Fire detection method and device based on multi-sensor fusion in complex environment, medium and product

Abstract

The application discloses a fire detection method and device based on multi-sensor fusion in complex environment, medium and product, and relates to the field of fire detection. The method uses infrared sensors and visible light sensors to detect fire threat areas respectively, obtains basic probability assignments of the infrared sensors and the visible light sensors, calculates a conflict coefficient in D-S evidence theory according to the two basic probability assignments, adopts a combination rule of the D-S evidence theory for multi-sensor fusion when the conflict coefficient is not equal to 1, adopts an improved combination rule of the D-S evidence theory for multi-sensor fusion when the conflict coefficient is equal to 1, and introduces a conflict distribution factor in the improved combination rule. The D-S evidence theory based on the improved combination rule can effectively correct the single-sensor misidentification, thereby guaranteeing the accuracy and reliability of a final fire danger area identification result, and realizing real-time fire detection in a complex scene.

Description

Technical Field

[0001] This invention relates to the field of fire detection, and in particular to a fire detection method, device, medium and product based on multi-sensor fusion in complex environments. Background Technology

[0002] Currently, with the rapid development of sensor technology, hazardous area detection methods based on multi-sensor fusion have become a research hotspot and challenge in the field of fire protection. Among them, fusion detection methods, represented by visible light sensors and infrared sensors, have extremely high civilian and military value. In modern firefighting, efficient detection of hazardous sources has become a decisive factor in determining the success or failure of firefighting efforts. Compared with conventional detection methods, multi-sensor fusion detection methods, due to their wide detection range, high detection accuracy, and short detection time, have become a modern detection platform that combines accurate and efficient detection with precise location of hazardous sources. Multi-source information fusion detection platforms, represented by visible light sensors and infrared sensors, have become important equipment and platforms for current firefighting processes due to their fast detection response, wide detection range, and certain anti-interference capabilities.

[0003] Currently, in the field of fire detection based on advanced sensors, most research focuses on improving the detection range and accuracy of single sensors, assuming that all sensors involved in the detection process are functioning normally. However, in actual firefighting operations, the stability and reliability of single sensors are difficult to guarantee. On the one hand, real-world fire scenarios place extremely high demands on the reliability and efficiency of detection methods; when encountering interference from physical facilities or signals, the robustness and efficiency of single sensors are hard to guarantee. On the other hand, during prolonged continuous detection, single sensors are prone to erroneous detection results due to thermal fatigue and electrical characteristic degradation, thus wasting valuable rescue time. Therefore, researching robust collaborative fire detection methods based on multi-sensor information fusion is one of the urgent challenges that needs to be addressed in the current field of fire detection.

[0004] In recent years, improving the accuracy and efficiency of single-sensor detection technology in fire protection has become a research hotspot and challenge, attracting considerable attention from scholars. However, collaborative fire protection detection methods based on multi-sensor information fusion in complex scenarios have received little research. Considering the need to quickly locate life in rubble and rescue trapped individuals during frequent disasters, Yan Zhi studied the application of radar-based life detectors in fire rescue, demonstrating the high efficiency of radar detectors in deep-seated personnel rescue by comparing them with traditional search and rescue methods such as search dogs. Li Ning analyzed the in-depth technical roadmap of life detectors, analyzed the advantages and disadvantages of pulse wave and continuous wave life detectors, summarized the problems encountered in their practical applications, and provided several methods and approaches to improve detection efficiency. Zhang Yawan et al. proposed a fire robot detection system based on OpenCV, utilizing features such as flame color and area to judge flame information in images and videos in real time. Experiments showed that the fire robot detection system based on visible light sensors achieved an accuracy rate exceeding 96% for image sample sets. Gu Yunpeng et al. proposed a temperature detection alarm device based on an infrared sensor. Utilizing the infrared sensor's ability to sense temperature, they designed a temperature detection alarm device with short response time, low maintenance cost, and high maintenance efficiency. This reduces the cost of previously requiring power outages for equipment maintenance, achieves contactless detection, lowers safety risks, and improves safety. However, current fire detection equipment and methods are mostly based on single sensors. Research focuses primarily on improving the detection accuracy and efficiency of single sensors, with little consideration given to fire detection methods and means under complex interference scenarios. The analysis of the stability and robustness of single sensors is insufficient. Furthermore, to date, there is no systematic robust collaborative fire detection method based on multi-sensor information fusion, and a lack of detection methods for complex interference scenarios is also lacking.

