TWI647664B - Data fusion based safety surveillance system and method - Google Patents

Abstract

A data fusion-based security monitoring system includes at least one physical monitoring system based on different algorithms to form first to Nth virtual monitoring systems and data fusion and decision making devices. The data fusion and decision device defines an ith detection model of the i-th virtual monitoring system, and according to the ith virtual monitoring system, multiple ith monitoring data under the condition of corresponding multiple locations, multiple environmental context data, and intruders The number of detected missing samples, the i-th detection model, multiple i-th monitoring data of multiple locations of the i-th virtual monitoring system, and multiple context data to estimate the i-th virtual monitoring system Detect misses. The data fusion and decision making device determines the fusion parameter group according to the first to the Nth detection loss probability, and performs data fusion on the first to Nth detection results of the first to Nth virtual monitoring systems according to the fusion parameter group to generate a decision. result. In this way, the overall detection and loss probability can be reduced through data fusion without increasing the cost of construction and adding an accurate physical monitoring system.

Description

Security monitoring system and method based on data fusion

The present invention relates to a security monitoring technology, and more particularly to a data fusion-based security that can combine data detection results of multiple virtual monitoring systems formed by one or more physical monitoring systems to perform different algorithms to perform judgment and decision making. Monitoring systems and methods.

In recent years, safety monitoring technology has been highly valued by industry and government agencies, especially in the monitoring of buildings or facilities with high-tech, military or administrative secrets. However, the physical monitoring system itself has individual physical limitations, and may affect the detection results of the physical monitoring system itself due to factors such as the environment (including weather), resulting in detection loss (Note: there are intruders, but No intruder detected or false alarm (Note: no intruders, but detected intruders).

In general, the loss caused by detecting a leak may be much greater than the loss of a false alarm. Therefore, the traditional approach is to increase the sensitivity of the physical monitoring system to avoid detecting misses. However, this practice causes the monitoring personnel to frequently receive the alarm message of the false alarm, which may cause the monitoring personnel to numb the alarm message, thereby ignoring the alarm message of the correct alarm. In addition, the establishment of a more accurate physical monitoring system and the replacement of the existing physical monitoring system can solve the above technical problems, but this approach leads to additional cost increases.

Therefore, in order to overcome the deficiencies of the prior art, the embodiments of the present invention provide a security monitoring system and method for integrating multiple detection results of multiple virtual monitoring systems formed by different physical algorithms of an existing entity monitoring system. The security monitoring system and method can greatly reduce the probability of detecting and error warning of the entire security monitoring system without replacing the existing physical monitoring system.

Based on at least one of the foregoing objectives, an embodiment of the present invention provides a data fusion-based security monitoring system, including first to Nth virtual monitoring systems and links formed by performing at least one entity monitoring system to perform different algorithms. Data fusion and decision making device to the Nth virtual monitoring system, where N is greater than or equal to 2. The data fusion and decision device defines first to Nth detection models of the first to Nth virtual monitoring systems, wherein the i-th detection model is used to represent the plurality of i-th monitoring data of the i-th virtual monitoring system and the corresponding multi-number i The relationship between the plurality of i-th detection results of the monitoring data, where i is an integer from 1 to N. The data fusion and decision-making apparatus according to the i-th virtual monitoring system, the plurality of detection data of the plurality of ith monitoring data corresponding to the plurality of locations, the plurality of environmental context data, and the intruder The plurality of i-th monitoring data and the plurality of environmental context data of the plurality of locations of the i-th virtual monitoring system estimate an ith detection and leakage probability of the i-th virtual monitoring system. The data fusion and decision making device determines the fusion parameter group according to the first to Nth detection loss probability factors, and the data fusion and decision making device performs the first to Nth detection results of the first to Nth virtual monitoring systems according to the fusion parameter group. Convergence to produce decision outcomes.

