disposal terminal database according to the disposal requirements in the collaborative disposal instruction, and generate an execution priority matrix to optimize the disposal response process.

[0012] Preferably, the generation of the execution priority matrix includes: Loading a monitoring area topology model, and positioning the spatial coordinate nodes of each terminal in the adapted disposal queue in the topology model; Calculating the optimal linkage path from the current position of each terminal to the target abnormal area based on a response timeliness algorithm, and sorting the adapted disposal queue according to the response efficiency; Integrating the optimal linkage path and the priority sorting into the topology model to generate a visual execution matrix.

[0013] Preferably, when the collaborative disposal unit performs cross-verification on the trigger threshold and the execution logic, a double-layer verification mode of a data integrity verification mechanism and a logic consistency verification mechanism is adopted. The data integrity verification mechanism is used to verify the effectiveness of data collection, and the logic consistency verification mechanism is used to eliminate response conflicts of the disposal terminal.

[0014] Preferably, the system further includes an instruction distribution module connected to the collaborative disposal unit. The instruction distribution module is configured to convert the collaborative disposal instruction into a terminal control instruction set, and send the terminal control instruction set to a target disposal terminal through the network communication gateway to initiate a collaborative response.

[0015] Preferably, the system further includes an event archiving module connected to the collaborative disposal unit. The event archiving module is configured to store the real-time perception data stream, the first abnormal feature map, the second abnormal feature map, the risk level, and the collaborative disposal instruction, and generate a complete event disposal record chain according to the response cycle.

[0016] Compared with the prior art, the beneficial effects of the present invention are: Based on the core detection dimensions of smoking behavior (heat source distribution identification, smoke diffusion identification, action trajectory identification) and auxiliary detection dimensions (environmental interference identification, light fluctuation identification), the behavior data collection module divides the perception data stream into far-infrared thermal sense data stream and visible light image data stream, achieving all-round capture of smoking behavior characteristics. Through the protocol adaptation of the network communication gateway and the security Internet of Things network, it ensures the efficient parsing of the mixed signal frame and the accurate separation of the core signal segment and the auxiliary signal segment, providing a reliable data basis for subsequent analysis.

[0017] Through dynamic temperature tracking and morphological analysis, the behavior analysis module generates the first abnormal feature map and the second abnormal feature map respectively. The processing of the visible light image data stream includes steps such as dynamic change sequence division, smoke contour feature calculation, brightness gradient region segmentation, and texture change feature extraction. Combining with the multi-dimensional matching mechanism, it effectively improves the comprehensiveness and accuracy of abnormal feature extraction, reducing the misjudgment problem caused by a single data dimension.

[0018] The behavior risk determination module realizes the spatio-temporal fusion and risk level determination of the abnormal feature map through a preset recognition strategy matrix. The dual mechanisms of the static association model and the dynamic association model enable the system to quickly determine the risk level through threshold mapping in normal scenarios, and also to use adaptive algorithms to calibrate the association parameters in real time in complex environments, improving the flexibility and adaptability of risk determination.

[0019] The response generation module matches the risk level with the target response strategy in the preset behavior library to ensure the pertinence and effectiveness of the disposal measures. The collaborative disposal unit ensures the reliability of the trigger threshold and the execution logic through a double-verification mode of data integrity verification and logical consistency verification, eliminating the response conflicts of the disposal terminals. The security linkage module combines the monitoring area topology model and the response timeliness algorithm to generate an execution priority matrix and a visual execution matrix, realizing the optimized sorting of the disposal queue and the optimal linkage path planning, significantly improving the efficiency and coordination of the disposal response.

[0020] The instruction distribution module ensures the accurate transmission of the collaborative disposal instructions to the target terminals, while the event archiving module provides detailed data support for subsequent supervision and traceability and system optimization by storing the whole-process data and generating a complete event disposal record chain. Overall, through multi-technology integration and full-process intelligent design, the system constructs a closed-loop system from data collection, analysis, determination to response disposal and event management, effectively solving the problems of insufficient detection accuracy, response lag, and poor system compatibility in the existing technologies, and providing an efficient and reliable technical solution for tobacco control management in public places. Brief Description of the Drawings

[0021] Figure 1This is the working principle diagram of the automatic recognition system for tobacco control areas in the present invention; Figure 2 This is the working principle diagram of the morphological analysis of the visible light picture data stream; Figure 3 This is the flow chart of the disposal of the security linkage module; Figure 4 This is the working principle diagram of the cross - verification and instruction distribution of the collaborative disposal unit. Specific implementation mode

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0023] Please refer to Figures 1 - 4 , an automatic recognition system for tobacco control areas involved in the present invention, the system includes: a behavior data acquisition module, a behavior analysis module, a behavior risk determination module, a response generation module and a collaborative disposal unit. The specific implementation is as follows: The behavior data acquisition module continuously obtains the perception data stream of the monitoring area, and divides the data stream into a far - infrared thermal perception data stream and a visible light picture data stream based on the smoking behavior characteristics. Among them, the smoking behavior characteristics include a core detection dimension and an auxiliary detection dimension. The core detection dimension covers a heat source distribution identifier, a smoke diffusion identifier and an action trajectory identifier. The auxiliary detection dimension includes an environmental interference identifier and a light fluctuation identifier, and each identifier corresponds to an independent data processing channel.

[0024] The behavior analysis module processes the far - infrared thermal perception data stream and the visible light picture data stream respectively: performs dynamic temperature tracking on the far - infrared thermal perception data stream to generate a first abnormal feature map; performs morphological analysis on the visible light picture data stream to generate a second abnormal feature map.

