Method for reducing false alarm rate of biological perimeter intrusion detection

Abstract

The application discloses a method for reducing the false alarm rate of biological induction perimeter alarm, and particularly relates to the field of infrared induction technology; background temperature data of a monitoring area is acquired, a background thermal field benchmark model is constructed, and a background benchmark temperature value is calculated; current background temperature is acquired according to a preset time interval, and a dynamic background temperature difference drift rate is calculated; heat source change information is extracted from a thermal imaging image sequence, the variance of the heat source change in a time window is calculated, and the stability of the static heat source signal is obtained; the above two features are normalized, and a feature parameter set is constructed; the set is input into a fuzzy logic reasoning model, a PIR sensor signal credibility parameter is output according to a preset fuzzy rule; according to the credibility result, an alarm judgment path is dynamically selected, intelligent fusion and judgment of the PIR signal and the video recognition result are realized; the method can effectively identify the PIR false triggering risk in a complex thermal environment, reduce the false alarm rate, and improve the reliability and adaptability of the perimeter security system.

Description

Technical Field

[0001] This invention relates to the field of infrared sensing technology, and more specifically to a method for reducing the false alarm rate of biosensor perimeter alarms. Background Technology

[0002] In existing technologies, false alarms are reduced by fusing information from multiple sensors (e.g., combining PIR with video) and employing logical rules (e.g., a "double confirmation mechanism"). However, in extreme environments, PIR sensors, highly dependent on the temperature difference between the environment and the target, often exhibit technical problems such as "no response" or "continuous response." When the background temperature is close to human body temperature, the sensor may fail to detect heat source movement, resulting in no alarm trigger. Conversely, in industrial sites with continuous heat dissipation, such as substations and oil refineries, the large-area heat radiation from high-temperature equipment may cause the PIR to misinterpret continuous human activity, leading to a continuous triggering state. Such anomalies prevent the system from correctly entering the double confirmation process. Even if the video system is operating normally, the alarm function is lost due to the abnormal PIR status, creating a long-term potential security blind spot that is extremely difficult to detect and poses a very high risk. Summary of the Invention

[0003] The purpose of this invention is to provide a method for reducing the false alarm rate of biosensor perimeter alarms, thereby addressing the shortcomings in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for reducing the false alarm rate of bio-sensor perimeter alarms, comprising:

[0005] Obtain background temperature data of the monitored area, and establish a background thermal field benchmark model based on the background temperature data to obtain the background benchmark temperature value T_bg_base;

[0006] The current background temperature T_bg_current is obtained according to a preset time interval Δt, and the dynamic background temperature difference drift rate ΔT_bg_dyn is calculated using the current background temperature T_bg_current and the background reference temperature value T_bg_base.

[0007] Acquire continuous image data within the monitored area, and extract heat source change information within a preset time window from the continuous image data;

[0008] Based on the extracted heat source change information, the variance of the heat source change within the preset time window is calculated to obtain the static heat source signal stability S_heat_sig.

[0009] The dynamic background temperature difference drift rate ΔT_bg_dyn and the static heat source signal stability S_heat_sig are normalized to form a set of characteristic parameters for unified calculation.

[0010] The set of characteristic parameters is input into a preset fuzzy logic reasoning model. Based on the preset fuzzy rules, the dynamic background temperature difference drift rate and the static heat source signal stability are jointly reasoned, and the PIR sensor signal reliability parameters are output.

[0011] Based on the reliability parameters of the output PIR sensor signal, the participation method of the PIR sensor output signal and the video recognition result is selected to generate the corresponding alarm judgment result.

[0012] Preferably, among which .

[0013] Preferably, the heat source change information includes spatial location, heat intensity, and contour area change information.

[0014] Preferably, the step of establishing a background thermal field benchmark model based on the background temperature data to obtain the background benchmark temperature value T_bg_base includes:

[0015] Acquire initial infrared temperature data from multiple spatial locations;

[0016] The initial infrared temperature data is spatially interpolated to construct a two-dimensional temperature distribution map covering the monitored area.

[0017] Based on the two-dimensional temperature distribution map, the temperature data of each sub-region within the monitoring area are comprehensively calculated using the regional weighted average method to obtain the background reference temperature value T_bg_base, which represents the overall environmental thermal state.