Based on at least one of the foregoing objectives, an embodiment of the present invention further provides a security monitoring method based on data fusion. First, the definition is performed by at least one entity monitoring system executing different algorithms The first to the Nth detection models of the first to the Nth virtual monitoring systems, wherein the i-th detection model is used to represent the plurality of i-th monitoring data of the i-th virtual monitoring system and the corresponding plurality of i-th monitoring data The relationship between the plurality of ith detection results, where i is an integer from 1 to N, and N is greater than or equal to two. According to the i-th virtual monitoring system, the plurality of detection data of the plurality of i-th monitoring data corresponding to the plurality of locations, the plurality of environmental context data, and the intruder are the number of the missing detection samples, the i-th detection model, and the i-th virtual monitoring The plurality of i-th monitoring data and the plurality of environmental context data in multiple locations of the system estimate the ith detection and leakage probability of the i-th virtual monitoring system. The fusion parameter set is determined according to the first to the Nth detection loss probability. Then, the first to Nth detection results of the first to Nth virtual monitoring systems are data fusion according to the fusion parameter group to generate a decision result.

Optionally, in the embodiment of the present invention, multiple pieces of the i-th monitoring data of the i-th virtual monitoring system under the assumption of multiple locations, multiple environmental context data, and no intruders are further selected. The number of erroneous alarm samples, the i-th detection model, multiple ith monitoring data of multiple locations of the i-th virtual monitoring system, and multiple environmental context data estimates the ith error warning probability of the i-th virtual monitoring system, and the fusion parameters The group is determined according to the first to Nth detection loss probability and the first to Nth false alarm probability.

Optionally, in the embodiment of the present invention, the plurality of known monitoring data based on the acquired i-th virtual monitoring system and the plurality of known detection results corresponding to the plurality of known monitoring data are based on a machine learning algorithm, The artificial intelligence algorithm or other algorithm for solving the model defines the i-th detection model

Optionally, in the embodiment of the present invention, the data fusion is performed by using a logical operation function, a reliability rule, or an environment context rule, and the fusion parameter group is used to determine a logical operation function, a reliability rule, and an environment. Thread rules.

Optionally, in the embodiment of the present invention, each of the plurality of environmental context data is weather data defining variables such as rainfall, wind, temperature, and brightness.

In short, the data fusion-based security monitoring system and method provided by the embodiments of the present invention define multiple detection models of multiple virtual monitoring systems formed by different one or more entity monitoring systems to execute different algorithms. Then, the data fusion-based security monitoring system estimates the detection loss based on the multiple monitoring data, the detection model, the multiple environmental context data of each virtual monitoring system, and the average number of multiple monitoring data under different conditions. Error alert probability. Then, the data fusion-based security monitoring system determines a fusion parameter set for data fusion of multiple detection results according to the detection loss and the false alarm probability. As a result, the data fusion-based security monitoring system and method provides a low-cost technical solution that does not require the addition of a physical monitoring system, and which can reduce the overall detection loss probability and false alarm probability.

1‧‧‧Security monitoring system based on data fusion

111‧‧‧First Virtual Surveillance System

112‧‧‧Second virtual monitoring system

11N‧‧‧Nth Virtual Monitoring System

12.3‧‧‧Information fusion and decision-making devices

S201~S205‧‧‧Steps

31‧‧‧Detection model definition unit

32‧‧‧ probability estimation unit

33‧‧‧Fused parameter setting unit

34‧‧‧Decision unit

1 is a block diagram of a data fusion based security monitoring system in accordance with an embodiment of the present invention.

2 is a flow chart of a data fusion based security monitoring method according to an embodiment of the present invention.

3 is a block diagram of a data fusion and decision apparatus according to an embodiment of the present invention.

In order to fully understand the objects, features and advantages of the present invention, the present invention will be described in detail by the accompanying drawings.

The embodiment of the invention provides a security monitoring system and method based on data fusion, which does not need to add an additional entity monitoring system, but will be implemented by one or more existing entity monitoring systems. The detection results of multiple virtual monitoring systems formed by different algorithms are combined to further generate decision results.