[0005] In fire detection using multi-sensor information fusion, the detection information from multiple sensors often takes different forms. Therefore, the raw detection data from multiple sensors needs to be processed and extracted before efficient fusion. Common multi-sensor fusion methods include data-level fusion, feature-level fusion, and decision-level fusion. Data-level fusion processes large amounts of data and has poor real-time performance. Feature-level fusion requires the extraction of data features. This is because real-time detection is crucial in fire detection, and the detection environment is unknown, making it impossible to determine the features in each actual scenario in advance. Decision-level data fusion provides a perfect solution for multi-sensor information fusion in fire detection. Therefore, how to research fire detection methods for complex scenarios based on decision-level fusion to achieve efficient and real-time fire detection is a pressing technical problem that needs to be solved. Summary of the Invention

[0006] The purpose of this invention is to provide a fire detection method, device, medium and product based on multi-sensor fusion in complex environments, which enables real-time and accurate fire detection in complex scenarios based on decision-level fusion.

[0007] To achieve the above objectives, the present invention provides the following solution:

[0008] A fire detection method based on multi-sensor fusion in complex environments includes:

[0009] Infrared and visible light sensors are used to detect fire-threatened areas. The temperature value of the fire source in the infrared sensor detection results constitutes the first decision vector, and the trust level of the fire source in the visible light sensor detection results constitutes the second decision vector.

[0010] The first decision vector and the second decision vector are normalized respectively, and the normalized first decision vector is used as the basic probability assignment for the infrared sensor, and the normalized second decision vector is used as the basic probability assignment for the visible light sensor.

[0011] Based on the basic probability assignments of infrared sensors and visible light sensors, the conflict coefficient in the DS evidence theory is calculated.

[0012] If the conflict coefficient is not equal to 1, the basic probability assignments of the infrared sensor and the visible light sensor are fused using the combination rules of DS evidence theory to obtain the multi-sensor decision-level fusion result.

[0013] If the conflict coefficient is equal to 1, the improved combination rule of DS evidence theory is used to fuse the basic probability assignments of the infrared sensor and the visible light sensor to obtain the multi-sensor decision-level fusion result; the improved combination rule introduces a conflict allocation factor.

[0014] Identify hazardous sources in fire-threatened areas based on the results of multi-sensor decision-level fusion.

[0015] Optionally, the formula for normalizing the first decision vector is:

[0016]

[0017] In the formula, m1(θ) u ) represents the normalized value of the u-th temperature value in the first decision vector, r u r v Let u and v represent the u-th and v-th temperature values ​​in the first decision vector, respectively.

[0018] The formula for normalizing the second decision vector is:

[0019]

[0020] In the formula, m2(θ) u′ ) represents the normalized value of the u'-th trust level in the second decision vector, c u′ c v′ Let u' and v' represent the trust levels in the second decision vector, respectively.

[0021] Optionally, the formula for calculating the conflict coefficient is:

[0022]

[0023] In the formula, K1 represents the conflict coefficient, m() represents the basic probability assignment function, and A i Let B represent the i-th proposition. j Z represents the j-th proposition. k Let m1(A) represent the k-th proposition. i Let m2(B) represent the first basic probability assignment function for the i-th proposition. j Let m denote the second basic probability assignment function for the j-th proposition. n (Z k Let ) denote the nth basic probability assignment function for the kth proposition. This represents the empty set.