[0018] Preferably, the spatial interpolation process employs a bilinear interpolation algorithm to estimate the temperature value of each grid cell within the monitored area based on the temperature values ​​of adjacent sampling points and their spatial coordinate relationships.

[0019] Preferably, the step of calculating the variance of the heat source change within the preset time window to obtain the static heat source signal stability S_heat_sig includes:

[0020] Based on the spatial location change information, heat intensity change information, and contour area change information of the heat source obtained within the preset time window, corresponding time series data are constructed and denoised.

[0021] For each denoised time series, calculate the variance of its change within the preset time window to obtain the variance values ​​of position change, thermal intensity change, and contour area change.

[0022] According to the preset weighting coefficients, the variance values ​​of position change, thermal intensity change, and contour area change are weighted and fused to generate a comprehensive variance value.

[0023] The comprehensive variation variance value is used as the value of the static heat source signal stability S_heat_sig.

[0024] Preferably, the step of outputting the PIR sensor signal confidence parameter includes:

[0025] Based on the set of feature parameters, fuzzy membership functions are constructed for the normalized dynamic background temperature difference drift rate and the normalized static heat source signal stability, respectively. The dynamic background temperature difference drift rate is divided into three fuzzy sets: low change, medium change, and high change, and the static heat source signal stability is divided into three fuzzy sets: stable, semi-stable, and unstable.

[0026] Preferably, based on the constructed fuzzy set, a fuzzy rule set is established using a rule-table-based fuzzy inference method. The fuzzy rules are used to describe the correspondence between the reliability of passive infrared sensor signals under different thermal environment change states and stable heat source states. The feature parameter set is input into the fuzzy rule set, and fuzzy inference operation is performed to obtain the fuzzy inference result of the reliability of the passive infrared sensor signal. The fuzzy inference result is defuzzified using the centroid method to obtain the reliability parameter of the passive infrared sensor signal with continuous numerical characteristics.

[0027] Preferably, the step of selecting the participation mode of the PIR sensor output signal and the video recognition result includes: setting a first confidence threshold and a second confidence threshold, wherein the first confidence threshold is higher than the second confidence threshold, for dividing the confidence parameter of the passive infrared sensor signal into intervals.

[0028] Preferably, when the confidence parameter is higher than the first confidence threshold, the alarm judgment is made by using a dual confirmation logic of passive infrared sensor signal and video recognition result, requiring both to meet the intrusion conditions simultaneously before an alarm judgment result is generated.

[0029] When the confidence parameter is between the first confidence threshold and the second confidence threshold, a weighted fusion judgment method is adopted, in which the video recognition result is given a high weight, and the alarm judgment result is generated by combining the passive infrared sensor signal.

[0030] When the confidence parameter is lower than the second confidence threshold, the passive infrared sensor signal is ignored, and the alarm judgment result is generated only based on the video recognition result and its behavioral feature analysis result.

[0031] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0032] 1. The method for reducing the false alarm rate of bio-sensor perimeter alarms provided by this invention introduces two key environmental characteristic parameters: dynamic background temperature difference drift rate and static heat source signal stability. This accurately captures the dynamic fluctuation characteristics of the thermal environment and the interference characteristics of stable heat sources within the monitored area, effectively identifying the risk of false triggering of PIR sensors in complex thermal fields. By combining the above parameters with a fuzzy logic reasoning model for joint calculation, the reliability of the PIR signal is dynamically output, achieving intelligent control of the traditional "double confirmation" alarm logic. This solves the problem of false alarms caused by heat source drift, sudden environmental changes, and other factors in existing technologies.

[0033] 2. This invention further automatically selects an alarm path strategy based on the PIR signal reliability parameter, flexibly switching between three judgment methods: PIR-led, video-led, or dual confirmation. While ensuring intrusion detection sensitivity, it significantly reduces redundant triggering and false alarm probability, improving the system's stability and robustness under extreme scenarios such as high temperature, strong interference, and long-term operation. This method has good practicality and scalability, and is suitable for intelligent perimeter detection tasks in various security environments. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0035] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] For examples, please refer to Figure 1 As shown in this embodiment, the method for reducing the false alarm rate of bio-sensor perimeter alarms includes:

[0038] Obtain background temperature data of the monitored area, and establish a background thermal field benchmark model based on the background temperature data to obtain the background benchmark temperature value T_bg_base.