In an embodiment, for each virtual monitoring system, the security monitoring system and method are first trained to define a detection model for each virtual monitoring system. The training method is to collect a plurality of known monitoring data of each virtual monitoring system and a plurality of known detection results corresponding to a plurality of known monitoring data, and then use machine learning algorithms, artificial intelligence algorithms or other The algorithm for solving the model obtains a detection model between multiple monitoring data of each virtual monitoring system and multiple detection results. Then, for each virtual monitoring system, the security monitoring system and method detects the number of missing samples, the environmental latitude of multiple locations, and multiple locations according to multiple monitoring data under multiple different conditions of the intruder. Corresponding multiple monitoring data and detection models estimate the probability of detecting leakage, and the number of multiple false alarm samples according to multiple monitoring data under different conditions of the intruder, and the environmental pulse probability of multiple locations Multiple monitoring data and detection models corresponding to multiple locations estimate the probability of false alarms.

After the training is completed, an inference is then made to produce a decision result. Further, the security monitoring system and method set a fusion parameter group according to the estimated detection failure probability and the false alarm probability of the plurality of virtual monitoring systems, wherein the fusion parameter group is used to determine a fusion manner of the multiple detection results. After that, the security monitoring system and method can generate the decision result according to the detection result of the fusion parameter group and the multiple virtual monitoring systems (that is, the data fusion of the multiple detection results according to the fusion parameter group to generate the decision result) ).

Please note that the loss due to detection loss may be much greater than the loss of false alarms. Therefore, in one embodiment of the present invention, the probability of false alarms may not be estimated, and the security supervision The control system and method only set the fusion parameter group according to the estimated detection and loss probability of multiple virtual monitoring systems.

Optionally, in one embodiment of the present invention, the data fusion manner of the multiple detection results may be performed by using a logical operation function, and the fusion parameter group may be used to determine detection results of multiple virtual monitoring systems. Logical operation function. For example, some detection results are logically ANDed, and another part of the detection results are logically ORed, and then multiple logical operations are logically ORed or ORed to generate decision results. .

Optionally, in one embodiment of the present invention, the data fusion manner of the multiple detection results may be performed by using a reliability rule, and the fusion parameter group may be used to determine a reliability rule. For example, if the final decision result needs to be greater than the reliability threshold, the decision result may be output, or the reliability of the detection result must be greater than the reliability threshold, so that the detection result can be used for data fusion.

Optionally, in an embodiment of the present invention, the data fusion manner of the multiple detection results may be performed by using an environment context rule, and the fusion parameter group may be used to determine an environment context rule. . For example, each detection result corresponds to environmental context information, such as coordinates, object type, number of objects, and accuracy rate, and multiple detection results are based on environmental context information for data fusion according to environmental context rules.

First, please refer to FIG. 1 of the present invention. FIG. 1 is a block diagram of a data fusion-based security monitoring system according to an embodiment of the present invention. The data fusion-based security monitoring system 1 includes a first virtual monitoring system 111, a second virtual monitoring system 112, ..., an Nth virtual monitoring system 11N, and a data fusion and decision making device. 12, wherein the first virtual monitoring system 111, the second virtual monitoring system 112, ..., the Nth virtual monitoring system 11N connects the data fusion and decision making device 12 by wire or wirelessly.

Please note here that, in fact, each of the one or more entity monitoring systems has different algorithms to form the first to Nth virtual monitoring systems 111~11N, that is, the first to Nth virtual monitoring systems. Each of 111~11N is generated by a physical monitoring system based on one of the algorithms. The monitoring system of the foregoing entity is, for example, a thermal image system, an electronic fence system, a laser ranging system or a virtual wall system, but the invention is not limited thereto. The foregoing variable N is a positive integer. For example, there are two entity monitoring systems, and one algorithm and two algorithms are respectively executed, so N can be 3. However, the present invention is not limited thereto.