[0024] Optionally, the combination rule of the DS evidence theory is:

[0025]

[0026] In the formula, C represents the fused proposition, and m(C) represents the degree of support of the evidence for the fused proposition C.

[0027] Optionally, the improved combination rule of the DS evidence theory is as follows:

[0028]

[0029] In the formula, m(A) represents the degree of support of the evidence for the fused proposition A, and A a Let m represent the a-th proposition. b (A a Let ) denote the probability assignment function of the b-th basic proposition for the a-th proposition. Let represent the probability assignment function for the empty set; q(A,m) represents the conflict assignment factor. 2 Θ This represents the power set of the recognition frame Θ.

[0030] Optionally, the method for determining the conflict allocation factor includes:

[0031] The degree of difference between two pieces of evidence is defined as: In the formula, d BPA (m e ,m f ) represents the degree of difference between the evidence, m e and m f Indicates evidence e and evidence f, m e (A p ) indicates that for the Ath p The e-th basic probability assignment function of a proposition, m f (A q Let ) denote the f-th basic probability assignment function for the Aq-th proposition, and D pq A represents p A q Similarity between them p,q=1,2,…,2 N , |·| represents the cardinal function; 2 N This represents the number of all subsets of the recognition frame Θ.

[0032] Based on the aforementioned evidence dissimilarity, the formula Sim(m) is used. e ,m f )=1-d BPA (m e ,m f Sim(m) determines the similarity between two pieces of evidence; where Sim(m) e ,m f ) represents two pieces of evidence m e and m f The degree of similarity between them.

[0033] Based on the aforementioned similarity, the formula is used. Determine the support level of one of the pieces of evidence; where Sup(m) e ) indicates evidence m e The support level, where n represents the number of pieces of evidence.

[0034] Based on the stated support level, using the formula Determine the credibility of one of the pieces of evidence; where Cred(m) e ) indicates evidence m e Credibility; Sup(m g ) indicates evidence m g Support level.

[0035] Based on the stated credibility, using the formula Determine the conflict allocation factor; where q(A,m) represents the conflict allocation factor, and m e(A) represents the e-th basic probability assignment function for the A-th proposition.

[0036] When the relative credibility of one of the pieces of evidence meets the following conditions When the conflict allocation factor is calculated, the formula is: In the formula, Rcred(m e ) indicates evidence m e The relative credibility.

[0037] Optionally, the hazard source in the fire threat area is determined based on the multi-sensor decision-level fusion result, specifically including: identifying the attribute of the threat area with the highest confidence in the multi-sensor decision-level fusion result as the hazard source.

[0038] A computer device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the fire detection method for multi-sensor fusion in complex environments as described in any of the preceding claims.

[0039] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the fire detection method for multi-sensor fusion in complex environments as described above.

[0040] A computer program product includes a computer program that, when executed by a processor, implements the steps of the fire detection method for multi-sensor fusion in complex environments as described above.

[0041] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0042] The fire detection method, device, medium, and product for multi-sensor fusion in complex environments according to embodiments of the present invention use the combination rules of DS evidence theory for multi-sensor fusion when the conflict coefficient is not equal to 1, and use the improved combination rules of DS evidence theory for multi-sensor fusion when the conflict coefficient is equal to 1. The improved combination rules introduce a conflict allocation factor. Based on the improved combination rules, DS evidence theory can effectively correct the misidentification of single sensors, thereby ensuring the accuracy and reliability of the final fire hazard area identification result, and realizing real-time fire detection in complex scenarios. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a flowchart illustrating the fire detection method using multi-sensor fusion in complex environments provided in Embodiment 1 of the present invention.

[0045] Figure 2 This is a schematic diagram of the fire detection method using multi-sensor fusion in complex environments provided in Embodiment 1 of the present invention.

[0046] Figure 3 This is a schematic diagram illustrating the uncertainty relationship of the proposition provided in Embodiment 1 of the present invention.