[0039] First, multiple infrared temperature monitoring points are deployed within the monitoring area. The number and layout of these points are determined based on the spatial dimensions and structural complexity of the monitored area. Preferably, the horizontal spacing between the points is set within the range of 1 to 3 meters to ensure uniform coverage and eliminate significant temperature blind spots.

[0040] Each location is sampled by an infrared thermal sensor at a fixed sampling frequency (e.g., once per second), and the temperature value is denoted as T1, T2, ..., Tn, where n represents the total number of locations. This set of temperature data constitutes the initial infrared temperature dataset for that moment, reflecting the instantaneous thermal state of each sub-region.

[0041] To construct a continuous temperature distribution covering the entire monitoring area from discrete initial infrared temperature data, spatial interpolation processing is required. The preferred interpolation method is bilinear interpolation, which estimates the temperature at the target location based on the temperature values ​​of four adjacent sampling points and their spatial coordinates.

[0042] Specifically, for the temperature point T(x,y) to be estimated, its value is calculated by weighting the temperature values ​​Ta, Tb, Tc, and Td of its four nearest sampling points and their relative position weighting coefficients. The interpolation process is performed in all grid cells within the monitoring area, ultimately forming a two-dimensional temperature distribution map with coordinates on the horizontal axis and temperature values ​​on the vertical axis, used to represent the thermal field structure of the current environment.

[0043] Based on the generated two-dimensional temperature distribution map, the monitoring area is divided into several sub-regions of equal area, each containing multiple interpolated temperature points. To comprehensively reflect the thermal state of the entire area, a region-weighted average method is used to fuse and calculate the temperature data of all sub-regions.

[0044] In the specific implementation, the average value of all temperature points within each sub-region is first calculated, denoted as T_zone1, T_zone2, ..., T_zonem, where m represents the number of sub-regions. Then, weighting coefficients W1, W2, ..., Wm are assigned to different sub-regions. The weighting can be set according to the importance of the region in the overall monitoring; for example, regions near boundaries or passageways are given higher weights. The final background baseline temperature value T_bg_base is calculated as follows: T_bg_base = (W1 × T_zone1 + W2 × T_zone2 + ... + Wm × T_zonem) ÷ (W1 + W2 + ... + Wm). This calculation ensures the comprehensiveness and specificity of the background temperature modeling, and the obtained T_bg_base is used as a reference value in the subsequent calculation of the dynamic background temperature difference drift rate.

[0045] The current background temperature T_bg_current is obtained according to a preset time interval Δt, and the dynamic background temperature difference drift rate ΔT_bg_dyn is calculated using the current background temperature T_bg_current and the background reference temperature value T_bg_base.

[0046] In the method for reducing the false alarm rate of biosensor perimeter alarms described in this invention, in order to monitor the dynamic changes in the background thermal environment of the monitored area in real time, it is necessary to periodically calculate the dynamic background temperature difference drift rate during operation, specifically including the following steps:

[0047] The preset temperature sampling time interval is denoted as Δt, in seconds, representing the time difference between two consecutive temperature samplings. The value of Δt is set according to the actual rate of change of ambient temperature and the required sampling accuracy, preferably ranging from 1 to 10 seconds. In this embodiment, Δt is set to 5 seconds.

[0048] At the end of each sampling period, the temperature data of the monitored area at the current moment is collected using an infrared temperature acquisition device. Using the same point layout and interpolation method as when establishing the background thermal field baseline model, a new two-dimensional temperature distribution map is obtained. Based on this distribution map, the representative temperature value of the current monitored area is calculated using the same regional weighted average method, denoted as the current background temperature T_bg_current.

[0049] The current background temperature T_bg_current represents the temperature expression value of the overall thermal environment state in the monitored area at the current time point, and has the same calculation logic and dimensions as the background reference temperature value.

[0050] The difference between the current background temperature T_bg_current and the pre-established background reference temperature T_bg_base is calculated to obtain the background temperature change. Then, this change is divided by the acquisition time interval Δt to calculate the dynamic background temperature drift rate, denoted as ΔT_bg_dyn. The calculation formula is: ΔT_bg_dyn = (T_bg_current - T_bg_base) ÷ Δt; where: the unit of ΔT_bg_dyn is degrees Celsius per second (°C / s); if the absolute value of ΔT_bg_dyn is large, it indicates that the environmental thermal field is undergoing drastic changes; if ΔT_bg_dyn is close to zero, it indicates that the environmental thermal field is relatively stable.

