Smoke detection method based on multi-sensor fusion analysis

Abstract

The application provides a smoke detection method based on multi-sensor fusion analysis, which comprises the following steps: when a smoke sensor detects that the smoke concentration in a detection range reaches a first threshold value, acquiring a temperature sensor to detect the temperature rising speed in the detection range, and acquiring a human body sensor to detect the environmental thermal radiation uniformity in the detection range; generating corresponding correction coefficients to correct the smoke concentration and the first threshold value respectively to obtain corresponding corrected smoke concentration and corrected threshold value; if the corrected smoke concentration reaches the corrected threshold value, calculating a comprehensive risk value based on the corrected smoke concentration and the temperature rising speed; and if the comprehensive risk value reaches a risk threshold value, determining that it is a real fire. In the application, the fusion processing of the smoke sensor, the temperature sensor and the human body sensor is combined to overcome the defect that the current smoke detection of fire is prone to false positives due to environmental interference.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a smoke detection method based on multi-sensor fusion analysis. Background Technology

[0002] Traditional fire detectors, such as those based on temperature, smoke, and light sensing principles, have many limitations. These detectors can only judge based on specific, single information characteristics and are highly susceptible to environmental and spatial influences. In open or harsh environments, interference from factors such as temperature and smoke concentration slows down sensor signal transmission, weakens signal strength, or even causes malfunction, resulting in poor recognition performance and a high false alarm rate. For example, in dusty industrial plants, ordinary smoke sensors often trigger false alarms due to dust interference.

[0003] With technological advancements, video image-based fire detection technology has gradually emerged. This technology overcomes the drawback of traditional fire detectors, which require proximity to the fire source to operate. It is suitable for complex and expansive outdoor environments and can provide rich visual information about early fires. However, this technology also faces challenges. Due to the complexity and variability of scenes and the uncertainty of environmental factors, it is prone to misjudgment when distinguishing smoke from similar interfering substances (such as fog, clouds, and water mist). For example, in foggy weather, video image-based smoke detection algorithms may misjudge fog as smoke, issuing false alarms.

[0004] Among numerous fire detection elements, smoke, as an early sign of a fire, plays a decisive role in the reliability of fire detection systems due to the accuracy of its detection results. Current technologies have shortcomings in smoke detection. Some algorithms rely solely on the smoke energy detected at a specific moment, ignoring potential temporary interferences, such as drifting white pollutants, which can cause energy attenuation and affect detection accuracy. Furthermore, existing algorithms struggle to accurately distinguish between interfering objects that closely resemble smoke characteristics, further reducing detection reliability.

[0005] Furthermore, traditional fire detection methods lack the ability to adapt to different environments and scenarios. In some special locations, such as dusty environments like kitchens and factories, or areas with large variations in day and night light intensity, fixed detection thresholds and sensitivity settings cannot meet the needs for accurate detection, easily leading to missed or false alarms. Summary of the Invention

[0006] The main objective of this invention is to provide a smoke detection method based on multi-sensor fusion analysis, which aims to overcome the shortcomings of current smoke detection methods for fires, which are easily affected by environmental interference and produce false alarms.

[0007] To achieve the above objectives, the present invention provides a smoke detection method based on multi-sensor fusion analysis, comprising the following steps:

[0008] When the smoke sensor detects that the smoke concentration within the detection range reaches a first threshold, the temperature sensor detects the rate of temperature rise within the detection range, and the human body sensor detects the uniformity of environmental thermal radiation within the detection range.

[0009] Based on the temperature rise rate and the uniformity of environmental thermal radiation, corresponding correction coefficients are generated to correct the smoke concentration and the first threshold, respectively, to obtain the corresponding corrected smoke concentration and corrected threshold.

[0010] If the corrected smoke concentration reaches the correction threshold, a comprehensive risk value is calculated based on the corrected smoke concentration and the rate of temperature rise.

[0011] If the overall risk value reaches the risk threshold, it is determined to be a real fire.

[0012] Furthermore, after determining it to be a real fire, this includes:

[0013] The risk level is determined based on the predicted heat source diffusion rate, the location of personnel located by the human body sensor, and the smoke concentration, and a differentiated alarm is output based on the risk level.

[0014] Furthermore, based on the rate of temperature rise and the uniformity of environmental thermal radiation, corresponding correction coefficients are generated to correct the smoke concentration and the first threshold, respectively, to obtain the corresponding corrected smoke concentration and corrected threshold, including:

[0015] If the rate of temperature rise is less than the second threshold, a smoke concentration correction coefficient is generated to correct the smoke concentration, resulting in a corrected smoke concentration.

[0016] If the uniformity of environmental thermal radiation exceeds the preset uniformity, a judgment threshold correction coefficient is generated to correct the first threshold, thus obtaining the corrected threshold.

[0017] Furthermore, based on the corrected smoke concentration and temperature rise rate, a comprehensive risk value is calculated, including:

[0018] Obtain the starting time T1 of the smoke concentration increase and the time T2 when the temperature rise rate exceeds the speed threshold; based on the time difference between T1 and T2, mark the corresponding causal relationship and assign the corresponding causal coefficient;

[0019] Obtain the preset characteristic ranges of the smoldering stage and the open flame stage. Different characteristic ranges of different stages correspond to different corrected smoke concentrations and temperature rise rates.

[0020] After the corrected smoke concentration and temperature rise rate are matched to the corresponding stage characteristic interval, a corresponding compensation factor is generated based on the fluctuation of the temperature rise rate; wherein, the generation rules of the compensation factor are different in different stage characteristic intervals.

[0021] Based on the labeled causal relationships and the matched stage feature intervals, the weights corresponding to the corrected smoke concentration and temperature rise rate are determined.

[0022] Based on the corresponding weights, the modified smoke concentration and temperature rise rate are fused and calculated to obtain a basic risk score; the basic risk score is then multiplied by the causality coefficient and the compensation factor in sequence to obtain the comprehensive risk value.

[0023] Furthermore, based on the labeled causal relationships and the matched stage feature intervals, the weights corresponding to the corrected smoke concentration and temperature rise rate are determined, including:

[0024] Obtain a preset weight mapping relationship; wherein, the weight mapping relationship includes weight combinations corresponding to different causal relationships and stage feature intervals;

[0025] Based on the causal relationship and the matched stage feature intervals, the weights corresponding to the corrected smoke concentration and temperature rise rate are determined through the weight mapping relationship.

[0026] Furthermore, before performing the fusion calculation on the corrected smoke concentration and temperature rise rate based on the corresponding weights, the following steps are included:

[0027] The corrected smoke concentration and temperature rise rate were normalized to the 0-1 range.

[0028] Further, acquiring the rate of temperature rise within the detection range detected by the temperature sensor includes:

[0029] The temperature sensor collects temperature data from multiple different points within the detection range at a preset sampling frequency;

[0030] Calculate the average temperature of each point at the same time, and calculate the rate of temperature rise based on the difference between the average temperature of adjacent times. If the fluctuation of the rate of temperature rise of multiple samples is less than the fluctuation threshold, the average of the rate of temperature rise of multiple samples is taken as the final rate of temperature rise.

[0031] If the fluctuation of the temperature rise rate after multiple samplings reaches the fluctuation threshold, the preset sampling frequency of the temperature sensor is increased, and after multiple additional samplings, abnormal sampling values ​​are removed; the average value of the remaining temperature rise rate is used as the final temperature rise rate.

[0032] Further, acquiring the uniformity of environmental thermal radiation within the detection range detected by the human body sensor includes:

[0033] The detection range is divided into a core area and an outer area by a preset radius; the core area is the core region in the detection range where the smoke concentration reaches the first threshold.

[0034] Thermal radiation intensity data were collected from multiple sampling points in the core and outer areas using human body sensors.