[0047] Figure 4 This is a diagram of the internal structure of a computer device. Detailed Implementation

[0048] 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, and 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.

[0049] This invention aims to study fire detection methods in complex scenarios. First, multiple sensors are needed to detect environmental information. Each sensor completes its own detection information summarization process based on its own physical characteristics. On this basis, each sensor provides the attribute judgment result of each hazard source in the real-time scenario and calculates its confidence level. Then, based on the DS evidence theory, the results of multiple sensors are fused at the decision level to give the final fusion result and confidence level, thereby obtaining the area with the largest hazard source in the complex scenario and completing an efficient and real-time fire detection process.

[0050] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0051] Example 1

[0052] like Figure 1 As shown in this embodiment, the fire detection method using multi-sensor fusion in complex environments includes the following steps.

[0053] Step 1: Use infrared sensors and visible light sensors to detect the fire threat area. The temperature value of the fire source in the detection results of the infrared sensor constitutes the first decision vector, and the trust level of the fire source in the detection results of the visible light sensor constitutes the second decision vector.

[0054] Infrared and visible light sensors are used to detect fire-threatened areas. Each sensor completes the detection process and obtains the detection result and the corresponding decision vector R = [r1, r2, r3, r4] and c = [c1, c2, c3, c3].

[0055] Step 2: Normalize the first decision vector and the second decision vector respectively, and use the normalized first decision vector as the basic probability assignment for the infrared sensor and the normalized second decision vector as the basic probability assignment for the visible light sensor.

[0056] The decision vector results from each sensor are normalized:

[0057]

[0058]

[0059] In the formula, m1(θ) u ) represents the normalized value of the u-th temperature value in the first decision vector, r u r v Let m2(θ) represent the u-th and v-th temperature values ​​in the first decision vector, respectively. u′ ) represents the normalized value of the u'-th trust level in the second decision vector, c u′ c v′ Let u' and v' represent the trust levels in the second decision vector, respectively.

[0060] The basic probability assignments for the infrared sensor and the visible light sensor are obtained respectively, i.e., the evidence:

[0061] m1=[m1(θ1),m1(θ2),m1(θ3),m1(θ4)]

[0062] m2=[m2(θ1),m2(θ2),m2(θ3),m2(θ4)]

[0063] See Figure 2 After normalization, it is necessary to determine whether a node is a reliable node. If it is, continue to step 3; otherwise, delete the node.

[0064] Step 3: Calculate the conflict coefficient in the DS evidence theory based on the basic probability assignments of the infrared sensor and the visible light sensor.

[0065] In addressing the problem of hazard identification during firefighting, the set of all areas to be identified constitutes the overall identification framework, denoted by Θ, and denoted by θ. iThis represents a specific target category within the framework. Before using DS evidence theory to solve the multi-source sensor fusion identification problem, the following definitions are given:

[0066] Definition 1: Recognize all subsets of frame Θ (2 N The set consisting of (n subsets) is called the power set of Θ, denoted as 2. Θ Let represent the set of all possible propositions that exist in the hypothesis space Θ:

[0067]

[0068] Define function m:2 Θ →[0,1]:

[0069]

[0070] m(A) is called the Basic Probability Assignment Function (BPA) or the Mass function. m(A) represents the degree of support of evidence for proposition A, excluding support for proper subsets of A. m(Θ) indicates the uncertainty caused by errors or incompleteness in information. To represent the empty set, This represents the probability assignment function for the empty set.

[0071] Definition 2: Let a certain recognition frame be Θ. If m:2 Θ →[0,1] is the basic probability assignment on Θ, and the function BEL:2 is defined. Θ →[0,1].

[0072]

[0073] This is called the trust function on Θ. This function represents the total trust in A, and it is easy to see that... BEL(Θ) = 1. Where, It is an empty set. This indicates that the confidence level in the empty set is 0.