[0025] The behavior risk determination module fuses the first abnormal feature map and the second abnormal feature map in space and time according to the preset recognition strategy matrix, associates them with the corresponding risk levels, and uses this risk level as the trigger threshold for the current scene.

[0026] The response generation module matches the target response strategy in the preset behavior library based on the risk level and uses it as the execution logic for the current scene.

[0027] The collaborative disposal unit cross - verifies the trigger threshold and the execution logic, generates a collaborative disposal instruction, and realizes the automatic recognition and response to smoking behavior.

[0028] The present invention will be further described below in conjunction with Embodiments 1 to 5: Embodiment 1:

[0029] This embodiment involves the dynamic temperature tracking of far-infrared thermal sense data stream and the morphological analysis process of visible light picture data stream. The system performs multi-step processing on the dynamic temperature tracking of far-infrared thermal sense data stream and visible light picture data stream through the behavior analysis module, specifically as follows: Use the Gaussian mixture model to perform background modeling on the far-infrared thermal sense data stream to separate the dynamic heat source area; calculate characteristic parameters such as temperature gradient and heat diffusion rate for each heat source area to generate a temperature feature vector; predict and correct the heat source trajectory through the Kalman filter algorithm to establish a dynamic temperature change model; compare the temperature feature vector with a preset abnormal threshold, and mark the area exceeding the threshold as an abnormal heat area; generate a first abnormal feature map based on the spatial distribution and temporal evolution law of the abnormal heat area, where different colors represent different levels of temperature abnormality.

[0030] Group the dynamic change sequences in the visible light picture data stream. The visible light picture data stream consists of continuous image frames, and each image frame contains real-time visual information within the monitoring area. The system analyzes the continuous image frames through the background modeling algorithm, classifying the pixel points in the picture into background pixels and foreground pixels. The background modeling algorithm can use common algorithms such as the Gaussian mixture model and the codebook model. By learning the historical frames, a background model is established, and the pixels in the current frame that are less different from the background model than the preset threshold are classified as background pixels to form a stable image group; the pixels with a difference greater than the preset threshold are classified as foreground pixels to form an abnormal deformation group. The stable image group mainly includes fixed objects within the monitoring area (such as walls, desks, chairs, fixed equipment, etc.) and targets that remain stationary for a long time, and their pixel values change little or remain basically unchanged in consecutive frames; the abnormal deformation group includes areas within the monitoring area where movement, deformation, or new targets occur, such as pixel areas in scenarios such as people walking, object movement, and smoke diffusion.

[0031] After completing the grouping of the dynamic change sequences, the system calculates the contour features of the abnormal deformation group based on a preset smoke analysis model. The preset smoke analysis model is a model obtained by training a large number of smoke images in smoking scenarios. This model contains typical contour feature parameters of smoke, such as the diffusion speed of smoke, the degree of blurring of the contour edge, and the irregularity of the shape. The system first extracts the contour of each foreground area in the abnormal deformation group, uses an edge detection algorithm (such as the Canny edge detection algorithm) to determine the edge contour of the foreground area, and then calculates the geometric feature parameters of the contour, including contour area, perimeter, circularity, aspect ratio, eccentricity, etc. Circularity is used to measure the degree to which the contour approaches a circle, and the calculation formula is , the outline of the smoke usually has a low circularity; the aspect ratio is the ratio of the length to the width of the circumscribed rectangle of the outline, which can reflect the diffusion direction of the smoke; the eccentricity is used to describe the elliptical degree of the outline and can reflect the irregularity of the smoke shape. In addition, the system also calculates the dynamic change characteristic parameters of the outline, such as the change rate of the outline area between adjacent frames, the moving trajectory of the outline centroid, the fluctuation frequency of the outline edge, etc. By comprehensively considering these geometric characteristic parameters and dynamic change characteristic parameters, a morphological feature set is generated, which comprehensively describes the outline shape and change law of each foreground area in the abnormal deformation group.

[0032] The system performs region segmentation processing on the brightness gradient data in the visible light image data stream. The brightness gradient data reflects the change of pixel brightness in the image and can be obtained by calculating the gradient amplitude and direction of the image. First, the system converts each image frame in the visible light image data stream into a grayscale image to simplify the calculation complexity. Then, the Sobel operator is used to calculate the gradient components of the grayscale image in the horizontal and vertical directions respectively, and then the gradient amplitude and gradient direction are obtained. According to the magnitude of the gradient amplitude, the image is divided into different regions: the region with a smaller gradient amplitude represents the region where the brightness change in the image is gentle, usually the background or the surface of a uniform object; the region with a larger gradient amplitude represents the region where there are edges or severe texture changes in the image, which may be the boundary of the target object or the junction of the smoke and the background.

[0033] After completing the region segmentation, the system extracts the texture change features for each region. The texture change features are used to describe the texture pattern of the local region in the image and its change over time. For each segmented region, the system calculates its texture feature parameters, including the gray-level co-occurrence matrix (GLCM) features and the local binary pattern (LBP) features. The GLCM features extract parameters such as energy, entropy, contrast, and correlation by statistically counting the frequency of pixel pairs with specific gray values and spatial relationships in the image. Energy reflects the uniformity of the image texture, entropy reflects the complexity of the image texture, contrast reflects the clarity of the image texture, and correlation reflects the directionality of the image texture; the LBP features generate a binary pattern by comparing the gray values of the central pixel with those of the surrounding neighborhood pixels, and then statistically count the histogram of the pattern to reflect the local texture structure of the image. In addition, the system also calculates the texture feature change parameters of each region in consecutive frames, such as the mean, variance, and change rate of the texture feature values, to construct an abnormal map. The abnormal map stores the texture feature parameters and change parameters of each region in matrix form. Each element in the matrix corresponds to a region in the image, and the element value is the quantization value of the texture abnormality degree of the region. This quantization value is obtained by comparing the texture feature parameters of the current frame with those of the historical frames. The larger the value, the more significant the texture change in the region and the more likely it is an abnormal scenario such as smoke diffusion.