[0035] Calculate the standard deviation of thermal radiation intensity in the core area, the standard deviation of thermal radiation intensity in the outer area, and the difference between thermal radiation intensity in the core area and the outer area;

[0036] Based on the standard deviation of the thermal radiation intensity in the core area, the standard deviation of the thermal radiation intensity in the outer area, and the difference in thermal radiation intensity between the core area and the outer area, the uniformity of environmental thermal radiation is calculated using a preset algorithm.

[0037] Furthermore, when calculating the comprehensive risk value based on the corrected smoke concentration, temperature rise rate, and environmental thermal radiation uniformity, the following is also included:

[0038] The duration during which the corrected smoke concentration reaches the correction threshold is obtained, and a duration correction factor is generated based on the duration.

[0039] The comprehensive risk value is calculated based on the duration correction factor, the corrected smoke concentration, and the rate of temperature rise.

[0040] The smoke detection method based on multi-sensor fusion analysis provided by this invention includes: when a smoke sensor detects that the smoke concentration within a detection range reaches a first threshold, acquiring the temperature rise rate within the detection range detected by a temperature sensor, and acquiring the environmental thermal radiation uniformity within the detection range detected by a human body sensor; based on the temperature rise rate and environmental thermal radiation uniformity, generating corresponding correction coefficients to correct the smoke concentration and the first threshold respectively, obtaining corresponding corrected smoke concentration and corrected threshold; if the corrected smoke concentration reaches the corrected threshold, calculating a comprehensive risk value based on the corrected smoke concentration and temperature rise rate; if the comprehensive risk value reaches the risk threshold, determining it as a real fire. In this invention, by combining the fusion processing of smoke sensors, temperature sensors, and human body sensors, the shortcomings of current smoke-based fire detection methods, which are easily affected by environmental interference and produce false alarms, are overcome. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the steps of a smoke detection method based on multi-sensor fusion analysis in one embodiment of the present invention;

[0042] Figure 2 This is a structural block diagram of a smoke detection device based on multi-sensor fusion analysis in one embodiment of the present invention;

[0043] Figure 3 This is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0044] The implementation, functional features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0046] Reference Figure 1 One embodiment of the present invention provides a smoke detection method based on multi-sensor fusion analysis, comprising the following steps:

[0047] Step S1: When the smoke sensor detects that the smoke concentration within the detection range reaches the first threshold, the temperature sensor detects the temperature rise rate within the detection range, and the human body sensor detects the environmental thermal radiation uniformity within the detection range.

[0048] Step S2: Based on the temperature rise rate and the uniformity of environmental thermal radiation, generate corresponding correction coefficients to correct the smoke concentration and the first threshold respectively, and obtain the corresponding corrected smoke concentration and corrected threshold.

[0049] Step S3: If the corrected smoke concentration reaches the correction threshold, calculate the comprehensive risk value based on the corrected smoke concentration and the rate of temperature rise.

[0050] Step S4: If the comprehensive risk value reaches the risk threshold, it is determined to be a real fire.

[0051] In this embodiment, as described in step S1 above, the detection result of the smoke sensor is used as the initial trigger condition. When the smoke sensor detects in real time that the smoke concentration value within its detection range reaches a preset first threshold (this threshold is the initial smoke concentration critical value for triggering a suspected fire determination), the multi-sensor linkage acquisition process is immediately initiated.

[0052] On one hand, a data acquisition command is sent to the temperature sensor, which then monitors temperature changes within the same detection range and calculates the rate of temperature change per unit time to obtain the core parameter of temperature rise rate. On the other hand, a mode switching and data acquisition command is sent to the human body sensor (PIR sensor). At this time, the thermal radiation distribution within the detection range of the human body sensor is calculated and output as an environmental thermal radiation uniformity parameter through a preset algorithm, providing data support for subsequent correction logic. The core purpose of this step is to quickly supplement two types of parameters directly related to fire characteristics after the smoke concentration triggers an initial suspected alarm. In one embodiment, the working principle of the human body infrared sensor is based on the infrared energy of thermal radiation and its own thermal characteristics, which can detect temperature changes caused by thermal radiation from heat sources. In smoke detection scenarios, it can be used to detect changes in human body thermal radiation in different areas. By analyzing the thermal radiation intensity data of multiple sampling points within the detection range, the standard deviation of thermal radiation intensity and the difference between regions are calculated to determine the environmental thermal radiation uniformity.

[0053] As described in step S2 above, for the correction of smoke concentration, a specific smoke concentration correction coefficient is generated based on the rate of temperature rise. The rate of temperature rise is compared with a preset second threshold (this threshold is used to distinguish between fire heat sources and non-fire ambient temperature environments) to determine whether the smoke is driven by a heat source. If the rate of temperature rise is lower than the second threshold, it indicates that there is no significant heat source at present, and a corresponding correction coefficient is generated to reduce the judgment weight of the original smoke concentration; if the rate of temperature rise is not lower than the second threshold, it indicates that there is a continuous heat source, and a corresponding correction coefficient is generated to maintain or adjust the weight of the original smoke concentration, ultimately obtaining the corrected smoke concentration.

[0054] For the correction of the first threshold, a specific judgment threshold correction coefficient is generated based on the uniformity of ambient thermal radiation. The uniformity of ambient thermal radiation is compared with a preset uniformity threshold (this threshold is used to distinguish between local high temperatures in a fire and a uniform thermal environment in a non-fire environment) to determine the correlation between smoke and fire. If the uniformity of ambient thermal radiation is higher than the preset uniformity threshold, it means that the thermal radiation distribution is balanced and there are no local high temperatures. A corresponding correction coefficient is generated to increase the first threshold to reduce false alarms. If the uniformity of ambient thermal radiation is not higher than the preset uniformity threshold, it indicates that there are local high temperatures, which is consistent with the characteristics of a fire heat source. The system then generates a corresponding correction coefficient to maintain or adjust the first threshold, and finally obtains the corrected threshold.

[0055] As described in step S3 above, when the corrected smoke concentration value obtained in step S2 reaches the correction threshold, it indicates that the current smoke has passed the initial authenticity verification. Then, step S3 is entered to calculate the comprehensive risk value, and the calculation process only revolves around two core parameters: corrected smoke concentration and temperature rise rate.

[0056] First, the modified smoke concentration and temperature rise rate are preprocessed using a preset algorithm, such as normalizing the two parameters to the 0-1 range to ensure their compatibility. Second, the weights of the two parameters are determined by preset rules. The weight allocation can be related to the characteristics of the fire development stage. For example, the weight of modified smoke concentration is emphasized in the smoldering stage, while the weight of temperature rise rate is emphasized in the open flame stage. The specific rules will be subject to the subsequent refinement of the plan. Finally, based on the determined weights, the preprocessed modified smoke concentration and temperature rise rate are fused and calculated to obtain a quantified comprehensive risk value. This value directly reflects the degree to which the current scenario conforms to the characteristics of a real fire.

[0057] As described in step S4 above, the comprehensive risk value calculated in step S3 is compared with a preset risk threshold. If the comprehensive risk value reaches or exceeds the risk threshold, it indicates that the smoke, temperature, and other multi-dimensional parameters in the current scenario all meet the characteristics of a fire, and the risk level has reached the level requiring an early warning. Therefore, it is determined to be a real fire, and the subsequent alarm and emergency response mechanisms are triggered. If the comprehensive risk value does not reach the risk threshold, it indicates that the current scenario may be a suspected situation caused by interfering factors (such as brief smoke, local temperature fluctuations, etc.). It is not determined to be a real fire, and the system returns to the initial monitoring state to continue monitoring the parameters within the detection range. This final determination step ensures the accuracy of fire early warning, avoiding the interference of false alarms with normal life and preventing safety hazards caused by missed alarms.