[0074] Definition 3: If A is a subset of the recognition frame Θ, and if m(A) > 0 holds, then A is called a focal element of the trust function BEL, and all focal elements of BEL are called the kernel.

[0075] Definition 4: Let a certain recognition frame be Θ, and define the function PL:2 Θ →[0,1].

[0076]

[0077] This is called the likelihood function. This function represents the degree of confidence in A without negation. It is easy to see that BEL(A)≤PL(A). Since BEL(A) represents the degree of confidence in A, the interval (BEL(A),PL(A)) represents the uncertainty of A.

[0078] Definition 5: [BEL(A),PL(A)] represents the confidence interval of a focal element A. According to this definition, it is easy to see that (1,1) means A is true; (0,0) means A is false; (0,1) means there is no information about A.

[0079] Based on the five definitions above, the certainty and uncertainty of a proposition are represented... Figure 3 As shown.

[0080] Let m1,…,m n These are n independent basic probability assignment functions, BEL1,…,BEL, on the same recognition frame Θ. n Let each be a trust function. Then, the combined basic probability allocation function can be calculated using the following formula:

[0081]

[0082] Where C represents the fused proposition, and m(C) is the basic probability assignment function (BPA) or Mass function. m(C) indicates the degree of support of the evidence for proposition C. This is called the conflict coefficient. A i Let B represent the i-th proposition. j Z represents the j-th proposition. k Let m1(A) represent the k-th proposition. i Let m2(B) represent the first basic probability assignment function for the i-th proposition. j Let m denote the second basic probability assignment function for the j-th proposition. n (Z k Let ) denote the nth basic probability assignment function for the kth proposition. This represents the empty set.

[0083] If K1 = 1, then the evidence m1,…,m is considered valid. n There are contradictions, and there is no combined basic probability assignment; if K1≠1, then m(C) is the basic probability assignment after combination. Called the normalization factor, K1 represents the degree of conflict between the fused pieces of evidence. One of the functions of K1 is to prevent non-zero probabilities from being assigned to an empty set during evidence synthesis. The value of K1 ranges from 0 to 1.

[0084] Commonly used This represents the fusion of two pieces of evidence, m1 and m2. Similarly... Let m1, m2, ..., m be the number of pieces of evidence. n The fusion of sensors. In multi-sensor fusion target recognition, using... This represents the result of fusion identification from n sensors. Furthermore, fusion based on DS evidence theory also satisfies some mathematical relations such as the commutative law:

[0085]

[0086] Associative Law:

[0087]

[0088] In practical fire detection problems, when using DS evidence theory for multi-sensor information fusion, two problems arise. First, when there is a high degree of conflict between the evidence provided by each source, the classic combination rules mentioned above cannot yield reasonable combination results, leading to incorrect decision-making. Second, when the number of sources is large, the traditional combination rules involve a large amount of computation, and the limited computing power of fire detection equipment results in low fusion efficiency and poor real-time performance. Therefore, we propose an improvement. Assume that for a recognition framework Θ = {θ1, θ2, θ3}, four pieces of evidence m1, m2, m3, m4 are known, as shown in Table 1.

[0089] Table 1 shows the four BPAs for Θ.

[0090]

[0091] Using the fusion rules described above, the resulting fused BPA and conflict coefficient K1 are shown in Table 2.

[0092] Table 2 Results after fusion

[0093]

[0094] Tables 1 and 2 show that evidence m1, m2, m3, and m4 supported θ1, θ3, θ1, and θ1 with high confidence levels before fusion, but the result after fusion was different. All evidence supports θ3, which leads to results that contradict reality. Therefore, the results obtained by fusing conflicting evidence using this fusion rule are unreliable. For combining such conflicting evidence, one approach is to modify the evidence source, and another is to improve the combination rule. In fire detection based on multi-source sensor information fusion, modifying the evidence source can easily result in the loss of original sensor observation data, leading to observational distortion and affecting the final identification effect. Therefore, an improved combination rule method is adopted for fusion identification.