[0034] Finally, the system performs multi-dimensional matching between the morphological feature set and the abnormal atlas to generate a second abnormal feature map. The multi-dimensional matching process includes spatial dimension matching, time dimension matching, and feature dimension matching. In the spatial dimension, the spatial position coordinates of each foreground region in the morphological feature set are matched with the position coordinates of the corresponding region in the abnormal atlas to ensure that both describe the same physical space region; in the time dimension, the timestamps of each feature parameter in the morphological feature set are aligned with the timestamps of the corresponding feature parameters in the abnormal atlas to ensure that the matched feature parameters belong to the same image frame at the same moment; in the feature dimension, an association relationship is established between the morphological feature parameters and the texture feature parameters. For example, the diffusion speed of the contour is associated with the texture change rate, and the irregularity of the contour is associated with the complexity of the texture. By calculating the similarity (such as Euclidean distance, cosine similarity, etc.) between the morphological feature parameters and the texture feature parameters, the comprehensive matching degree value of each foreground region is obtained. The higher the comprehensive matching degree value, the more likely it is that the region has both typical smoke morphological features and texture change features, and is more likely to be the smoke region generated by smoking. The system marks the abnormal regions in the image frame according to the comprehensive matching degree values of each foreground region, and represents the level of the matching degree with different colors or gray levels to generate a second abnormal feature map. The second abnormal feature map intuitively shows the abnormal regions where smoking behavior may exist in the monitoring area and their feature matching degrees, providing rich visual and data support for subsequent behavior risk determination.

[0035] During the entire processing process, the system realizes a comprehensive analysis of the smoke morphology and texture changes generated by smoking behavior through steps such as grouping the dynamic change sequence of the visible light picture data stream, calculating contour features, segmenting the brightness gradient region, extracting texture features, and multi-dimensional matching, laying a foundation for accurately identifying smoking behavior. This process does not rely on specific experimental effect data, and only realizes the analysis of the data stream and the generation of the feature map through the logical processing of algorithms and models, ensuring the feasibility and repeatability of the technical solution.

[0036] Example 2: In the process of involving the network communication architecture and data stream division mechanism of the system, the system sets up a network communication gateway. This gateway serves as a bridge connecting the internal modules of the system and the security Internet of Things network, realizing protocol adaptation between the behavior data acquisition module, the behavior analysis module, the behavior risk determination module, and the response generation module and the security Internet of Things network. The security Internet of Things network transmits data based on standard communication protocols (such as MQTT, CoAP, etc.). The network communication gateway supports the parsing and conversion between different protocols by integrating a multi-protocol conversion engine, ensuring that the internal modules of the system can interact with various devices (such as cameras, sensors, disposal terminals, etc.) in the security Internet of Things network.

[0037] The behavior data acquisition module receives the mixed signal frames transmitted in the security Internet of Things network through the network communication gateway. The mixed signal frames are data units collected and encapsulated by the front-end devices (such as surveillance cameras, thermal sensors, etc.) in the security Internet of Things network, and contain core detection information and auxiliary detection information. The core detection information corresponds to the core detection dimensions in the smoking behavior characteristics, including the heat source distribution identifier, the smoke diffusion identifier, and the action trajectory identifier; the auxiliary detection information corresponds to the auxiliary detection dimensions, including the environmental interference identifier and the light fluctuation identifier. The mixed signal frames follow a specific frame structure, usually including a protocol header, a data payload, and a frame tail. The protocol header contains metadata such as the frame type, length, and identifier, the data payload contains the specific detection data, and the frame tail is used to verify the data integrity.

[0038] When performing data stream partitioning on the mixed signal frames based on the smoking behavior characteristics, the system first performs the separation operation of the core signal segment. According to the heat source distribution identifier, the smoke diffusion identifier, and the action trajectory identifier in the core detection dimensions, the behavior data acquisition module parses the protocol header of the mixed signal frame through the network communication gateway and identifies the preset feature codes in the protocol header. For example, the feature code corresponding to the heat source distribution identifier can be set to a specific binary sequence (such as 0x01), the smoke diffusion identifier corresponds to 0x02, and the action trajectory identifier corresponds to 0x03. When the protocol header contains these feature codes, the system determines that the mixed signal frame contains core detection information and extracts the corresponding core detection data segment in the data payload to form the core signal segment. The core signal segment mainly contains the key data directly related to the smoking behavior, such as the heat source position and temperature change data detected by the thermal sensor, the video stream of the personnel action trajectory captured by the camera, etc.

[0039] After completing the separation of the core signal segment, the system extracts the auxiliary signal segment. According to the environmental interference identifier and the light fluctuation identifier in the auxiliary detection dimensions, the behavior data acquisition module traverses the extended header of the mixed signal frame (the extended header is a supplementary part of the protocol header for storing auxiliary information) and parses the identifier information in the extended header fields. The environmental interference identifier may correspond to the identification codes of parameters such as the environmental noise intensity and the air flow velocity, and the light fluctuation identifier may correspond to the identification codes of parameters such as the light intensity and the color temperature. The system extracts the corresponding auxiliary detection data by matching the identification codes in the extended header to form the auxiliary signal segment. The auxiliary signal segment contains the environmental factor data affecting the smoking behavior detection, such as the real-time light intensity change data in the surveillance area, the air flow direction and velocity data of the ventilation system, etc. These data are used for subsequent calibration and interference filtering of the core detection data.