The data fusion and decision making device 12 includes a plurality of circuits and is configured to cause the data fusion and decision device 12 to have computing power. For example, the data fusion and decision making device 12 is a general computer that is equipped with a corresponding software program to perform related operations to perform the steps required to perform the data fusion and decision device 12 and the desired functions.

In general, thermal imaging systems are susceptible to temperature. Electronic fence systems or laser ranging systems are susceptible to moisture interference because they detect the characteristics of the signal band. In addition, the virtual wall system is provided by the visible light camera as the basis for identification, so if the image quality is not good or affected by the innate conditions (for example, no light source at night, excessive wind speed, leaf sway or insufficient image pixels (resolution), etc. ), the final detection results will be directly or indirectly affected. Therefore, it can be known that the environmental context data is an important factor affecting the detection results of the above virtual monitoring systems. From the above example, the environmental context data may be, for example, weather data including variables such as humidity, wind speed, temperature, and brightness, but the present invention does not limit the environmental context data to weather data.

Since the environment will have an impact on the detection results of the virtual monitoring system, the environmental context factor must still be taken into account when estimating the probability of detecting the missed rate and the probability of the false alarm. Taking environmental context data as weathering data as an example, humidity can be expressed as a binary value of "rainy weather" or "no rain", or simply by rainfall; wind speed can use "no wind", "breezes" or "strong wind" The ternary value is expressed simply by the wind; the temperature can be expressed as a binary value of "low temperature" or "high temperature", or simply by temperature value; and the brightness can be binary of "day" or "night" Value is expressed, or simply expressed as a brightness value.

The data fusion and decision making device 12 first collects a plurality of known first to Nth monitoring data and multiple pens of the first virtual monitoring system 111, the second virtual monitoring system 112, ..., the Nth virtual monitoring system 11N. Know the first to Nth detection results. The plurality of known first monitoring data, the second monitoring data, and the Nth monitoring data are detection errors of the first virtual monitoring system 111, the second virtual monitoring system 112, ..., and the Nth virtual monitoring system 11N, respectively. Missing and false alarm source data, in order to reduce the number of multiple detected missing samples and multiple false alarm samples for multiple monitoring data in multiple different conditions of intruder and non-intruder Calculation. However, the present invention is not limited thereto, and the plurality of known first monitoring data, second monitoring data, and Nth monitoring data may also be the first virtual monitoring system 111, the second virtual monitoring system 112, respectively. The Nth virtual monitoring system 11N does not detect the original data of error and non-error alarms, but this method requires additional recalculation of multiple monitoring under different conditions of intruders and non-intrusives. The number of missing samples of the data and the number of multiple false alarm samples.

Then, the data fusion and decision making device 12 can transmit through the plurality of known i-th monitoring data of the i-th virtual monitoring system and the plurality of known i-th detecting results corresponding to the plurality of known i-th monitoring data. A machine learning algorithm, an artificial intelligence algorithm, or other algorithm for solving a model (that is, the present invention does not limit behavior such as solving a model) to define the relationship between the i-th monitoring data and the ith detection result. Test model, where i is equal to an integer from 1 to N. After that, the data fusion and decision-making apparatus 12 detects, according to the plurality of positions, multiple pieces of the i-th monitoring data of the i-th virtual monitoring system corresponding to the plurality of locations, the plurality of environmental context data, and the intruder. Environmental context data (to obtain the environmental context probability of multiple locations), multiple ith monitoring data corresponding to multiple locations, and detection model of the ith virtual monitoring system to estimate the ith detection loss of the ith virtual monitoring system The probability of the plurality of erroneous alarm samples of the plurality of ith monitoring data of the i-th virtual monitoring system under the condition of corresponding multiple locations, multiple environmental context data, and no intruders, Multiple environmental context data (to obtain the environmental context probability of multiple locations), multiple ith monitoring data corresponding to multiple locations, and detection model of the ith virtual monitoring system to estimate the ith error of the ith virtual monitoring system Alarm probability.