[0058] In one embodiment, after determining that a fire is a real fire, the process includes:

[0059] The risk level is determined based on the predicted heat source diffusion rate, the location of personnel located by the human body sensor, and the smoke concentration, and a differentiated alarm is output based on the risk level.

[0060] In this embodiment, after determining that a real fire has occurred, the rate of heat source diffusion is first calculated based on the rate of temperature rise using a preset model (the faster the temperature rises, the higher the rate of heat source diffusion). Simultaneously, human body sensors are used to determine the specific locations of personnel within the detection range. This, combined with smoke concentration (the higher the concentration, the greater the fire's impact and hazard), forms the basis for a risk level classification (e.g., high, medium, and low: high risk if the heat source diffuses quickly, personnel are in the core area, and smoke concentration is high; low risk if the heat source diffuses slowly, personnel are in a safe area, and smoke concentration is low). Finally, differentiated alarms are output according to different risk levels. For example, high risk triggers audible and visual alarms, emergency evacuation broadcasts, and fire alarm linkage; medium risk triggers local audible and visual alarms and personnel guidance notifications; and low risk triggers only basic audible and visual alarms, ensuring that the alarm response matches the actual severity of the fire.

[0061] In one embodiment, based on the rate of temperature rise and the uniformity of ambient thermal radiation, corresponding correction coefficients are generated to correct the smoke concentration and the first threshold, respectively, to obtain the corresponding corrected smoke concentration and corrected threshold, including:

[0062] If the rate of temperature rise is less than the second threshold, a smoke concentration correction coefficient is generated to correct the smoke concentration, resulting in a corrected smoke concentration.

[0063] If the uniformity of environmental thermal radiation exceeds the preset uniformity, a judgment threshold correction coefficient is generated to correct the first threshold, thus obtaining the corrected threshold.

[0064] In this embodiment, a smoke concentration correction mechanism is activated when the rate of temperature rise is less than a second threshold. The second threshold is a critical value used to distinguish between fire heat sources and non-fire environments. A temperature rise rate less than this threshold means there is no significantly and continuously increasing heat source within the detection range, and the smoke at this time is more likely caused by non-fire interference factors such as dust and fumes. Based on the difference between the temperature rise rate and the second threshold, a corresponding smoke concentration correction coefficient is generated using a preset algorithm. This coefficient is less than 1 and is used to reduce the weight of the original smoke concentration in the judgment. The original smoke concentration is multiplied by this correction coefficient to obtain the corrected smoke concentration, i.e., the corrected smoke concentration. The core of this correction process is to filter out non-fire smoke without heat source support through temperature parameters, ensuring that subsequent judgments are based on smoke intensity data that more closely reflects the characteristics of a real fire.

[0065] When the uniformity of ambient thermal radiation exceeds a preset uniformity threshold, a first threshold correction mechanism is activated. The preset uniformity threshold is a critical value that distinguishes between localized high temperatures in a fire and a uniform thermal environment in a non-fire context. An ambient thermal radiation uniformity exceeding this threshold indicates a balanced thermal radiation distribution within the detection range, with no obvious localized high-temperature points. In this case, the smoke has a low correlation with a fire and is more likely generated by non-fire scenarios such as cooking or equipment cooling. Based on the deviation between the ambient thermal radiation uniformity and the preset uniformity threshold, a corresponding judgment threshold correction coefficient is generated using a preset algorithm. This coefficient is greater than 1 and is used to improve the judgment standard of the first threshold. Multiplying the first threshold by this correction coefficient yields the corrected judgment threshold. The core of this correction process is to adjust the judgment standard based on the characteristics of thermal radiation distribution, reducing false alarms in non-fire scenarios and ensuring that only smoke with fire heat source characteristics passes subsequent judgment.

[0066] In one embodiment, calculating the comprehensive risk value based on the corrected smoke concentration and temperature rise rate includes:

[0067] Obtain the starting time T1 of the smoke concentration increase and the time T2 when the temperature rise rate exceeds the speed threshold; based on the time difference between T1 and T2, mark the corresponding causal relationship and assign the corresponding causal coefficient;

[0068] Obtain the preset characteristic ranges of the smoldering stage and the open flame stage. Different characteristic ranges of different stages correspond to different corrected smoke concentrations and temperature rise rates.

[0069] After the corrected smoke concentration and temperature rise rate are matched to the corresponding stage characteristic interval, a corresponding compensation factor is generated based on the fluctuation of the temperature rise rate; wherein, the generation rules of the compensation factor are different in different stage characteristic intervals.

[0070] Based on the labeled causal relationships and the matched stage feature intervals, the weights corresponding to the corrected smoke concentration and temperature rise rate are determined.

[0071] Based on the corresponding weights, the modified smoke concentration and temperature rise rate are fused and calculated to obtain a basic risk score; the basic risk score is then multiplied by the causality coefficient and the compensation factor in sequence to obtain the comprehensive risk value.

[0072] In this embodiment, two key time points are first recorded: the starting time T1 when the smoke concentration begins to rise, and the time T2 when the rate of temperature rise exceeds a preset threshold. By calculating the time difference between T1 and T2, the correlation between the rise in smoke concentration and the rise in temperature is analyzed, thereby marking the causal relationship between the two and assigning a corresponding causal coefficient. For example, if the time difference between T1 and T2 is extremely small (e.g., less than 3 seconds), it indicates that the smoke and temperature rise occur almost synchronously, and there is a strong causal relationship between the two, so the system assigns a high causal coefficient (e.g., 1.5); if the time difference is large (e.g., more than 10 seconds), it indicates that the correlation between the two is weak and may be caused by different factors, so a low causal coefficient (e.g., 0.6) is assigned. The core of this step is to strengthen the characteristic weight of the synchronous changes of smoke and temperature in real fires through correlation analysis in the time dimension, and weaken the interference of isolated occurrences of smoke and temperature in non-fire scenarios.

[0073] Next, two preset characteristic intervals for different fire stages are invoked: the smoldering stage characteristic interval and the open flame stage characteristic interval. The smoldering stage characteristic interval corresponds to a parameter combination of high but slow-rising corrected smoke concentration and low, stable temperature rise rate. The open flame stage characteristic interval corresponds to a parameter combination of rapidly spreading corrected smoke concentration and high, potentially fluctuating, temperature rise rate. The real-time values ​​of corrected smoke concentration and temperature rise rate are compared with the characteristic parameters of these two intervals to determine the current suspected fire stage. This matching process provides a staged basis for subsequent compensation factor generation and weight allocation, ensuring that subsequent calculations adapt to the characteristic patterns of different fire stages.

[0074] After correcting the smoke concentration and temperature rise rate to match the corresponding stage characteristic range, the fluctuation of the temperature rise rate is further analyzed (e.g., by calculating the fluctuation amplitude through the standard deviation or range of multiple consecutive samples), and a corresponding compensation factor is generated. The generation rules vary depending on the stage. In the smoldering stage characteristic range, the temperature rise rate fluctuates relatively little under normal circumstances. If the actual fluctuation amplitude is less than the preset fluctuation threshold for this stage, it indicates that the heat source is stable and conforms to the smoldering characteristics, generating a compensation factor greater than 1 (e.g., 1.1). If the fluctuation amplitude exceeds the threshold, it indicates the presence of abnormal interference (e.g., the influence of local airflow), generating a compensation factor less than 1 (e.g., 0.9). In the open flame stage characteristic range, the temperature rise rate is allowed to fluctuate to a certain extent but must be within a reasonable range. If the fluctuation amplitude is within the preset normal range for this stage, it indicates that it conforms to the dynamic characteristics of open flame combustion, generating a higher compensation factor (e.g., 1.2). If the fluctuation amplitude is too large, it indicates abnormal heat source stability (e.g., non-fire high temperature interference), generating a significantly reduced compensation factor (e.g., 0.7). Through this differentiated rule, the authenticity of the heat source at different stages can be accurately identified, reducing the impact of environmental interference on risk assessment.