[0040] After the core signal segment and the auxiliary signal segment are separated, the system needs to perform time synchronization processing on the two types of signal segments to ensure the accuracy of subsequent analysis. Time synchronization is achieved through the timestamps in the signal segments. The timestamps use a unified time reference (such as UTC time) and are accurate to the millisecond level. The behavior data acquisition module first reads the timestamp information in the core signal segment and the auxiliary signal segment, and then performs frame alignment operations on the two types of signal segments according to the timestamps. Specifically, for the core signal segment and the auxiliary signal segment with the same timestamp, the system regards them as the detection data at the same moment and stores them associatively; for the signal segments with timestamp deviations, the system performs time calibration through methods such as linear interpolation or nearest neighbor interpolation to make the two types of signal segments consistent in the time series.

[0041] After the time synchronization is completed, the core signal segment and the auxiliary signal segment are written into different data storage areas respectively. Since the core signal segment contains key data related to thermal sensing detection, it is written into the thermal sensing data storage area. The thermal sensing data storage area uses cache technology (such as DDR memory) or dedicated storage media (such as SSD) to support fast read and write operations on real-time dynamic data, facilitating the behavior analysis module to timely retrieve the far-infrared thermal sensing data stream for dynamic temperature tracking. Since the auxiliary signal segment mainly involves visible light image data and environmental auxiliary information, it is written into the visible light data buffer. The visible light data buffer uses a queue structure to store data in chronological order, supports access by timestamp index, and is convenient for the behavior analysis module to perform morphological analysis on the visible light image data stream.

[0042] During the network communication process, the network communication gateway is also responsible for handling error checking and retransmission mechanisms in the data transmission process. For example, when it is detected that the frame tail check code (such as CRC check code) of the mixed signal frame is inconsistent with the actual calculation result, the gateway will send a retransmission request to the sending end device of the security Internet of Things network to ensure the integrity of the received data. In addition, the gateway supports data encryption transmission function, encrypts the core signal segment and the auxiliary signal segment through encryption protocols such as SSL / TLS, prevents data from being stolen or tampered with during transmission, and ensures the security and reliability of the system.

[0043] The network communication architecture and data flow division mechanism of the system achieve the efficient acquisition and preprocessing of multi-source heterogeneous data in the monitoring area through standardized protocol adaptation, refined signal segment separation, precise time synchronization, and reliable data storage. This process strictly follows the detection dimensions of smoking behavior characteristics, ensuring the independent processing and collaborative analysis of core detection data and auxiliary detection data, providing an accurate and orderly data foundation for the subsequent behavior analysis module to generate abnormal feature maps and the behavior risk determination module to determine risk levels. The entire process does not involve any assumptions or verifications of experimental effect data, and only realizes data transmission and processing through hardware architecture design and software algorithm logic, reflecting the engineering implementation path and logical rigor of the technical solution.

[0044] Embodiment 3: Regarding two application modes and a risk level generation mechanism involving a preset recognition strategy matrix. The behavior risk determination module, as the core decision-making unit of the system, performs fusion processing on the first abnormal feature map and the second abnormal feature map output by the behavior analysis module through the preset recognition strategy matrix. The core lies in adopting different feature fusion logics according to the matrix type (static association model or dynamic association model), and finally generating a risk level corresponding to the risk degree of smoking behavior, which is used as the threshold basis for triggering subsequent responses.

[0045] When the preset recognition strategy matrix adopts a static association model, the system, based on the prior knowledge and historical data of smoking behavior, predefines the fusion weight and threshold mapping rules for the first abnormal feature map (far-infrared thermal sensation feature) and the second abnormal feature map (visible light morphology feature). The construction process of the static association model is as follows: First, by analyzing a large amount of sample data of smoking scenarios, determine the contribution degrees of the far-infrared thermal sensation features (such as heat source temperature, heat zone diffusion rate) and visible light morphology features (such as smoke contour irregularity, diffusion speed) to the recognition of smoking behavior. For example, determine the thermal sensation feature weight as 60% and the morphology feature weight as 40% through expert experience or offline training. These weight values remain fixed after model deployment and do not dynamically adjust with real-time data.

[0046] In specific implementation, the processing flow of the static association model is as follows: First, perform spatio-temporal alignment on the first abnormal feature map and the second abnormal feature map to ensure that their spatial coordinates (such as pixel positions in the monitoring screen) and timestamps (such as data acquisition moments) are exactly corresponding. Spatio-temporal alignment is achieved through an image registration algorithm, such as a rigid transformation algorithm based on feature point matching, to unify the coordinate systems of the far-infrared thermal sensation image and the visible light image and eliminate spatial deviations caused by differences in sensor installation positions. Second, perform pixel-level fusion on the two aligned feature maps, that is, for each pixel point, calculate the fusion feature value according to the preset weight. Suppose the thermal sensation feature value of a certain pixel point is , and the morphology feature value is , the fusion feature value The calculation formula is as follows:

[0047] (The formula is only for logical illustration. In actual applications, the quantization method of characteristic values needs to be determined according to the output characteristics of the sensor).