Thereafter, the data fusion and decision making device 12 detects the first to Nth detection failure rate and the first to Nth errors of the first to Nth virtual monitoring systems 111 to 11N according to the first to Nth virtual monitoring systems 111 to 11N. The alarm probability setting fusion parameter group is used to determine the fusion manner of the first to Nth detection results of the first to Nth virtual monitoring systems 111~11N. Then, the data fusion and decision making device 12 generates a decision result according to the first to Nth detection results of the fusion parameter group and the first to Nth virtual monitoring systems 111-11N to determine whether an intruder breaks in.

For example, the data fusion manner of the first to the Nth detection results of the first to Nth monitoring systems 111 to 11N may be performed by using a logical operation function, and the fusion parameter group may be used to determine the first to Logical operation function of the first to Nth detection results of the Nth monitoring system 111~11N number. For example, if N is 3, the first and second detection results are logically ANDed, and the obtained operation result is logically ORed with the third detection result, thereby obtaining a decision result.

Next, the details of estimating the ith detection loss probability and the ith error warning probability of the i-th virtual monitoring system are further explained as follows. The ith detection and loss probability of the i-th virtual monitoring system can be expressed as M _i,l =Σ _k ʃʃ(1- R _i,l ( y ^i,k )) p ( y ^i,k ,W _m ,L _k | H ₁ ) dy ^i,k dW _m (referred to as Formula 1 for short), wherein the variables i, k and l respectively represent the virtual monitoring system index value, the position index value and the parameter group index value, and the variable L _k represents the position index value k corresponding to The position, the variable y ^i,k represents the i-th monitoring data acquired by the i-th virtual monitoring system at the position L _k , and the function R _i,l ( y ^i,k ) indicates that the i-th virtual monitoring system adopts the first parameter group based on the i monitors the ith detection result of the data y ^i,k , the function R _i,l ( y ^i,k ) can be 1 or 0 to indicate whether an intruder is detected, and the variable H ₁ indicates a condition in which an intruder invades The variable W _m represents a vector of environmental context data (which may, for example, be defined as weather data including four variables such as humidity, wind speed, temperature, and brightness), and the parameter group index value l is determined by the position L _k and The environmental context data W _m is determined, and the function p ( y ^i,k , W _m , L _k | H ₁ ) indicates that under the condition of intruder intrusion, the i-th virtual monitoring system is at the position L _k and the environmental context data is W _m and the i-th monitoring data are the joint probability of y ^i,k .

Since both the thermal image system and the virtual wall system may affect the detection results due to the sway of the surrounding trees, the electronic fence system is sensitive to groundwater or puddles and is easily affected, and the above-mentioned entity monitoring systems perform different algorithms. The resulting virtual surveillance systems are also sensitive to the terrain of the hillside. Therefore, the function p ( y ^i,k , W _m , L _k | H ₁ ) in the above formula ₁ can be simplified to be p ( y ^i,k , W _m , L _k | H ₁ )= p ( y ^{i, k} | H ₁ , W _m , L _k ) p ( W _m , L _k ) (referred to as formula 2), where the function p ( y ^i,k | H ₁ , W _m , L _k ) is at position L _k , environment Under the condition that the network data W _m and there are intruders, the i-th monitoring data of the i-th virtual monitoring is the probability of y ^i,k , and the function p ( W _m , L _k ) represents the position L _k and the environmental context data W _m Probability (that is, the environmental systolic probability of position L _k ).

Since the variable p ( W _m , L _k ) is independent of the condition of H ₁ , after formula 2 is substituted into formula 1, formula 1 can be rewritten as (referred to as formula 3), among them The expected value of the i-th monitoring data y ^i,k of the i-th virtual monitoring system under the condition of the location L _k , the environmental context data W _m and the intruder.