[0095] Step 4: If the conflict coefficient is not equal to 1, the basic probability assignments of the infrared sensor and the visible light sensor are fused using the combination rules of DS evidence theory to obtain the multi-sensor decision-level fusion result.

[0096] Using DS evidence theory to fuse basic probability assignments (i.e., evidence):

[0097]

[0098] Step 5: If the conflict coefficient is equal to 1, the basic probability assignments of the infrared sensor and the visible light sensor are fused using the improved combination rule of DS evidence theory to obtain the multi-sensor decision-level fusion result; the improved combination rule introduces a conflict allocation factor.

[0099] Using the concept of weighted average, when two pieces of evidence conflict, the evidence conflict allocation factor is as follows:

[0100]

[0101] Among them, A j Let m represent the j-th proposition. i (A j Let represent the probability assignment function (BPA, also known as evidence or credibility) for the j-th proposition. While this conflict factor can resolve evidence conflict issues, it does not consider the different credibility levels of each piece of evidence. Therefore, a conflict assignment factor that considers the credibility of evidence is proposed. The following definition is given first:

[0102] Define two pieces of evidence m e and m f The degree of difference in evidence between them d BPA (m e ,m f )for:

[0103]

[0104] in, m e and m f Indicates evidence e and evidence f, m e (A p ) indicates that for the Ath p The e-th basic probability assignment function of a proposition, m f (A q ) indicates that for the Ath q The f-th basic probability assignment function of a proposition, D pq A represents p A q The similarity between them is calculated as follows:

[0105]

[0106] Where |·| represents the cardinality function, which calculates the number of elements; 2 N This represents the number of all subsets of the recognition frame Θ.

[0107] Then, define m e and m f The similarity between them is Sim(m) e ,m f )for:

[0108] Sim(m e ,m f )=1-d BPA (m e ,m f )

[0109] Based on this, define evidence m e The support level is Sup(m) e The specific calculation method is as follows:

[0110]

[0111] Where n represents the number of pieces of evidence.

[0112] Next, define evidence m e The credibility is Cred(m) e The calculation method is as follows:

[0113]

[0114] Among them, Sup(m g ) indicates evidence m g Support level.

[0115] Finally, based on the basic definitions given above, this invention proposes a modified conflict allocation factor for evidence credibility as follows:

[0116]

[0117] Where, m e (A) represents the e-th basic probability assignment function for the A-th proposition, Rcred(m) e ) is evidence m e The relative credibility is expressed as follows:

[0118]

[0119] Furthermore, we can obtain:

[0120]

[0121] Furthermore, q(A,m) must also satisfy a normalization condition to be used as a conflict allocation factor, namely:

[0122]

[0123] Then we have:

[0124]

[0125] In summary, the modified combination rules can be obtained as follows:

[0126]

[0127] The conflict allocation factor is calculated as follows:

[0128]

[0129] Step 6: Identify the hazard sources in the fire threat area based on the multi-sensor decision-level fusion results.

[0130] The identification framework for the fire detection area to be identified is represented as Θ={θ1,θ2,θ3,θ4}, where θ1,θ2,θ3,θ4 represent different danger areas, and one of the danger sources is the actual danger source.

[0131] Hazardous sources in the fire protection zone are determined based on the result m(A) after multi-sensor decision-level fusion. The threat area with the highest confidence level in the multi-sensor decision-level fusion result is identified as the hazard source.