[0048] After the calculation of the fusion characteristic value is completed, the system converts it into a risk level through a preset threshold mapping rule. The threshold mapping rule is set based on the risk level stratification standard. For example, the risk level is divided into three levels: low risk (threshold interval ), medium risk (threshold interval ), high risk (threshold interval ). When the fusion characteristic value of a certain area falls into the low-risk interval, it is determined that there may be environmental heat source interference (such as electrical appliance heating); when it falls into the medium-risk interval, it is determined that there is a suspected smoking behavior; when it falls into the high-risk interval, it is determined that there is a confirmed smoking behavior. The division of the risk level needs to be adjusted in combination with specific application scenarios. For example, in high-sensitivity areas (such as hospitals, schools), the high-risk threshold can be appropriately reduced to improve the detection sensitivity. When the preset recognition strategy matrix adopts a dynamic association model, the system calibrates the association parameters of the first abnormal feature map and the second abnormal feature map in real time through an adaptive algorithm to meet the detection requirements in complex environments. The core of the dynamic association model is to construct an algorithm framework that can automatically adjust feature weights and association rules according to real-time data. Its typical implementation methods include online learning algorithms based on machine learning (such as Adaptive Neuro-Fuzzy Inference System ANFIS, Recursive Least Squares RLS) or heuristic rule engines.

[0049] In the processing flow of the dynamic association model, it is first necessary to define dynamically adjustable association parameters, including but not limited to the thermal sensation feature weight , the morphological feature weight (satisfying ), the time decay factor (used to measure the influence degree of historical data on the current decision), the abnormal feature change rate threshold , etc. These parameters are dynamically updated through real-time input environmental data (such as light intensity, air flow velocity) and detection data (such as the entropy value, variance of the feature map). For example, when the environmental light intensity increases significantly (obtained through the light fluctuation identification of the auxiliary detection dimension), the reliability of the visible light morphological feature may decrease. At this time, the system automatically reduces and increases to enhance the proportion of the thermal sensation feature in risk determination.

[0050] The dynamic calibration process is achieved through the following steps: First, collect the environmental parameters at the current moment (such as light intensity , air flow velocity ), and feature map statistics (such as the average temperature of the thermal sensation feature map , the contour complexity of the morphological feature map ); Secondly, input these parameters into the adaptive algorithm module, and calculate the latest associated parameter values through preset mapping functions (such as linear regression models, neural networks); Finally, use the updated associated parameters to perform fusion processing on the feature maps. Taking adaptive weight adjustment as an example, assuming that the light intensity and the morphological feature weight satisfy a linear negative correlation relationship, that is:

[0051] (where is the initial weight, is the light influence coefficient), when the light intensity increases, automatically decreases according to this formula, thereby dynamically balancing the contribution degrees of the two types of features.

[0052] In the feature fusion stage, the dynamic association model generates risk levels in the form of a collaborative decision-making set. The collaborative decision-making set contains decision-making results in multiple dimensions. For example: (1) The independent decision-making result based on the thermal sensation feature , divided into "no abnormality", "suspected heat source", "confirmed heat source"; (2) The independent decision-making result based on the morphological feature , divided into "no abnormality", "suspected smoke", "confirmed smoke"; (3) The associated decision-making result based on the spatio-temporal joint feature , divided into "unrelated", "weak association", "strong association". By combining these three types of decision-making results (such as confirmed heat source confirmed smoke strong association), different risk level combinations are formed. For example, "high risk" corresponds to the situation where all dimensions are determined to be abnormal and strongly associated, "medium risk" corresponds to the situation where some dimensions are determined to be abnormal and weakly associated, and "low risk" corresponds to the situation of single-dimensional abnormality or no abnormality.

[0053] The advantage of the dynamic association model is that it can adapt to environmental changes in real time. For example, when the shape of the smoke diffuses irregularly due to the influence of air flow, by adjusting the time decay factor to enhance the weight of recent data and avoid misjudgment caused by interference from historical data; in the scenario of multi-heat source interference, by calibrating the associated parameters to distinguish the thermal sensation feature differences (such as the duration of the temperature peak, the moving trajectory of the hot area) between normal heat sources (such as coffee machines) and smoking heat sources (such as cigarettes). In addition, the dynamic association model supports continuous optimization of decision-making rules through online learning. For example, by accumulating real smoking event data, automatically adjusting the feature thresholds and association logics to improve the long-term detection accuracy of the system.

[0054] Whether it is a static association model or a dynamic association model, the core lies in achieving accurate identification of smoking behavior through a reasonable feature fusion strategy. The static model is suitable for scenarios with stable environments and fewer interference factors, and has the characteristics of high computational efficiency and simple deployment; the dynamic model is suitable for complex and changeable environments, and improves the detection robustness through an adaptive mechanism. The selection of the two models can be flexibly switched through system configuration parameters, or the optimal model can be automatically selected according to real-time environmental parameters, so as to achieve reliable risk level determination in different application scenarios. The entire process does not rely on any assumptions about experimental effect data, and only through model architecture design and algorithm logic, feature fusion and risk grading are realized, ensuring the feasibility and logical rigor of the technical solution.

[0055] Example 4: In the process of realizing the functions of the security linkage module and generating the execution priority matrix, the system sets up a security linkage module connected to the collaborative disposal unit, and this module establishes a communication connection with the disposal terminal database through the network communication gateway. The disposal terminal database stores the basic information of various disposal devices, including device types (such as sound and light alarms, sprinkler devices, ventilation system controllers, on-site patrol robots, etc.), physical location coordinates, communication protocols, response ability parameters (such as sprinkler coverage range, alarm volume threshold, robot moving speed), and current working status (such as online / offline, faulty / normal), etc. The core function of the security linkage module is to screen suitable disposal devices from the database according to the disposal requirements in the collaborative disposal instruction, and optimize the disposal response process by generating an execution priority matrix to achieve the efficiency and orderliness of multi-device collaborative operations.