Then, using the statistical concept that the number of samples approximates the expected value, Equation 3 can be rewritten as (referred to as Formula 4), where The number of detected missing samples of the i-th monitoring data y ^i,k of the i-th virtual monitoring system under the condition of the location L _k , the environmental context data W _m and the intruder.

Since the i-th detection result R _i,l ( y ^i,k ) cannot be obtained from the i-th monitoring data y ^i,k of the i-th virtual monitoring system, the machine learning algorithm, the artificial intelligence algorithm or other solution is obtained. The algorithm of the model can define the i-th monitoring data y ^i,k and the i-th based on the known i-th monitoring data y ^i,k and the known corresponding detection result R _i,l ( y ^i,k ) measurement results ^{_{R i, l (y i,}} k) of the detection model, i.e., can know the function ^{_{R i, l (y i,}} k) of the model. Then, through Equation 4, based on the detection model of the i-th virtual monitoring system, according to the environmental context data W _m (the environmental probability of the position L _k can be statistically learned), the position L _k , the position L _k , the environmental context data The number of detected missing samples of the i-th monitoring data y ^i,k of the i-th virtual monitoring system under the condition of W _m and intruder And the i-th monitoring data y ^{i,k of} each position L _k calculates the ith detection leakage probability of the i-th virtual monitoring system.

Similarly, the ith error alarm probability of the ith virtual monitoring system can be expressed as F _i,l =Σ _k ʃʃ R _i,l ( y ^i,k ) p ( y ^i,k ,W _m ,L _k | H ₀ ) dy ^i,k dW _m (abbreviated as Equation 5), where the function p ( y ^i,k , W _m , L _k | H ₀ ) indicates that the i-th virtual monitoring system is in position without intruder intrusion L _k , the environmental context data W _m and the i-th monitoring data are the joint probability of y ^i,k , and the variable H ₀ indicates the condition that no intruder invades. Based on a similar derivation, Equation 5 can therefore be reduced to M _i,l =Σ _k ʃ _Ri,l ( y ^i,k ) p ( W _m ,L _k ) dW _m (referred to as formula 6), wherein The average number of the i-th monitoring data of the i-th virtual monitoring system under the condition of the location L _k , the environmental context data W _m and no intruders.

Then, through Equation 6, the detection model based on the ith virtual monitoring system can be used, according to the environmental context data W _m (the environmental probability of the location L _k can be statistically learned), the position L _k , the position L _k , the environmental context data Number of false alarm samples of the i-th monitoring data y ^i,k of the i-th virtual monitoring system under the condition of W _m and no intruder And the i-th monitoring data y ^{i,k of} each position L _k calculates the false alarm probability of the i-th virtual monitoring system. After obtaining the first to Nth detection leakage probability and the first to Nth false alarm probability of the first to Nth virtual monitoring systems 111 to 11N, the first to Nth detection errors can be calculated based on the demand. The probability and the first to Nth false alarm probability determine the fusion parameter set, and then merge the first to Nth detection results of the first to Nth virtual monitoring systems 111~11N based on the fusion parameter group to generate a decision result. . In this way, the detection result of each virtual monitoring system formed by different algorithms can be generated by at least one entity monitoring system without changing the existing entity monitoring system to generate more accurate decision results, thereby reducing data fusion. The overall security detection system 1 detects the loss and the probability of false alarms.

2 and FIG. 3, FIG. 2 is a flowchart of a data fusion-based security monitoring method according to an embodiment of the present invention, and FIG. 3 is a block diagram of a data fusion and decision device according to an embodiment of the present invention. The data fusion based security monitoring method of FIG. 2 can be implemented by the data fusion and decision device 12 of FIG. 1, and the data fusion and decision device 12 can be implemented by the data fusion and decision device 3 of FIG.