[0075] Then, based on the labeled causal relationships (i.e., the strength of the relationship corresponding to the time difference between T1 and T2) and the matched stage characteristic intervals (smoldering or open flame stages), the weights corresponding to the corrected smoke concentration and temperature rise rate are retrieved from the preset weight mapping relationship. For example, in the smoldering stage and in scenarios with strong causal relationships, the corrected smoke concentration has a greater impact on fire determination, with a weight allocation of 70%, while the temperature rise rate has a weight of 30%; in the open flame stage and in scenarios with strong causal relationships, the weight of the temperature rise rate is increased to 60%, while the weight of the corrected smoke concentration is reduced to 40%; if the causal relationship is weak, the weight difference between the two types of parameters will be reduced (e.g., 50% each) to reduce the excessive influence of a single parameter. This step, by dynamically adjusting the weights, makes the risk calculation more closely match the core characteristics of fires at different stages.

[0076] Finally, the modified smoke concentration and temperature rise rate are normalized (e.g., converted to standardized values ​​in the 0-1 range), and then weighted and summed according to the weights determined in the previous step to obtain the basic risk score. Subsequently, the basic risk score is multiplied by the causality coefficient and the compensation factor in sequence to obtain the final comprehensive risk value. For example, if the basic risk score is 0.6, the causality coefficient is 1.5, and the compensation factor is 1.1, then the comprehensive risk value is 0.6 × 1.5 × 1.1 = 0.99. This calculation process, through multi-level parameter correction (strengthening the temporal logic of causal relationships, optimizing the adaptability of the compensation factor stage, and highlighting core features through weight allocation), ultimately outputs a comprehensive risk value that can objectively quantify the degree to which the current scenario conforms to the characteristics of a real fire, providing accurate quantitative basis for subsequent fire determination.

[0077] In one embodiment, based on the labeled causal association and the matched stage feature intervals, the weights corresponding to the corrected smoke concentration and temperature rise rate are determined, including:

[0078] Obtain a preset weight mapping relationship; wherein, the weight mapping relationship includes weight combinations corresponding to different causal relationships and stage feature intervals;

[0079] Based on the causal relationship and the matched stage feature intervals, the weights corresponding to the corrected smoke concentration and temperature rise rate are determined through the weight mapping relationship.

[0080] In this embodiment, a pre-defined three-dimensional weight mapping table of causal association, stage feature interval, and weight combination is first obtained from local storage or a cloud database. This mapping table is not a fixed, single rule, but a standardized system formed after calibration based on a large amount of experimental data from fire scenarios and risk assessment models. Its horizontal dimension covers two types of causal association states: strong causal association and weak causal association. Its vertical dimension covers two types of stage feature intervals: smoldering stage and open flame stage. The intersection of these two dimensions forms four combined scenarios, each corresponding to a unique weight combination (weight for corrected smoke concentration and weight for temperature rise rate). For example, in a strong causal-smoldering scenario, the numerical setting of the weight combination is based on the principle of highlighting core risk parameters and weakening interference parameters. For instance, in a strong causal scenario, priority is given to matching the core stage parameters (heavy smoke in smoldering, heavy temperature in open flame), while in a weak causal scenario, priority is given to strengthening the basic judgment parameter of smoke. The core value of this step lies in avoiding the randomness of weight allocation through a pre-defined standardized mapping relationship, ensuring that the weight determination logic is consistent and traceable across different scenarios.

[0081] During matching, two key matching conditions are defined: first, the causal relationship marked in the previous causal logic verification stage (strong or weak causal relationship); and second, the stage characteristic interval determined in the previous fire stage matching stage (smoldering stage or open flame stage). Subsequently, using these two conditions as search keywords, a cross-search is performed in the weight mapping relationship table obtained in step one. After the search is completed, the weight values ​​in this combination are directly extracted and assigned to the corrected smoke concentration and temperature rise rate, respectively. This dual-condition precise search method ensures that the weight allocation is causally reasonable and highly adapted to the current scenario and fire stage, avoiding risk calculation errors caused by a disconnect between weights and the scenario.

[0082] In one embodiment, before performing a fusion calculation on the corrected smoke concentration and temperature rise rate based on the corresponding weights, the following steps are included:

[0083] The corrected smoke concentration and temperature rise rate were normalized to the 0-1 range.

[0084] In one embodiment, acquiring the rate of temperature rise within the detection range detected by the temperature sensor includes:

[0085] The temperature sensor collects temperature data from multiple different points within the detection range at a preset sampling frequency;

[0086] Calculate the average temperature of each point at the same time, and calculate the rate of temperature rise based on the difference between the average temperature of adjacent times. If the fluctuation of the rate of temperature rise of multiple samples is less than the fluctuation threshold, the average of the rate of temperature rise of multiple samples is taken as the final rate of temperature rise.

[0087] If the fluctuation of the temperature rise rate after multiple samplings reaches the fluctuation threshold, the preset sampling frequency of the temperature sensor is increased, and after multiple additional samplings, abnormal sampling values ​​are removed; the average value of the remaining temperature rise rate is used as the final temperature rise rate.

[0088] In this embodiment, firstly, multi-point, standardized sampling ensures that the temperature data comprehensively reflects the distribution and changing trends of heat sources within the detection range, avoiding the limitations of single-point sampling. Specifically, a preset sampling frequency is configured for the temperature sensor (set based on the environmental stability of the detection scenario, such as 1 second / sample in normal scenarios and 0.5 seconds / sample in complex scenarios). This frequency needs to balance data timeliness and sensor power consumption. Simultaneously, based on the spatial size of the smoke sensor's detection range, multiple different points are divided within the detection range, and the temperature sensor synchronously collects real-time temperature data from each point at a preset frequency. Multi-point synchronous sampling avoids single-data deviations caused by localized temperature anomalies (such as brief device overheating), providing a comprehensive and objective raw data foundation for subsequent accurate calculations of the temperature rise rate.

[0089] Next, through mean calculation and fluctuation verification, stable temperature rise rates are selected to ensure that the data reflects the true temperature change trend. The specific operation consists of three stages: The first stage is mean aggregation, which takes the arithmetic mean of temperature data from multiple points at the same sampling time point (e.g., if the temperatures at three points at a certain moment are 25℃, 25.2℃, and 24.8℃, then the average temperature at that moment is 25℃), converting multi-point data into single-value data reflecting the overall temperature level and eliminating the interference of point differences on trend judgment; the second stage is velocity calculation, which calculates the single temperature rise rate (e.g., 0.3℃ / second) based on the average temperature of two adjacent sampling time points (e.g., 25℃ at time t1 and 25.3℃ at time t2, with a time interval of 1 second), and continuously calculates the temperature rise rate corresponding to multiple samplings (e.g., 3 times); the third stage... The third stage is fluctuation verification. A preset fluctuation threshold is retrieved (based on the temperature sensor accuracy and ambient temperature stability, such as 0.1℃ / second). The maximum difference between multiple sampled temperature rise rates is calculated (e.g., three rates are 0.28℃ / second, 0.3℃ / second, and 0.31℃ / second, with a maximum difference of 0.03℃ / second). If this difference is less than the fluctuation threshold, it indicates a stable temperature rise trend with no significant abnormal interference. The arithmetic mean of these multiple rates (e.g., 0.297℃ / second) is then used as the final temperature rise rate. This process of mean aggregation, rate calculation, and fluctuation verification ensures that the final temperature rise rate accurately reflects the stable trend of overall temperature change, avoiding misjudgments caused by single data fluctuations.