[0056] After the collaborative disposal unit generates a collaborative disposal instruction, the security linkage module first parses the disposal requirements in the instruction. The disposal requirements at least include the spatial location of the target abnormal area (such as coordinates in the monitoring screen or physical area number), risk level (low / medium / high), and required disposal type (such as alarm, smoke reduction, personnel dissuasion, etc.). For example, when the risk level is high, the disposal requirements may include multiple tasks such as "activate the sound and light alarm", "turn on the sprinkler in the nearby area", and "dispatch the patrol robot to the scene" at the same time. According to these requirements, the security linkage module executes the screening logic: first filter out the devices with corresponding disposal capabilities (such as sprinkler devices that can cover the target area), and then exclude the devices with abnormal current status (such as offline or faulty devices), and finally form an adapted disposal queue.

[0057] The first step in generating the execution priority matrix is to load the monitoring area topology model. The topology model presents the spatial structure of the monitoring area in the form of a digital map, including information such as building floor plans, passage distributions, and equipment installation locations. It is stored in a vector graphics format, with each physical location corresponding to a unique coordinate node. For example, in a shopping mall monitoring scenario, the topology model can be refined to each store, corridor, and stairwell on each floor, and the locations of each sprinkler device and alarm are marked on the corresponding coordinate nodes. The security linkage module maps the physical locations of each terminal in the adaptation disposal queue to the nodes of the topology model through a coordinate matching algorithm, forming a visual device distribution map.

[0058] Next, the system calculates the optimal linkage path from each terminal to the target abnormal area based on the response timeliness algorithm. The response timeliness algorithm comprehensively considers factors such as path length, passage obstacles (such as closed doors, temporary obstacles), and real-time traffic conditions (such as personnel density). For fixed-position devices (such as sprinkler devices, alarms), the path length is directly the straight-line distance from the device to the target area or the shortest reachable path (such as the travel distance along the corridor); for mobile devices (such as patrol robots), an optimal route that avoids obstacles needs to be generated through path planning algorithms (such as the A* algorithm, Dijkstra algorithm). For example, if a patrol robot needs to pass through two corridors from its current location to the target area, the algorithm will analyze the personnel density in the corridors based on the real-time video stream and select the route with fewer people to shorten the arrival time.

[0059] While calculating the optimal path, the system calculates the response time of each terminal based on the path length and the device response speed (such as the robot's moving speed, sprinkler activation delay time). The formula for calculating the response time is: Response time = Device activation delay time + Path length / Device moving speed (for mobile devices) or Device activation delay time (for fixed devices). For example, if the activation delay of a certain sound and light alarm is 2 seconds and the signal transmission delay corresponding to the straight-line distance from the device to the target area is negligible, the response time is 2 seconds; if a certain patrol robot has an activation delay of 5 seconds, a moving speed of 1 m / s, and a path length of 30 meters, the response time is 5 + 30 = 35 seconds.

[0060] After completing the response time calculation, the system performs a priority sorting on the adaptation disposal queue according to the response efficiency. The core principle of the priority sorting is that "the shorter the response time, the higher the priority", that is, the device that can reach the target area fastest or execute the disposal action fastest is triggered first. For example, in a high-risk scenario, the sound and light alarms and sprinkler devices closest to the target area will be given the highest priority to ensure that the alarm is issued and the smoke spread is controlled immediately; the patrol robot has a lower priority due to its longer response time and is mainly used for subsequent on-site confirmation and personnel dissuasion.

[0061] After sorting is completed, the system integrates the optimal linkage path and the priority sorting result into the topological model to generate a visual execution matrix. The visual execution matrix is displayed in the form of a graphical interface and includes the following elements: (1) Highlighted marking of the target abnormal area (such as a red rectangular box); (2) The current positions and planned paths of each disposal terminal (such as blue lines representing the movement routes of robots and green lines representing signal transmission paths); (3) Terminal priority identification (such as numerical numbers or color grades, with red being the highest priority); (4) Estimated arrival time or countdown of the start time. This matrix is synchronized to the security monitoring center and each disposal terminal in real time through the network communication gateway, facilitating operators to intuitively grasp the scheduling of disposal resources and providing path navigation and execution instructions for terminal devices at the same time.

[0062] During the integration process, the system also needs to handle the conflict problems of multi-device collaborative operations. For example, when the coverage areas of multiple sprinkler devices overlap, it is necessary to avoid simultaneous activation, which may cause waste of water resources or excessive wetness in the area; when the path of a patrol robot overlaps with the evacuation route of personnel, it is necessary to adjust the path priority to ensure safety. Through the spatial analysis function of the topological model, the system pre-identifies potential conflict points and avoids them during priority sorting and path planning. For example, a time-sharing start strategy is set for the sprinkler devices in the overlapping area, or a detour route is re-planned for the robot.

[0063] The entire workflow of the security linkage module closely depends on the accuracy of the topological model of the monitoring area and the real-time nature of network communication. The topological model needs to be updated regularly to reflect environmental changes (such as new devices, channel renovations), which can be achieved through manual input or automatic scanning by sensors; the network communication gateway is responsible for ensuring the low-latency transmission of disposal instructions and status data, and uses TCP / IP protocol or a dedicated industrial bus protocol (such as PROFINET) to ensure reliability. Through this mechanism, the system realizes the automated process from disposal requirements to device scheduling, improves the response speed and collaborative efficiency of smoking behavior disposal, and no assumptions about experimental effect data are involved in the whole process. Only through topological modeling, path planning, and priority logic, the engineering implementation of the technical solution is realized.

[0064] Example 5: Regarding the cross-validation mechanism and the functions of the system expansion module involving collaborative disposal units. The collaborative disposal unit, as the core hub of the system decision-making and execution, undertakes the dual verification responsibilities for the trigger threshold and execution logic to ensure the accuracy and reliability of the system response. The data integrity verification mechanism and the logical consistency verification mechanism it adopts constitute a double-layer verification mode, which conducts quality control for the data source and execution logic respectively.