The data fusion and decision device 3 of FIG. 3 is configured by a plurality of hardware circuits (or a hardware circuit), and includes a detection model definition unit 31, a probability estimation unit 32, a fusion parameter setting unit 33, and a decision unit. 34, wherein the detection model definition unit 31 is connected to the probability estimation unit 32, the probability estimation unit 32 is connected to the fusion parameter setting unit 33, and the fusion parameter setting unit 33 is connected to the decision unit 34.

First, in step S201, the detection model definition unit 31 defines a plurality of known first to Nth monitoring data of the first to Nth virtual monitoring systems and a plurality of corresponding first to Nth monitoring materials. Know the first to the Nth detection results. Then, in step S202, the detection model definition unit 31 defines the first to the Nth monitoring system for solving the model's algorithm (for example, through a machine learning algorithm, an artificial intelligence algorithm, or other algorithms for solving the model). The detection model, wherein the detection model of the i-th virtual monitoring system is obtained based on a plurality of known i-th monitoring data and corresponding plurality of known i-th detection results.

Then, in step S203, the probability estimating unit 32 responds to the plurality of environmental context data (the environmental probability of each location can be known through statistics), the plurality of i-th monitoring data corresponding to the plurality of locations, and the i-th virtual monitoring system. Position, multiple environmental context data, and multiple detection samples of the i-th monitoring data under the condition of the intruder and the detection model of the i-th virtual monitoring system to estimate the i-th virtual monitoring system Detecting the probability of missed, and according to multiple environmental context data (which can be used to know the environmental context probability of each location through statistics), multiple ith monitoring data corresponding to multiple locations, and the i-th virtual monitoring system corresponding to multiple locations and multiple pens Environmental context data and more than the number of i-monitoring data under the conditions of intruders The number of error alarm samples and the detection model of the i-th virtual monitoring system are used to calculate the probability of the ith error alarm of the i-th virtual monitoring system.

Then, in step S204, the fusion parameter setting unit 33 is configured to set the fusion parameter group according to the estimated i-th to N-th detection leakage probability and the i-th to Nth false alarm probability of the i-th to N-th virtual monitoring systems. Then, in step S205, the determining unit 34 performs data fusion of the i-th to N-th detection results of the i-th to N-th virtual monitoring systems according to the fusion parameter group to generate a decision result.

Accordingly, the data fusion-based security monitoring system and method provided by the embodiments of the present invention can perform multiple virtual monitoring systems formed by different algorithms by using at least one physical monitoring system without adding an entity monitoring system. Multiple detection results, data fusion of multiple detection results, resulting in decision-making results, to reduce the overall detection of missed probability and false alarm probability. In other words, the data fusion-based security monitoring system and method provides a low-cost technical solution for reducing the overall detection loss probability and false alarm probability.

The present invention has been disclosed in its preferred embodiments, and it should be understood by those skilled in the art that the present invention is not intended to limit the scope of the invention. It should be noted that variations and permutations equivalent to those of the foregoing embodiments are intended to be included within the scope of the present invention. Therefore, the scope of protection of the present invention is defined by the scope of the patent application.

Claims (10)