[0090] For scenarios with significant fluctuations in the rate of temperature rise, the accuracy and reliability of the final data are ensured by increasing the sampling frequency and removing outliers. The specific operational logic is as follows: First, when the fluctuation in the rate of temperature rise from multiple samples reaches or exceeds the fluctuation threshold (e.g., three rates are 0.2℃ / sec, 0.35℃ / sec, and 0.4℃ / sec, with a maximum difference of 0.2℃ / sec, exceeding the threshold of 0.1℃ / sec), it is determined that there are unstable factors in the current temperature change (e.g., fluctuating intensity of local heat sources, airflow interference, etc.). At this time, the preset sampling frequency of the temperature sensor is automatically increased to twice the original frequency (e.g., from 1 second / sample to 0.5 seconds / sample) to capture temperature change details with higher temporal resolution and reduce information loss. Second, multiple additional samples are taken at the increased frequency to obtain more dimensions of temperature rise rate data. Subsequently, outlier sampled values ​​are removed, and the average and standard deviation of all sampled rates are calculated. Values ​​exceeding three times the standard deviation of the average are identified as outliers and removed from the dataset. Finally, the arithmetic mean of the remaining non-outlier sampled values ​​is taken as the final temperature rise rate. For unstable scenarios, a dual calibration strategy of high-frequency sampling and anomaly rejection is adopted. This avoids data distortion caused by excessive fluctuations and ensures that the final temperature rise rate reflects the real heat source change trend through precise screening, providing high-quality temperature parameters for subsequent multi-sensor fusion analysis.

[0091] In one embodiment, obtaining the uniformity of environmental thermal radiation within the detection range detected by the human body sensor includes:

[0092] The detection range is divided into a core area and an outer area by a preset radius; the core area is the core region in the detection range where the smoke concentration reaches the first threshold.

[0093] Thermal radiation intensity data were collected from multiple sampling points in the core and outer areas using human body sensors.

[0094] Calculate the standard deviation of thermal radiation intensity in the core area, the standard deviation of thermal radiation intensity in the outer area, and the difference between thermal radiation intensity in the core area and the outer area;

[0095] Based on the standard deviation of the thermal radiation intensity in the core area, the standard deviation of the thermal radiation intensity in the outer area, and the difference in thermal radiation intensity between the core area and the outer area, the uniformity of environmental thermal radiation is calculated using a preset algorithm.

[0096] In this embodiment, the overall detection range is first divided into two sub-regions: a core area and an outer area, based on a preset radius. The core area is defined directly based on smoke concentration; specifically, it is the area within the detection range where the smoke concentration first reaches a first threshold. This area is a key monitoring area where smoke initially accumulates and the fire risk is high. The outer area is the region covered by a preset radius extending outward from the core area. It is used to capture the thermal radiation distribution characteristics around the core area, forming supplementary monitoring of the core area. This zoning method enables differentiated monitoring of the initial smoke generation area and its surrounding environment.

[0097] After the area division is completed, the human body sensor switches to thermal radiation intensity acquisition mode, performing multi-point sampling on both the core and peripheral areas. For the core area, the sensor collects thermal radiation intensity data from multiple locations according to a preset sampling point distribution rule (such as a uniform grid distribution). For the peripheral area, the same sampling rule is used, collecting equal or proportional amounts of thermal radiation intensity data within its range. The purpose of multi-point sampling is to avoid the random errors of a single sampling point, ensuring that the acquired thermal radiation intensity data accurately reflects the overall thermal distribution characteristics of both areas, providing reliable data support for subsequent uniformity calculations.

[0098] Statistical analysis was performed on the collected thermal radiation intensity data of the core and peripheral areas, generating three key parameters: First, the standard deviation of the thermal radiation intensity in the core area, reflecting the dispersion of thermal radiation intensity at different sampling points within the core area; a smaller standard deviation indicates a more uniform thermal radiation distribution within the core area. Second, the standard deviation of the thermal radiation intensity in the peripheral area, serving the same purpose as the core area's standard deviation, measuring the uniformity of thermal radiation distribution within the peripheral area. Third, the difference between the thermal radiation intensity of the core and peripheral areas, calculated by the absolute difference between the average thermal radiation intensity of the two areas; a smaller difference indicates a more balanced thermal radiation distribution between the areas. These three parameters, from the perspectives of internal uniformity and inter-regional consistency, jointly characterize the thermal radiation distribution features within the detection range.

[0099] The three statistical parameters (standard deviation of the core area, standard deviation of the outer area, and inter-regional difference) are input into a preset algorithm for fusion calculation to obtain the environmental thermal radiation uniformity. The core logic of the preset algorithm is to weight the three parameters, where the standard deviation and inter-regional difference are negatively correlated with uniformity (i.e., the smaller the parameter value, the higher the uniformity). For example, the preset algorithm first normalizes the three parameters to the 0-1 range, then calculates a weighted sum using preset weights (e.g., 40% for the standard deviation of the core area, 30% for the standard deviation of the outer area, and 30% for the inter-regional difference), and finally subtracts this weighted sum from 1 to obtain the environmental thermal radiation uniformity (value range 0-1, the closer the value is to 1, the more uniform the thermal radiation distribution). Through this multi-parameter fusion calculation, the overall uniformity of thermal radiation within the detection range can be comprehensively reflected, providing a quantitative basis for the subsequent correction of the first threshold.

[0100] In one embodiment, when calculating the comprehensive risk value based on the corrected smoke concentration, temperature rise rate, and environmental thermal radiation uniformity, the method further includes:

[0101] The duration during which the corrected smoke concentration reaches the correction threshold is obtained, and a duration correction factor is generated based on the duration.

[0102] The comprehensive risk value is calculated based on the duration correction factor, the corrected smoke concentration, and the rate of temperature rise.

[0103] In this embodiment, firstly, the stability and severity of the risk are determined by the duration of the corrected smoke concentration exceeding the threshold, providing a supplementary time dimension for the risk value. The specific operation consists of two stages: the first stage is duration acquisition, starting from the moment the corrected smoke concentration first reaches the correction threshold, continuously monitoring and recording the continuous duration for which the concentration remains at or above the correction threshold. This duration directly reflects the sustained intensity of the risk; the longer the duration, the more stable the smoke risk and the lower the probability of non-transient interference. The second stage involves generating a duration correction factor. A pre-defined factor mapping rule based on duration is used, which follows the logic that the longer the risk persists, the higher the fire probability: if the duration is ≤3 seconds (instantaneous exceeding of the threshold, possibly due to interference), a duration correction factor of 0.7-0.8 is generated (reducing the risk weight to avoid misjudgment due to instantaneous interference); if 3 seconds < duration ≤8 seconds (short-term duration, risk requires vigilance), a duration correction factor of 1.0-1.1 is generated (maintaining or slightly increasing the risk weight); if the duration >8 seconds (long-term duration, highly stable risk), a duration correction factor of 1.3-1.4 is generated (significantly increasing the risk weight and strengthening the impact of stable risks). Factor values ​​are calibrated using historical fire data to ensure accurate matching of the actual risk probability corresponding to different durations. This overcomes the limitations of traditional risk assessment relying solely on concentration values. By verifying the time dimension, it eliminates misjudgments caused by instantaneous interference, while highlighting the threat level of stable risks, providing a more comprehensive basis for determining the overall risk value.

[0104] When calculating the comprehensive risk value, the duration correction factor is calculated collaboratively with the original two parameters (corrected smoke concentration and temperature rise rate) to generate a comprehensive quantitative result that reflects concentration intensity, temperature trend, personnel risk, and time duration. Integrating the duration correction factor as an independent dimension into the risk assessment forms a three-dimensional synergy of numerical value, trend, and duration with the original parameters. This retains the advantages of multi-parameter fusion while strengthening the impact of risk duration through the time factor. Its technical value lies in ensuring that the comprehensive risk value more accurately matches the actual fire development pattern (fire risk increases over time), avoiding misjudgments of low risk or omissions of high risk due to neglecting the time dimension, and further improving the scientific rigor and reliability of the risk assessment.