[0065] The data integrity verification mechanism is mainly used to verify the effectiveness of data collection. When the behavior data collection module obtains the real-time perception data stream, data loss, distortion, or timing disorder may occur due to sensor failures, network transmission interruptions, or environmental interference. The verification mechanism is implemented through the following steps: First, perform basic format verification on the far-infrared thermal sensor data stream and the visible light image data stream, check whether the data frames conform to the preset protocol (such as whether they contain complete timestamps, device identifiers, and data payloads), and eliminate invalid data with incorrect formats; Second, perform range rationality verification. For example, the temperature value of the far-infrared thermal sensor data needs to be within the normal operating range of the sensor (such as -40°C to 80°C), and the brightness value of the visible light image needs to be within the gray scale range of 0-255. Data outside the range is marked as abnormal and triggers re-collection or discard; Finally, implement timing continuity verification. Determine whether data collection is continuous by the time difference between adjacent frame timestamps. If the time interval exceeds the preset threshold (such as 500 milliseconds), it is determined that the data is interrupted, and the system automatically triggers the sensor to restart or switches to redundant sensors to collect data. Through these three layers of verification, ensure that the data stream received by the behavior analysis module is complete and reliable, and avoid deviations in the generation of abnormal feature maps due to data quality problems.

[0066] The logical consistency verification mechanism focuses on eliminating response conflicts in the disposal terminals. The execution logic output by the response generation module may contain multiple disposal instructions. When multiple terminals execute simultaneously, operation conflicts (such as starting the ventilation system and the sprinkler device at the same time, resulting in airflow interference with the sprinkler effect) or resource competition (such as multiple terminals competing for the same communication channel) may occur. The verification mechanism is implemented through the following process: First, establish a conflict rule library for disposal instructions, and pre-define the compatibility relationships of various instructions. For example, "acoustic and optical alarm" and "voice prompt" are compatible instructions and can be executed simultaneously; "sprinkler" and "start / stop of high-temperature equipment" are mutually exclusive instructions and need to be prohibited from being triggered simultaneously. The rule library is stored in the form of a state transition matrix, and the matrix elements indicate whether the instruction combination is allowed to be executed simultaneously; Second, traverse and check the instruction set in the execution logic, compare it with the conflict rule library one by one, and identify mutually exclusive instruction pairs; Finally, for conflicting instructions, automatically adjust the execution order or selection according to the risk level and disposal priority. For example, in a high-risk scenario, give priority to executing the sprinkler instruction to reduce smoke, postpone non-urgent ventilation operations, and start the ventilation system after the smoke is initially controlled. Through this mechanism, ensure that the disposal instructions are logically self-consistent, and the terminals do not interfere with each other during collaborative operations, improving the stability of the response process.

[0067] In addition to the two-factor authentication mode, the system also sets up an instruction distribution module and an event archiving module as extended functional components of the collaborative disposal unit. The instruction distribution module undertakes the dual responsibilities of protocol conversion and instruction transmission: First, it converts the collaborative disposal instructions generated by the collaborative disposal unit (such as "Start the sprinkler in Area A" and "Dispatch Robot No. B to Point C") into a terminal control instruction set recognizable by the target disposal terminal. Different terminals may adopt different communication protocols (such as Modbus and Zigbee). The instruction distribution module parses the abstract disposal instructions into specific device control codes by integrating a multi-protocol conversion engine. For example, it converts the "Start sprinkler" instruction into a register write operation (coil setting) in the Modbus protocol. Second, it sends the instruction set to the target disposal terminal through a network communication gateway, supporting unicast, multicast, or broadcast transmission modes, and selects the optimal transmission path according to the terminal type to ensure that the instruction reaches the destination in the shortest time. The confirmation response mechanism is adopted during the transmission process. If the terminal does not return a reception confirmation signal within the specified time, the system automatically retransmits the instruction until it is successful or marks the terminal as faulty.

[0068] The event archiving module is responsible for the full-process data recording and management. Its storage objects include key data such as real-time perception data streams, the first abnormal feature map, the second abnormal feature map, risk levels, collaborative disposal instructions, and terminal execution results. The archiving process follows the time series principle and generates an event disposal record chain according to the response cycle (such as the complete process from detecting an anomaly to the end of disposal). Each record chain contains the following elements: event trigger time, monitoring area identifier, original value of perception data, process parameters for generating the feature map, basis for risk level determination, content of disposal instructions, terminal response time, and execution status (success / failure / timeout). The data storage adopts a distributed database architecture (such as a MySQL cluster), supporting multi-dimensional retrieval by time, area, risk level, etc., which is convenient for subsequent audit traceability (such as verifying the reason for a false alarm) or system optimization (such as analyzing the disposal efficiency in different scenarios). In addition, the event archiving module supports data backup and recovery functions, regularly synchronizing key data to an off-site disaster recovery center to prevent data loss due to hardware failures.

[0069] In the collaborative work of the cross-validation and extension module, the system forms a complete closed-loop control process: after the behavior data acquisition module obtains the original data, it generates a disposal instruction through behavior analysis and risk determination; the collaborative disposal unit ensures the accuracy and feasibility of the instruction through double-layer verification; the instruction distribution module is responsible for the accurate transmission and execution of the instruction; the event archiving module records the whole-process data for traceability. For example, when the system detects that the temperature in a certain area rises abnormally and the visible light picture shows a smoky deformation, the behavior risk determination module generates a high-risk level based on the preset policy matrix, and the response generation module matches the execution logic of "activate alarm + sprinkle + robot go"; the collaborative disposal unit first verifies the integrity of the sensing data (such as confirming that the thermal sensor and the camera are working properly), then verifies the compatibility between instructions (such as no conflict between alarm and sprinkling), and then sends the instruction to the corresponding terminal through the instruction distribution module; the event archiving module synchronously records all the information from data acquisition to terminal response, forming a traceable disposal record. The whole process does not rely on any hypothetical description of experimental effects, and only realizes the implementation of the technical solution through the function definition of hardware modules and the logical connection of software processes, ensuring the operability and verifiability of each link.