A data fusion-based security monitoring system includes: a first to an Nth virtual monitoring system, wherein one or more entity monitoring systems execute different multiple algorithms to form the first to Nth virtual monitoring systems, and And greater than or equal to 2; and a data fusion and decision making device, connecting the first to Nth virtual monitoring systems for defining a first to Nth detection model of the first to Nth virtual monitoring systems, wherein the i-th detecting The measurement model is used to indicate the relationship between the plurality of i-th monitoring data of the i-th virtual monitoring system and the plurality of i-th detecting results corresponding to the plurality of i-th monitoring data, wherein i is an integer from 1 to N; The data fusion and decision-making apparatus according to the ith virtual monitoring system, the number of the plurality of detected erroneous samples of the plurality of ith monitoring data under the condition of corresponding multiple locations, multiple environmental context data, and an intruder, The ith detection model, the plurality of ith monitoring data at the locations of the ith virtual monitoring system, and the plurality of environmental context data estimates an ith detection loss probability of the ith virtual monitoring system ; the data fusion and decision The device determines a fusion parameter group according to the first to Nth detection loss probability rates; and the data fusion and decision device detects the first to Nth detections of the first to Nth virtual monitoring systems according to the fusion parameter group The results were fused to produce a decision result. The data fusion-based security monitoring system according to Item 1, wherein the data fusion and decision-making device is further configured according to the condition of the i-th virtual monitoring system corresponding to the environment context data and the intruder. The number of the plurality of erroneous alarm samples of the plurality of ith monitoring data, the ith detection model, and the ith virtual monitoring system The plurality of ith monitoring data of the equal position and the plurality of environmental context data estimates an ith error warning probability of the ith virtual monitoring system; and the data fusion and decision device according to the first to Nth Detecting The measurement failure rate and the first to Nth false alarm probability determine the fusion parameter set. The data fusion-based security monitoring system of claim 1, wherein the plurality of known monitoring data obtained by the ith virtual monitoring system and the plurality of known monitoring materials corresponding to the plurality of The detection result is based on a machine learning algorithm or an artificial intelligence algorithm to define the i-th detection model. The data fusion-based security monitoring system of claim 1, wherein the data fusion is performed by using a logical operation function, a reliability rule or an environmental rule, and the fusion parameter group is used. The logical operation function, the reliability rule, or the environmental rule is determined. The data fusion-based security monitoring system of claim 1, wherein each of the plurality of environmental context data is a one-day data defining variables such as rainfall, wind, temperature, and brightness. A data fusion-based security monitoring method includes: defining a first to Nth detection models of a first to an Nth virtual monitoring system, wherein one or more entity monitoring systems execute different multiple algorithms to form the The first to the Nth virtual monitoring system, the ith detection model is configured to indicate between the plurality of i-th monitoring data of the i-th virtual monitoring system and the plurality of i-th detecting results corresponding to the plurality of i-th monitoring data Relationship, where i is an integer from 1 to N, and N is greater than or equal to 2; According to the ith virtual monitoring system, the plurality of Detected Missing samples, the ith detection model, and the plurality of ith monitoring data, corresponding to the plurality of locations, the plurality of environmental context data, and the intruder The plurality of ith monitoring data at the locations of the i-th virtual monitoring system and the plurality of environmental context data estimates an ith detection probability of the ith virtual monitoring system; according to the first to the first The N detection miss rate determines a fusion parameter group; and the first to Nth detection results of the first to Nth virtual monitoring systems are data fusion according to the fusion parameter group to generate a decision result. The data fusion-based security monitoring method according to Item 6 of the claim further includes: according to the ith virtual monitoring system, corresponding to the location, the plurality of environmental context data, and no intruder The plurality of erroneous alarm samples of the plurality of ith monitoring data, the ith detection model, the plurality of ith monitoring data of the ith virtual monitoring system, and the plurality of environmental context data An ith error alarm probability of the ith virtual monitoring system is estimated, wherein the fused parameter set is determined according to the first to Nth detection loss probability and the first to Nth false alarm probability. The data fusion-based security monitoring method according to Item 6 of the claim, wherein the plurality of known monitoring data of the ith virtual monitoring system obtained and the plurality of known monitoring materials corresponding to each other are obtained. The detection result is based on defining the i-th detection model based on a machine learning algorithm or an artificial intelligence algorithm. The data fusion-based security monitoring method according to Item 6 of the claim, wherein the data fusion is performed by using a logical operation function, a reliability rule or an environment. The rules are used for data fusion, and the fusion parameter set is used to determine the logical operation function, the reliability rule or the environment rule. The data fusion-based security monitoring method of claim 6, wherein each of the plurality of environmental context data is a one-day data defining variables such as rainfall, wind, temperature, and brightness.