[0051] The calculated ΔT_bg_dyn is stored in a temperature difference record array, forming a temperature difference drift rate sequence over a period of time, which is used to subsequently determine the trend of thermal environment fluctuations. In the system logic, a threshold for temperature difference drift rate changes can be set, for example, 0.15 degrees Celsius per second. When ΔT_bg_dyn exceeds this threshold, the current thermal environment is considered to be in a state of drastic change, which may cause the passive infrared sensor to malfunction. This threshold can be preset according to different deployment scenarios or adaptively set through sample training.

[0052] Acquire continuous image data within the monitored area, and extract heat source change information within a preset time window from the continuous image data.

[0053] Thermal imaging cameras are deployed in the monitored area to continuously acquire images of the area. The image acquisition frequency is set according to the required analysis accuracy, preferably between 1 frame per second and 5 frames per second.

[0054] During operation, a fixed-length time window, denoted as T_window, is defined in seconds. This time window is used to cache and analyze the changes in the heat source's appearance across consecutive image frames. In this embodiment, T_window is set to 10 seconds. If the sampling frequency is 1 frame per second, 10 frames of image data can be obtained within one time window, forming a continuous image sequence.

[0055] In the continuous image sequence, a threshold-based heat source extraction algorithm is applied to each frame. This algorithm filters out potential heat source targets by setting a temperature threshold T_thresh, filtering areas in the image above this threshold. The temperature threshold can be adaptively adjusted based on the actual ambient temperature baseline or set manually, for example, by adding 3°C to the background temperature.

[0056] The extracted heat source regions are represented as outlines in each image frame, which facilitates subsequent statistical analysis of area, location, and heat intensity.

[0057] For the extracted heat source region, the following three change characteristic indicators are calculated in each frame of the image:

[0058] Spatial location information: By calculating the centroid coordinates (x, y) of the heat source region, its center position in the image coordinate system is obtained, and its motion trajectory in consecutive frames is tracked.

[0059] Thermal intensity information: The thermal intensity value of each heat source in the current frame is obtained by summing or averaging the pixel temperature values ​​within each heat source region. This value is denoted as Ht. This value reflects the overall temperature radiation capability of the heat source.

[0060] Outline area information: The total number of pixels in the heat source region is counted and converted into actual area units to obtain the projected area of ​​the heat source in the frame, denoted as At. The outline area reflects the size of the target or its thermal radiation coverage.

[0061] The above three heat source characteristic indicators are stored according to frame sequence number to form a heat source change time series data structure, as follows:

[0062] Centroid trajectory sequence: P={(x1,y1),(x2,y2),...,(xn,yn)};

[0063] Thermal intensity sequence: H={H1,H2,...,Hn};

[0064] Area sequence: A = {A1, A2, ..., An};

[0065] Where n represents the number of image frames within the time window. This data structure will serve as the input basis for subsequent static heat source signal stability calculation steps, supporting quantitative analysis of the dynamic fluctuation behavior of the heat source in the time dimension.

[0066] Based on the extracted heat source change information, the variance of the heat source change within the preset time window is calculated to obtain the static heat source signal stability S_heat_sig.

[0067] First, from the acquired continuous thermal imaging images, three main change features of the target heat source within a preset time window are extracted, including spatial location change information, thermal intensity change information, and contour area change information.

[0068] Spatial location change information: By calculating the centroid coordinates of the heat source in each frame of the image, a location time series P={(x1,y1),(x2,y2),...,(xn,yn)} is constructed;

[0069] Thermal intensity change information: Calculate the average pixel temperature value within the heat source area and construct a thermal intensity time series H={H1,H2,...,Hn};

[0070] Contour area change information: Count the number of pixels in the heat source area, convert them into actual area units, and construct the area time series A={A1,A2,...,An}.

[0071] After the time series data were constructed, they were denoised using a moving average filtering algorithm. The moving average filtering window length was set to 3 to 5 data points, preferably 3, to remove data fluctuations caused by image jitter or single-frame anomalies, thereby enhancing feature stability.

[0072] For the three time series after denoising, their statistical variances within the time window were calculated, yielding the following three variance values:

[0073] Variance of position change: The variance of the Euclidean distance from each centroid point in the time series P to the average centroid of the series is calculated, which reflects the spatial stability of the target within the time window;

[0074] Variance of heat intensity variation: The variance of the deviation between the heat intensity value of each frame in time series H and the average heat intensity value is calculated to reflect the temperature fluctuation of the heat source.