[0070] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device.

[0071] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An automatic recognition system for a tobacco control area, characterized in that, Including: A behavior data acquisition module, configured to obtain real-time perception data streams of a monitored area, and divide the perception data streams into far-infrared thermal perception data streams and visible light picture data streams based on smoking behavior characteristics; A behavior analysis module, configured to perform dynamic temperature tracking on the far-infrared thermal perception data stream to generate a first abnormal feature map, and perform morphological analysis on the visible light picture data stream to generate a second abnormal feature map; A behavior risk determination module, configured to, according to a preset recognition strategy matrix, perform spatio-temporal fusion on the first abnormal feature map and the second abnormal feature map and then associate them with corresponding risk levels, and use the risk levels as trigger thresholds for the current scene; A response generation module, configured to match a target response strategy in a preset behavior library based on the risk level, and use the target response strategy as the execution logic for the current scene; A collaborative disposal unit, configured to perform cross-verification on the trigger threshold and the execution logic to generate a collaborative disposal instruction.

2. The automatic recognition system for tobacco control areas according to claim 1, wherein Performing morphological analysis on the visible light picture data stream includes: Dividing the dynamic change sequence in the visible light picture data stream into a stable image group and an abnormal deformation group, and calculating contour features of the abnormal deformation group based on a preset smoke analysis model to generate a morphological feature set; Performing regional segmentation processing on the brightness gradient data in the visible light picture data stream, extracting texture change features of each segment of the region and constructing an abnormal map; Performing multi-dimensional matching on the morphological feature set and the abnormal map to generate the second abnormal feature map.

3. The automatic recognition system for tobacco control areas according to claim 1, characterized in that The smoking behavior characteristics include a core detection dimension and an auxiliary detection dimension; the core detection dimension includes a heat source distribution identifier, a smoke diffusion identifier, and an action trajectory identifier; the auxiliary detection dimension includes an environmental interference identifier and a light fluctuation identifier, and each identifier corresponds to an independent data processing channel.

4. The automatic smoking control area recognition system according to claim 3, characterized in that, The system further includes a network communication gateway, and the network communication gateway is configured to implement protocol adaptation of the behavior data acquisition module, the behavior analysis module, the behavior risk determination module, and the response generation module to the security Internet of Things network respectively; The behavior data acquisition module dividing the perception data stream based on smoking behavior characteristics includes: Receiving a mixed signal frame from the security Internet of Things network through the network communication gateway, and matching the protocol header of the mixed signal frame according to the identifier in the core detection dimension to separate the core signal segment; Traversing the extended header of the mixed signal frame according to the identifier in the auxiliary detection dimension to extract the auxiliary signal segment; Performing frame synchronization on the core signal segment and the auxiliary signal segment according to the time stamp and then writing them into the thermal perception data storage area and the visible light data buffer respectively.

5. The automatic smoke control area recognition system according to claim 1, characterized in that When the preset recognition strategy matrix adopts a static association model, the risk level is the threshold mapping result of the fusion weight of the first abnormal feature map and the second abnormal feature map; When the preset recognition strategy matrix adopts a dynamic association model, the risk level is a collaborative determination set obtained by performing real-time calibration on the association parameters of the first abnormal feature map and the second abnormal feature map through an adaptive algorithm.

6. The automatic recognition system for tobacco control areas according to claim 4, wherein It further includes a security linkage module connected to the collaborative disposal unit, and the security linkage module is connected to the disposal terminal database through the network communication gateway; The security linkage module is configured to screen an adapted disposal queue from the disposal terminal database according to the disposal requirements in the collaborative disposal instruction, and generate an execution priority matrix to optimize the disposal response process.

7. The automatic recognition system for tobacco control areas according to claim 6, characterized in that, The generating of the execution priority matrix includes: Loading a monitoring area topology model, and locating the spatial coordinate nodes of each terminal in the adapted disposal queue in the topology model; Calculating the optimal linkage path from the current position of each terminal to the target abnormal area based on the response timeliness algorithm, and sorting the adapted disposal queue according to the response efficiency; Integrating the optimal linkage path and the priority sorting into the topology model to generate a visual execution matrix.

8. The automatic recognition system for tobacco control areas according to claim 1, wherein When the collaborative disposal unit performs cross-verification on the trigger threshold and the execution logic, a double-verification mode of a data integrity verification mechanism and a logic consistency verification mechanism is adopted. The data integrity verification mechanism is used to verify the effectiveness of data collection, and the logic consistency verification mechanism is used to eliminate response conflicts of the disposal terminals.

9. The automatic recognition system for tobacco control areas according to claim 4, characterized in that, It further includes an instruction distribution module connected to the collaborative disposal unit. The instruction distribution module is configured to convert the collaborative disposal instruction into a terminal control instruction set, and send the terminal control instruction set to the target disposal terminal through the network communication gateway to initiate a collaborative response.

10. The automatic recognition system for tobacco control areas according to claim 1, characterized in that, It further includes an event archiving module connected to the collaborative disposal unit. The event archiving module is configured to store the real-time perception data stream, the first abnormal feature map, the second abnormal feature map, the risk level, and the collaborative disposal instruction, and generate a complete event disposal record chain according to the response cycle.

Images