[0105] In one embodiment, before the smoke sensor detects that the smoke concentration within the detection range has reached a first threshold, the method further includes:

[0106] After the smoke sensor is activated, it first collects the initial particulate matter content and day and night light intensity within the detection range through the built-in environment type recognition module;

[0107] Based on the initial particulate matter baseline, the first threshold is automatically adjusted, and the detection sensitivity level of the smoke sensor is adjusted at the same time.

[0108] Based on day and night light intensity, the sampling interval of the smoke sensor is adjusted, and the power consumption mode of the smoke sensor is also adjusted.

[0109] In this embodiment, after the smoke sensor completes its power-on startup, it does not directly enter the conventional detection mode. Instead, it first activates the built-in environment type identification module (integrating a micro particulate matter detector and a light sensor) to collect two key environmental parameters within the detection range: one is the initial particulate matter baseline, which is the baseline value of particulate matter concentration (unit: mg / m³) in the air with a particle size of 0.3-10 μm when there is no significant smoke interference within the detection range. 3 The system uses two main parameters: first, it assesses the degree of dust pollution in the environment (e.g., kitchens and factories have higher initial particulate matter levels, while bedrooms and offices have lower levels); second, it measures day and night light intensity, i.e., real-time light intensity values ​​(unit: lux) within the detection range, to distinguish whether the current scene is daytime (light intensity ≥ 100 lux) or nighttime (light intensity < 50 lux), and to determine the frequency of human activity in the environment and the sensor's power consumption requirements (there is less human activity at night, requiring a balance between detection accuracy and power consumption). By collecting environmental parameters in the early stages, the sensor accurately perceives the inherent characteristics of the current detection scene, providing a basis for subsequent adjustments and avoiding the inapplicability of fixed parameters in complex environments.

[0110] Based on the initial particulate matter concentration, the first threshold and detection sensitivity are optimized to reduce the interference of inherent environmental dust on smoke detection and ensure that the sensor can accurately identify fire smoke. The specific operation consists of two parts: the first part is the adjustment of the first threshold, which presets a threshold adjustment rule based on the initial particulate matter concentration. This rule follows the logic that the higher the environmental dust concentration, the higher the first threshold should be to avoid false alarms: if the initial particulate matter concentration is <0.1 mg / m³... 3 (In low-dust environments, such as bedrooms), if environmental interference is deemed minimal, the original first threshold (e.g., 0.3 mg / m³) should be maintained. 3 ) remains unchanged; if 0.1 mg / m 3 ≤Initial particulate matter baseline <0.2 mg / m³ 3 (In a medium-dust environment, such as an office), if the environment is deemed to have slight disturbance, the first threshold is increased by 10% (e.g., 0.33 mg / m³). 3 If the initial particulate matter concentration is ≥0.2 mg / m³ 3 (In high-dust environments, such as kitchens and workshops), if the environmental interference is deemed significant, the first threshold will be increased by 20% (e.g., 0.36 mg / m³). 3 The second part involves adjusting the detection sensitivity level. The sensitivity level is inversely related to the initial particulate matter concentration: in low-dust environments, the sensitivity is set to the high-efficiency level (capable of identifying tiny fire smoke particles of 0.3-0.5μm) to ensure that early, weak smoke can be captured; in medium-dust environments, it is set to the standard level (identifying only fire characteristic particles of 0.3-2μm, filtering out some dust particles); in high-dust environments, it is set to the anti-interference level (identifying only typical fire smoke particles of 0.3-1μm, further eliminating interference from large-particle dust). The threshold and sensitivity adjustment parameters have been calibrated through numerous simulation experiments in different dust environments to ensure a balance between preventing false alarms and missed alarms. Through environmental dust adaptation adjustments, the first threshold and sensitivity can dynamically match the inherent interference of the environment, avoiding frequent false alarms caused by excessively low thresholds in high-dust environments, while ensuring detection accuracy in low-dust environments and improving the sensor's environmental adaptability.

[0111] Based on the distinction between daytime and nighttime scenes according to daytime and nighttime light intensity, the sampling interval and power consumption mode are optimized to ensure both detection requirements and energy efficiency, especially suitable for battery-powered sensor scenarios. The specific operation is divided into two parts: The first part is the adjustment of the sampling interval, which presets the judgment threshold of daytime and nighttime light intensity (e.g., 50 lux as the daytime and nighttime boundary): If the daytime and nighttime light intensity is ≥50 lux, the sampling interval is set to 1 second / time to ensure real-time capture of smoke changes through high-frequency sampling; if the daytime and nighttime light intensity is <50 lux, the sampling interval is extended to 2 seconds / time to reduce the data collection frequency while ensuring basic detection capabilities. The second part involves power consumption mode adjustment, which is coordinated with the sampling interval: During high-frequency sampling in the daytime, the sensor switches to high-performance mode, with all detection modules (particulate matter detection, signal processing) operating at full load to ensure real-time data acquisition and signal analysis. During low-frequency sampling at night, it switches to low-power mode, shutting down non-core modules (such as the environment type recognition module) during sampling intervals, retaining only the core smoke detection circuit for low-power operation, while reducing the operation frequency of the signal processing unit, thus reducing sensor power consumption to 60%-70% of the daytime mode. If the light intensity is in the transitional range of 50-100 lux (such as dawn and dusk), a compromise mode is adopted, with the sampling interval set to 1.5 seconds / sample, and the power consumption mode adjusted to a balanced level (core modules operate normally, non-core modules work intermittently). Through day and night scene adaptation, the sensor achieves an optimal balance between sampling efficiency and power consumption at different times: high-frequency sampling during the day ensures safety, while low power consumption at night extends battery life, meeting the detection needs of different scenarios and improving the sensor's practical lifespan and economy, especially suitable for independent installation scenarios without continuous power supply.

[0112] In one embodiment, the method further includes:

[0113] Collect usage time data from smoke sensors, temperature sensors, and human body sensors, and construct a time data matrix;

[0114] Using the detection range as the origin, the three-dimensional coordinates of the three types of sensors are obtained to generate a relative position graph;

[0115] A preset data array containing corrected smoke concentration, temperature rise rate, and environmental thermal radiation uniformity is retrieved. The preset data array is then mutated based on the relative position graph to obtain the first data array.

[0116] The time data matrix is ​​cross-mutated with the first data array to obtain a mutated data array and a mutated data matrix.

[0117] The relative position graphics, the variation data array, and the variation data matrix are superimposed according to preset rules, and encoding factors are generated based on the features of the superimposed area to encode and transmit the data collected by the smoke sensor, temperature sensor, and human body sensor.

[0118] In this embodiment, the timing module and status monitoring unit built into the sensor first collect three types of core usage time data for each type of sensor: first, cumulative working time, reflecting the overall wear and tear of the sensor; second, effective working time over the past 30 days (actual detection time after excluding faults and dormant states, accurate to the hour), reflecting the recent operational stability of the sensor; and third, single continuous working time (uninterrupted operation time within the current detection cycle, accurate to the minute), reflecting the current working status of the sensor. Subsequently, a time data matrix is ​​constructed according to a two-dimensional structure of sensor type-time data type: the matrix rows correspond to the three types of sensors mentioned above, and the columns correspond to the three types of time data (cumulative working time, effective working time over the past 30 days, and single continuous working time). Each cell in the matrix is ​​filled with the corresponding time data value for the corresponding sensor. This transformation of the sensor usage time—a key state feature—into structured matrix data provides a unified data format for subsequent cross-fusion with detection parameter data. Simultaneously, the time data reflects the sensor's wear and tear status, ensuring that subsequent coding factors can be associated with the sensor's reliability characteristics.