[0075] Variance of contour area change: The variance of the deviation between the area value of each frame in time series A and the average area value is calculated, which reflects the degree of change in the target shape or the change in occlusion.

[0076] The units and dimensions of each variance value should be kept consistent. The smaller the variance, the more stable the corresponding feature is over time.

[0077] To comprehensively reflect the overall stability of the heat source, the above three variance values ​​are weighted and integrated. The weighting coefficients are set as follows: W1: weight of spatial location variance; W2: weight of thermal intensity variance; W3: weight of contour area variance.

[0078] The three satisfy the relationship: W1 + W2 + W3 = 1. The weight values ​​are set according to the dynamic sensitivity of each feature in the actual scenario, and the preferred values ​​are: W1 = 0.4, W2 = 0.3, W3 = 0.3.

[0079] The weighted average algorithm is used for calculation: Comprehensive variation variance = (W1 × location variation variance) + (W2 × thermal intensity variation variance) + (W3 × area variation variance).

[0080] The above-mentioned comprehensive variation variance value is used as the value of the stability of the static heat source signal, denoted as S_heat_sig. The smaller the value of S_heat_sig, the smaller the change of the heat source within the time window, the higher the stability, and the more likely it is a continuous static heat source, such as a fixed heat dissipation device or a background heat object; when S_heat_sig approaches zero, it can be used as the input condition of "static heat source" in subsequent fuzzy logic reasoning.

[0081] The dynamic background temperature difference drift rate ΔT_bg_dyn and the static heat source signal stability S_heat_sig are normalized to form a set of characteristic parameters for unified calculation.

[0082] In the method for reducing the false alarm rate of biosensor perimeter alarms described in this invention, to facilitate the joint analysis of the dynamic background temperature difference drift rate and the static heat source signal stability by inputting them into a unified feature calculation model, both need to be normalized to ensure consistency in their numerical ranges and comparability in calculations. This process includes the following specific steps:

[0083] The target range for normalization is set to 0 to 1. The normalized parameter values ​​can be directly used as input variables for fuzzy logic inference systems or neural network models, adapting to various standardized computing frameworks.

[0084] Normalized upper and lower limits were set for the dynamic background temperature difference drift rate and the static heat source signal stability, respectively. These limits were derived from statistical analysis of a large amount of historical sampling data and experimental samples, as follows:

[0085] Normalization parameter settings for dynamic background temperature drift rate ΔT_bg_dyn:

[0086] The minimum value is set to 0 degrees Celsius per second, indicating a completely static environment;

[0087] The maximum value is set to 1 degree Celsius per second, which represents the upper limit of drastic fluctuations in the thermal environment;

[0088] Values ​​exceeding the maximum value are truncated to 1, and values ​​below the minimum value are truncated to 0.

[0089] Normalization parameter settings for static heat source signal stability S_heat_sig:

[0090] The minimum value is set to 0, indicating a completely stable and unchanging heat source;

[0091] The maximum value is set to 5 (a standardized measure of variance), representing a dynamic heat source with strong fluctuations.

[0092] The same truncation strategy is used to handle outliers outside the boundary.

[0093] The above upper and lower limits can be adjusted during training or preset according to different deployment environments.

[0094] Applying linear normalization to both parameters, the normalization formula is: Normalized value = (Original value - Minimum value) ÷ (Maximum value - Minimum value). The specific implementation is as follows:

[0095] Dynamic background temperature drift rate normalization: Subtract the set minimum temperature difference value of 0 from ΔT_bg_dyn, and then divide by the maximum temperature difference value of 1 to calculate the normalized dynamic thermal environment change index, which is denoted as: Normalized dynamic temperature drift rate = ΔT_bg_dyn ÷ 1.

[0096] Static heat source signal stability normalization processing: Subtract the set minimum variance value of 0 from S_heat_sig, and then divide by the maximum variance value of 5 to calculate the normalized heat source dynamic index, which is denoted as: Normalized heat source stability = S_heat_sig ÷ 5; The normalization result is retained to two decimal places and used in subsequent calculations in floating-point format.

[0097] The normalized dynamic background temperature difference drift rate and the normalized static heat source signal stability are sequentially constructed into a two-dimensional feature vector: feature parameter set = {normalized dynamic temperature difference drift rate, normalized heat source stability}; this feature parameter set will be used as an input variable in the subsequent fuzzy logic reasoning steps to comprehensively evaluate the reliability of the passive infrared sensor signal under the current thermal environment.

[0098] The resulting set of feature parameters is input into a preset fuzzy logic reasoning model. Based on preset fuzzy rules, the dynamic background temperature difference drift rate and the static heat source signal stability are jointly reasoned, and the PIR sensor signal reliability parameters are output.

[0099] In the method for reducing the false alarm rate of biosensor perimeter alarms described in this invention, in order to achieve dynamic judgment of the reliability of passive infrared sensor signals under complex thermal environment conditions, a fuzzy logic reasoning model needs to be constructed based on the normalized dynamic background temperature difference drift rate and the stability of the static heat source signal. This process includes the following steps:

[0100] Fuzzy membership functions are constructed for the normalized dynamic background temperature difference drift rate and the static heat source signal stability, respectively. The membership functions are defined in the form of triangle or trapezoidal functions to map continuous values ​​to the membership degree of the corresponding fuzzy language set.

[0101] Fuzzy set partitioning of dynamic background temperature difference drift rate: The parameter is divided into three fuzzy sets: "low change", "medium change" and "high change"; the numerical range is 0 to 1; the membership function of "low change" is defined as follows: the membership degree is 1 in the range of 0 to 0.3, and linearly decreases to 0 in the range of 0.3 to 0.5; "medium change" forms a symmetrical triangle in the range of 0.3 to 0.7; "high change" increases linearly between 0.5 and 1, and the membership degree is 1 after 0.7.

[0102] The fuzzy set partitioning of the static heat source signal stability is as follows: It is divided into three fuzzy sets: "stable," "semi-stable," and "unstable." The numerical range is 0 to 1. "Stable" membership functions with a degree of 0 to 0.3 represent high membership; "semi-stable" represents an intermediate transition zone; and "unstable" membership functions with a degree of 0.6 to 1 represent high membership. These functions are implemented using either a lookup table or function calculation and are pre-defined in the fuzzy inference module during deployment.

[0103] Based on the fuzzy sets of the two input variables mentioned above, a fuzzy rule set is established using a rule table approach. Each rule reflects the confidence judgment result of the sensor signal corresponding to a set of input states.

[0104] The fuzzy set of the output variable "passive infrared sensor signal confidence level" is defined as three levels: "low confidence level", "medium confidence level" and "high confidence level".

[0105] Example fuzzy rules include:

[0106] If the dynamic background temperature difference drift rate is "high change" and the heat source signal stability is "stable", then the confidence level is "low confidence level".

[0107] If the dynamic background temperature difference drift rate is "medium variation" and the heat source signal stability is "semi-stable", then the confidence level is "medium confidence level".

[0108] If the dynamic background temperature difference drift rate is "low change" and the heat source signal stability is "unstable", then the confidence level is "medium confidence".

[0109] If the dynamic background temperature difference drift rate is "low change" and the heat source signal stability is "stable", then the confidence level is "high confidence".

[0110] There are 3×3=9 rules, which fully cover the combination of input variables. The rules can be formulated by experts or automatically generated based on training samples.

[0111] Substitute the two normalized input values ​​from the feature parameter set into the constructed membership function, calculate its membership degree value for each fuzzy set, and use the Mamdani-type fuzzy inference method to synthesize the "minimum membership degree" according to all rules that meet the conditions, so as to obtain the activation degree of each output fuzzy set and form a comprehensive fuzzy output set.

[0112] This reasoning process can be implemented using a fuzzy reasoning engine, supporting real-time computation.

[0113] The fuzzy output set is defuzzified using the centroid method, which calculates the geometric centroid position of the output fuzzy set as the final confidence value. The specific method is as follows:

[0114] Construct a numerical function from each set of activated fuzzy outputs;

[0115] Calculate the area of ​​each set multiplied by its center position;

[0116] Use the weighted average of all results as the output value.

[0117] The final passive infrared sensor signal confidence parameter is a continuous value within the range of 0 to 1, used for subsequent alarm determination path selection. Higher confidence indicates a more reliable passive infrared sensor output under the current thermal environment conditions, while low confidence suggests the system may need to reduce the weight of the sensor signal.