[0119] Next, through 3D coordinate positioning and graphical processing, the spatial relationships of the three types of sensors are transformed into visual graphics, providing a basis for the spatial dimensional variations of the data array. The specific operation consists of two parts: The first part is 3D coordinate acquisition. Using the geometric center of the preset detection range of the smoke detector as the origin, the 3D coordinates of the smoke sensor, temperature sensor, and human body sensor are collected through the built-in IoT positioning module of the sensors. The coordinate values ​​reflect the spatial position of each sensor relative to the origin. The second part is relative position graphic generation. The collected 3D coordinates of the three types of sensors are imported into the graphics generation module to generate a 3D graphic centered on the origin, containing position markers for the three types of sensors. The graphic distinguishes the three types of sensors with different colors or shapes, and also marks the straight-line distance between each sensor and the origin, as well as the relative distance between the sensors. In addition, the graphic also marks the distance between each sensor and the boundary of the detection range to show whether the sensor position is in the optimal detection area.

[0120] Then, combining the spatial relationships reflected in the relative position graphs of the sensors, the data array of the core detection parameters is mutated to reflect the influence of spatial location on the detection data. The specific operation consists of three steps: The first step is to retrieve the preset data array. Core parameter data for smoke detection within a preset time period is retrieved from the database to construct the preset data array: the array's row dimensions correspond to detection time nodes (each row represents the detection data for one time node), and the column dimensions correspond to three types of core parameters (corrected smoke concentration, temperature rise rate, and environmental thermal radiation uniformity). Each cell is filled with the corresponding parameter detection value for the corresponding time node. The second step is to determine the spatial mutation rules. Key spatial features are extracted from the relative position graphs: one is the distance between each sensor and the origin, and the other is the distance between each sensor and the detection target area. Data mutation weights are set based on these two types of distances. The third step is the data mutation operation. According to the matching relationship between the parameter and the sensor-spatial weight, each type of parameter data in the preset data array is mutated: the spatial weight of the smoke sensor corresponds to the corrected smoke concentration, the spatial weight of the temperature sensor corresponds to the temperature rise rate, and the spatial weight of the human body sensor corresponds to the environmental thermal radiation uniformity. Each parameter value is multiplied by its corresponding weight to obtain the mutated parameter value.

[0121] Furthermore, the sensor state data in the time dimension (time data matrix) and the spatially mutated detection data (first data array) are cross-operated to achieve deep fusion and bidirectional mutation of the two types of data, generating structured data with both time and detection features. The specific operation consists of two parts: The first part is data dimension adaptation. Since the time data matrix and the first data array may have different dimensions, the time data matrix is ​​first expanded: the time data matrix is ​​copied N times according to the number of time nodes in the first data array and arranged in chronological order to form an expanded time matrix; simultaneously, the first data array is expanded according to sensor type to form an associated detection matrix, ensuring that the two types of data are completely matched in terms of time nodes and sensor type dimensions. The second part is the cross-mutation operation, which can use a cross-mutation algorithm of element-wise multiplication and eigenvalue compensation: for each time node, the time data of the corresponding sensor in the expanded time matrix is ​​multiplied point-by-point with the detection data of the corresponding sensor in the associated detection matrix to obtain basic cross data; then, the row mean of the time data matrix and the column mean of the first data array are extracted as compensation values ​​and superimposed on the basic cross data to enhance data features. After cross-mutation, the detection parameter dimension data and the time data dimension data are extracted separately: the cross-mutation data of the detection parameter dimension are arranged according to the time node to form a mutation data array; the cross-mutation data of the time data dimension are arranged according to the sensor type and the time node to form a mutation data matrix.

[0122] Finally, by overlaying spatial graphics, arrays, and matrices in multiple forms, features are extracted to generate encoding factors, enabling secure encoded transmission of sensor-collected data. In one embodiment, multiple key features are extracted from the overlay area, then converted into binary form and combined to generate encoding factors. During data encoding and transmission, the real-time data collected by the sensor is XORed with the encoding factors to generate encrypted encoded data. The encoding factors generated by overlaying multiple forms of data can uniquely associate the sensor's spatial location, usage status, and detection data characteristics, ensuring the security of encoded transmission and enabling data traceability and integrity verification through the encoding factors, preventing data tampering or leakage during transmission.

[0123] In one embodiment, the method further includes:

[0124] The preset smoke concentration, temperature rise rate, and environmental thermal radiation uniformity are combined with the smoke sensor signal response time, temperature sensor element aging degree, and human body sensor detection angle deviation to form six types of basic coded parameters.

[0125] Construct a pre-defined standard encoding table, including the correspondence between the original data and the encoded symbols; map the six types of basic encoding parameters to different row / column dimensions of the encoding table, so that each row of the encoding table corresponds to one type of parameter and each column corresponds to different value ranges of the parameter, forming a parameter-related initial encoding table;

[0126] A multi-parameter time-series stacked graph is generated with time as the horizontal axis and six types of basic coding parameters as the vertical axis. The stacked structure features of the graph are extracted. The number of columns and row order of the coding table are adjusted according to the stacked structure features to complete the dimensional structure variation of the coding table.

[0127] For each of the six basic coding parameters, a parameter trend curve is generated, and the curvature change characteristics of each parameter trend curve are extracted. Based on the curvature change characteristics, the coding symbols are mutated to obtain the final mutated coding table.

[0128] The real-time data generated during smoke detection is encoded using a variation coding table to generate a data coding string; the stacked structure features and curvature change features are used as coding verification information and embedded at the end of the data coding string.

[0129] In this embodiment, the first stage is the coding preparation phase, which focuses on screening and integrating two types of key parameters: one type is the core parameters for smoke detection (smoke concentration, temperature rise rate, and environmental thermal radiation uniformity), reflecting the safety status of the detection scenario; the other type is the sensor operating parameters (signal response time, component aging, and detection angle deviation), reflecting the sensor's reliability. The units and value ranges of these parameters will be standardized to form a unified set of basic coding parameters, providing a data foundation for the subsequent construction of the coding table.

[0130] Next, an initial coding framework is built. First, a standard coding table corresponding to the original data and coding symbols is constructed according to industry standards. Then, six types of basic parameters are bound to the coding table: each row corresponds to one type of parameter, and each column corresponds to a value range of the parameter, forming an initial coding table specifically for smoke detection scenarios.

[0131] Then, the coding table structure is adjusted. First, a multi-parameter time-series stacked plot is generated (time on the horizontal axis, parameters on the vertical axis, with parameter curves arranged in layers according to importance). Next, the stacking order of parameters in the plot (reflecting parameter importance) and the cross-frequency (reflecting parameter correlation) are extracted. Finally, the coding table is adjusted according to these characteristics: the number of columns corresponding to parameters with high cross-frequency increases (resulting in finer value ranges), and the row order is reordered according to parameter importance, completing the coding table structure variation. The above rules for adjusting the coding table structure can be customized and will not be elaborated here.

[0132] Next, the encoding characters in the encoding table are modified. A trend curve is generated for each type of parameter, and the curvature peaks (the moments of parameter abrupt change) and curvature frequencies (the frequency of parameter fluctuations) of the curves are extracted. The encoding symbols are then modified according to these characteristics: the symbol complexity at parameter abrupt changes is increased, and the symbol format for parameters with frequent fluctuations is adjusted, ultimately forming an encoding table with variations in both structure and content. The rules for modifying the encoding characters in the above encoding table can be customized and will not be elaborated here.