[0118] Based on the reliability parameters of the output PIR sensor signal, the participation method of the PIR sensor output signal and the video recognition result is selected to generate the corresponding alarm judgment result.

[0119] In the method for reducing the false alarm rate of bio-sensor perimeter alarms described in this invention, in order to intelligently select the participation mode of passive infrared sensor signals and video recognition results in alarm decision-making based on the current thermal environment state, multi-level decision path switching is required based on the aforementioned passive infrared sensor signal confidence parameter. Specifically, the method includes the following steps:

[0120] Two consecutive confidence thresholds are set to classify the confidence parameters of passive infrared sensor signals into intervals: the first confidence threshold is set to 0.7 to distinguish high confidence states; the second confidence threshold is set to 0.4 to identify low confidence states. The confidence parameter ranges from 0 to 1 and is a continuous numerical variable. When the confidence parameter of the passive infrared sensor signal is higher than 0.7, it is defined as a "confidential state"; between 0.4 and 0.7, it is defined as a "medium confidence state"; and below 0.4, it is defined as an "untrustworthy state". These thresholds can be determined through training sample analysis or manually set according to the false alarm tolerance of the deployment scenario.

[0121] When the reliability parameter of the passive infrared sensor signal is higher than the first reliability threshold of 0.7, a standard dual-confirmation logic is used for alarm determination, specifically including:

[0122] Simultaneously, it monitors whether the passive infrared sensor outputs an intrusion trigger signal; checks whether the video recognition module identifies a target with human characteristics; and generates a valid alarm judgment result only when both intrusion conditions are met simultaneously; if either signal fails to meet the intrusion conditions, the judgment result is a non-alarm state. This logic is used to fully trust sensor signals, enhancing the rigor and accuracy of alarms.

[0123] When the reliability parameter of the passive infrared sensor signal is between 0.4 and 0.7, the system automatically enters the weighted fusion judgment mode, using the following processing method: The video recognition result is assigned a higher weight, preferably set to 0.7; the passive infrared sensor signal is assigned a weight of 0.3; the weighted score is calculated as follows: Alarm Comprehensive Score = 0.7 × Video Result Value + 0.3 × Passive Infrared Sensor Result Value; If the comprehensive score exceeds the fusion alarm threshold (e.g., 0.6), an alarm judgment result is generated; otherwise, no alarm is triggered. This logic is suitable for environments where sensor signals are unstable but have some reference value, balancing the risks of false alarms and missed alarms.

[0124] When the reliability parameter of the passive infrared sensor signal is lower than the second reliability threshold of 0.4, the sensor signal is considered unreliable, and the processing method is as follows:

[0125] The current trigger signal from the passive infrared sensor is ignored; only the video recognition result and its associated behavioral feature analysis result are analyzed. Behavioral features include the target's movement trajectory, duration, and proximity to the area boundary. If the video recognition system independently determines that intrusion behavior exists, an alarm judgment result is generated; otherwise, no alarm is triggered. This strategy is suitable for situations where the sensor is severely interfered with by the environment, such as strong heat sources or equipment failure, ensuring that the system still possesses a minimum level of detection capability.

[0126] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for reducing the false alarm rate of bio-sensor perimeter alarms, characterized in that: include: Obtain background temperature data of the monitored area, and establish a background thermal field benchmark model based on the background temperature data to obtain the background benchmark temperature value T_bg_base; The current background temperature T_bg_current is obtained according to a preset time interval Δt, and the dynamic background temperature difference drift rate ΔT_bg_dyn is calculated using the current background temperature T_bg_current and the background reference temperature value T_bg_base. Acquire continuous image data within the monitored area, and extract heat source change information within a preset time window from the continuous image data; Based on the extracted heat source change information, the variance of the heat source change within the preset time window is calculated to obtain the static heat source signal stability S_heat_sig. The dynamic background temperature difference drift rate ΔT_bg_dyn and the static heat source signal stability S_heat_sig are normalized to form a set of characteristic parameters for unified calculation. The set of characteristic parameters is input into a preset fuzzy logic reasoning model. Based on the preset fuzzy rules, the dynamic background temperature difference drift rate and the static heat source signal stability are jointly reasoned, and the PIR sensor signal reliability parameters are output. Based on the reliability parameters of the output PIR sensor signal, the participation method of the PIR sensor output signal and the video recognition result is selected to generate the corresponding alarm judgment result.

2. The method for reducing the false alarm rate of bio-sensor perimeter alarms according to claim 1, characterized in that: in .

3. The method for reducing the false alarm rate of bio-sensor perimeter alarms according to claim 1, characterized in that: The heat source change information includes information on spatial location, heat intensity, and changes in outline area.

4. The method for reducing the false alarm rate of bio-sensor perimeter alarms according to claim 1, characterized in that: The steps for establishing a background thermal field baseline model based on the background temperature data and obtaining the background baseline temperature value T_bg_base include: Acquire initial infrared temperature data from multiple spatial locations; The initial infrared temperature data is spatially interpolated to construct a two-dimensional temperature distribution map covering the monitored area. Based on the two-dimensional temperature distribution map, the temperature data of each sub-region within the monitoring area are comprehensively calculated using the regional weighted average method to obtain the background reference temperature value T_bg_base, which represents the overall environmental thermal state.

5. The method for reducing the false alarm rate of bio-sensor perimeter alarms according to claim 4, characterized in that: The spatial interpolation process employs a bilinear interpolation algorithm to estimate the temperature value of each grid cell within the monitored area based on the temperature values ​​of adjacent sampling points and their spatial coordinate relationships.

6. The method for reducing the false alarm rate of bio-sensor perimeter alarms according to claim 1, characterized in that: The step of calculating the variance of the heat source change within the preset time window to obtain the static heat source signal stability S_heat_sig includes: Based on the spatial location change information, heat intensity change information, and contour area change information of the heat source obtained within the preset time window, corresponding time series data are constructed and denoised. For each denoised time series, calculate the variance of its change within the preset time window to obtain the variance values ​​of position change, thermal intensity change, and contour area change. According to the preset weighting coefficients, the variance values ​​of position change, thermal intensity change, and contour area change are weighted and fused to generate a comprehensive variance value. The comprehensive variation variance value is used as the value of the static heat source signal stability S_heat_sig.

7. The method for reducing the false alarm rate of bio-sensor perimeter alarms according to claim 1, characterized in that: The step of outputting the PIR sensor signal reliability parameter includes: Based on the set of feature parameters, fuzzy membership functions are constructed for the normalized dynamic background temperature difference drift rate and the normalized static heat source signal stability, respectively. The dynamic background temperature difference drift rate is divided into three fuzzy sets: low change, medium change, and high change, and the static heat source signal stability is divided into three fuzzy sets: stable, semi-stable, and unstable.

8. The method for reducing the false alarm rate of bio-sensor perimeter alarms according to claim 7, characterized in that: Based on the constructed fuzzy set, a fuzzy rule set is established using a rule table-based fuzzy reasoning method. The fuzzy rules are used to describe the correspondence between the reliability of passive infrared sensor signals under different thermal environment change states and stable heat source states. The set of feature parameters is input into the set of fuzzy rules, and fuzzy inference is performed to obtain the fuzzy inference result of the reliability of the passive infrared sensor signal. The fuzzy inference result is then defuzzified using the centroid method to obtain the reliability parameter of the passive infrared sensor signal with continuous numerical characteristics.

9. The method for reducing the false alarm rate of bio-sensor perimeter alarms according to claim 1, characterized in that: The steps for selecting the participation mode of PIR sensor output signal and video recognition result include: setting a first confidence threshold and a second confidence threshold, wherein the first confidence threshold is higher than the second confidence threshold, and is used to divide the confidence parameter of passive infrared sensor signal into intervals.

10. The method for reducing the false alarm rate of bio-sensor perimeter alarms according to claim 9, characterized in that: When the confidence parameter is higher than the first confidence threshold, the alarm judgment is made by using the dual confirmation logic of passive infrared sensor signal and video recognition result. Both are required to meet the intrusion conditions at the same time before an alarm judgment result is generated. When the confidence parameter is between the first confidence threshold and the second confidence threshold, a weighted fusion judgment method is adopted, in which the video recognition result is given a high weight, and the alarm judgment result is generated by combining the passive infrared sensor signal. When the confidence parameter is lower than the second confidence threshold, the passive infrared sensor signal is ignored, and the alarm judgment result is generated only based on the video recognition result and its behavioral feature analysis result.

Images