[0133] Finally, the real-time detection data is matched to the mutation coding table, and the corresponding coding symbols are extracted and concatenated into an initial coding string. Then, the previously extracted graphic features (layering order, cross frequency, curvature features) are converted into check codes and embedded at the end of the coding string to form a complete coding string. After transmission, the receiving end verifies the data integrity through the check code and then decodes and restores the detection data.

[0134] Reference Figure 2 In another embodiment of the present invention, a smoke detection device based on multi-sensor fusion analysis is also provided, comprising:

[0135] The detection unit is used to acquire the rate of temperature rise within the detection range detected by the temperature sensor and the uniformity of environmental thermal radiation within the detection range detected by the human body sensor when the smoke sensor detects that the smoke concentration within the detection range reaches a first threshold.

[0136] The correction unit is used to generate corresponding correction coefficients based on the temperature rise rate and the uniformity of ambient thermal radiation to correct the smoke concentration and the first threshold respectively, so as to obtain the corresponding corrected smoke concentration and correction threshold.

[0137] The calculation unit is used to calculate a comprehensive risk value based on the corrected smoke concentration and the rate of temperature rise if the corrected smoke concentration reaches the correction threshold.

[0138] The determination unit is used to determine a real fire if the comprehensive risk value reaches the risk threshold.

[0139] In this embodiment, the specific implementation of each unit in the above device embodiment is described in the above method embodiment, and will not be repeated here.

[0140] Reference Figure 3 This invention also provides a computer device, which can be a server, and its internal structure can be as follows: Figure 3 As shown, the computer device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores the data corresponding to this embodiment. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0141] Those skilled in the art will understand that Figure 3 The structures shown are merely block diagrams of some structures related to the present invention and do not constitute a limitation on the computer devices on which the present invention is applied.

[0142] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0143] In summary, the smoke detection method based on multi-sensor fusion analysis provided in this embodiment of the invention includes: when the smoke sensor detects that the smoke concentration within the detection range reaches a first threshold, acquiring the temperature rise rate within the detection range detected by the temperature sensor, and acquiring the environmental thermal radiation uniformity within the detection range detected by the human body sensor; based on the temperature rise rate and the environmental thermal radiation uniformity, generating corresponding correction coefficients to correct the smoke concentration and the first threshold respectively, obtaining corresponding corrected smoke concentration and corrected threshold; if the corrected smoke concentration reaches the corrected threshold, calculating a comprehensive risk value based on the corrected smoke concentration and the temperature rise rate; if the comprehensive risk value reaches the risk threshold, determining it as a real fire. In this invention, by combining the fusion processing of smoke sensors, temperature sensors, and human body sensors, the shortcomings of current smoke-based fire detection methods, which are easily affected by environmental interference and produce false alarms, are overcome.

[0144] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the present invention and embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0145] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0146] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A smoke detection method based on multi-sensor fusion analysis, characterized in that, Includes the following steps: When the smoke sensor detects that the smoke concentration within the detection range reaches a first threshold, the temperature sensor detects the rate of temperature rise within the detection range, and the human body sensor detects the uniformity of environmental thermal radiation within the detection range. Based on the temperature rise rate and the uniformity of environmental thermal radiation, corresponding correction coefficients are generated to correct the smoke concentration and the first threshold, respectively, to obtain the corresponding corrected smoke concentration and corrected threshold. If the corrected smoke concentration reaches a correction threshold, a comprehensive risk value is calculated based on the corrected smoke concentration and the rate of temperature rise. This includes: obtaining the smoke concentration rise start time T1 and the time T2 when the rate of temperature rise exceeds the speed threshold; marking the corresponding causal relationship and assigning a corresponding causal coefficient based on the time difference between T1 and T2; obtaining preset smoldering stage characteristic intervals and open flame stage characteristic intervals, with different corrected smoke concentrations and rates of temperature rise corresponding to different stage characteristic intervals; generating a corresponding compensation factor based on the fluctuation of the rate of temperature rise after matching the corrected smoke concentration and the rate of temperature rise to the corresponding stage characteristic interval; wherein the generation rules of the compensation factor are different in different stage characteristic intervals; determining the weights corresponding to the corrected smoke concentration and the rate of temperature rise based on the marked causal relationship and the matched stage characteristic interval; performing a fusion calculation on the corrected smoke concentration and the rate of temperature rise based on the corresponding weights to obtain a basic risk score; and multiplying the basic risk score by the causal coefficient and the compensation factor in sequence to obtain the comprehensive risk value. If the overall risk value reaches the risk threshold, it is determined to be a real fire.

2. The smoke detection method based on multi-sensor fusion analysis according to claim 1, characterized in that, After it is determined to be a real fire, the following applies: The risk level is determined based on the predicted heat source diffusion rate, the location of personnel located by the human body sensor, and the smoke concentration, and a differentiated alarm is output based on the risk level.

3. The smoke detection method based on multi-sensor fusion analysis according to claim 1, characterized in that, Based on the temperature rise rate and the uniformity of environmental thermal radiation, corresponding correction coefficients are generated to correct the smoke concentration and the first threshold, respectively, to obtain the corresponding corrected smoke concentration and corrected threshold, including: If the rate of temperature rise is less than the second threshold, a smoke concentration correction coefficient is generated to correct the smoke concentration, resulting in a corrected smoke concentration. If the uniformity of environmental thermal radiation exceeds the preset uniformity, a judgment threshold correction coefficient is generated to correct the first threshold, thus obtaining the corrected threshold.

4. The smoke detection method based on multi-sensor fusion analysis according to claim 1, characterized in that, Based on the labeled causal relationships and the matched stage feature intervals, the weights corresponding to the corrected smoke concentration and temperature rise rate are determined, including: Obtain a preset weight mapping relationship; wherein, the weight mapping relationship includes weight combinations corresponding to different causal relationships and stage feature intervals; Based on the causal relationship and the matched stage feature intervals, the weights corresponding to the corrected smoke concentration and temperature rise rate are determined through the weight mapping relationship.

5. The smoke detection method based on multi-sensor fusion analysis according to claim 1, characterized in that, Before performing the fusion calculation on the corrected smoke concentration and temperature rise rate based on the corresponding weights, the following steps are included: The corrected smoke concentration and temperature rise rate were normalized to the 0-1 range.

6. The smoke detection method based on multi-sensor fusion analysis according to claim 1, characterized in that, Acquiring the rate of temperature rise within the detection range detected by the temperature sensor includes: The temperature sensor collects temperature data from multiple different points within the detection range at a preset sampling frequency; Calculate the average temperature of each point at the same time, and calculate the rate of temperature rise based on the difference between the average temperature of adjacent times. If the fluctuation of the rate of temperature rise of multiple samples is less than the fluctuation threshold, the average of the rate of temperature rise of multiple samples is taken as the final rate of temperature rise. If the fluctuation of the temperature rise rate after multiple samplings reaches the fluctuation threshold, the preset sampling frequency of the temperature sensor is increased, and after multiple additional samplings, abnormal sampling values ​​are removed; the average value of the remaining temperature rise rate is used as the final temperature rise rate.

7. The smoke detection method based on multi-sensor fusion analysis according to claim 1, characterized in that, Acquiring the uniformity of environmental thermal radiation within the detection range detected by the human body sensor includes: The detection range is divided into a core area and an outer area by a preset radius; the core area is the core region in the detection range where the smoke concentration reaches the first threshold. Thermal radiation intensity data were collected from multiple sampling points in the core and outer areas using human body sensors. Calculate the standard deviation of thermal radiation intensity in the core area, the standard deviation of thermal radiation intensity in the outer area, and the difference between thermal radiation intensity in the core area and the outer area; Based on the standard deviation of the thermal radiation intensity in the core area, the standard deviation of the thermal radiation intensity in the outer area, and the difference in thermal radiation intensity between the core area and the outer area, the uniformity of environmental thermal radiation is calculated using a preset algorithm.

Images