[0132] Considering fire detection methods and means in complex interference scenarios, the stability and robustness of single sensors are poor. On the one hand, in actual fire scenarios, the reliability and efficiency of detection methods are extremely demanding. When encountering interference from physical facilities or signals, the robustness and efficiency of single sensors are difficult to guarantee. On the other hand, during long-term continuous detection, single sensors are prone to erroneous detection results due to thermal fatigue and electrical characteristic degradation, thus wasting valuable rescue time. Therefore, a robust collaborative fire detection method based on multi-sensor information fusion for complex interference scenarios is proposed. First, multiple sensors detect environmental information, and each sensor completes its own detection information summarization and judgment process based on its own physical characteristics. On this basis, each sensor provides the attribute judgment result of each hazard source in the scene in real time and calculates its confidence level. Then, based on DS evidence theory, the results of multiple sensors are fused at the decision level to give the final fusion result and confidence level. Thus, the area with the largest hazard source in the complex scenario is obtained, completing an efficient and real-time fire detection process.

[0133] The effectiveness of the method of the present invention is verified below:

[0134] Assume that at a certain moment, the fire hazard zone identification results calculated by two sensors are shown in Tables 3 and 4. The fire detection zone identification framework to be identified is represented as Θ={θ1,θ2,θ3,θ4}, where θ1,θ2,θ3,θ4 represent different hazard zones, and one of them is the actual hazard source.

[0135] Table 3 Infrared sensor identification results

[0136]

[0137] Table 4. Visible light sensor identification results

[0138]

[0139] Based on the single-sensor identification above, the DS evidence theory with the proposed modified combination rules is fused, and the fusion results are shown in Table 5:

[0140] Table 5. Fusion Results Based on Improved DS Evidence Theory

[0141]

[0142]

[0143] Where R represents infrared sensor-based identification, C represents visible light sensor-based identification, and R_C represents fused identification. Analyzing the original and fused identification results obtained by each sensor in Tables 3 and 4, it can be seen that for the same type of target, the fusion identification using the improved combination rule-based DS evidence theory yields a higher level of confidence and more reliable identification results. In the identification of the second type of target, the infrared sensor-based identification incorrectly identified the second-type hazard area θ2 as the fourth-type hazard area θ4 due to some reasons, while the multi-sensor fusion identification using DS evidence theory yielded θ2 as the correct identification result. Furthermore, it can be seen that the DS evidence theory fusion identification method based on the improved combination rule can effectively correct single-sensor misidentification, thus ensuring the accuracy and reliability of the final fire hazard area identification results.

[0144] This invention first considers a scenario where multiple infrared and visible light sensors collaboratively detect the same hazardous area. It assumes that each sensor node can independently complete the detection and output its own judgment result and the confidence level of each result. Then, considering the influence of various complex interference factors, the detection results of each sensor are different. Based on the DS evidence theory, the detection results of each sensor are fused at the decision level to obtain the fused confidence level. Finally, the attribute of the threat area with the highest confidence level is used as the final fused identification result. Finally, a case study is provided to compare the identification results of each single sensor and the multi-sensor fused identification result for the same hazardous area, demonstrating the accuracy and reliability of the fire detection and identification method based on multi-sensor multi-source information fusion.

[0145] Example 2

[0146] A computer device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the fire detection method for multi-sensor fusion in complex environments as described in Embodiment 1.

[0147] Example 3

[0148] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the fire detection method for multi-sensor fusion in complex environments as described in Embodiment 1.

[0149] Example 4

[0150] A computer program product includes a computer program that, when executed by a processor, implements the steps of the fire detection method for multi-sensor fusion in complex environments as described in Embodiment 1.

[0151] Example 5

[0152] A computer device, the internal structure of which can be shown in the diagram below. Figure 4As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores pending transactions. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network. When the computer program is executed by the processor, it implements the multi-sensor fusion fire detection method in complex environments described in Embodiment 1.

[0153] It should be noted that the object information (including but not limited to object device information, object personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this invention are all information and data authorized by the object or fully authorized by all parties.

[0154] 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 described above. Any references to memory, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided by this invention may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided by this invention may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0155] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0156] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A fire detection method using multi-sensor fusion in complex environments, characterized in that, include: Infrared and visible light sensors are used to detect fire-threatened areas. The temperature value of the fire source in the infrared sensor detection results constitutes the first decision vector, and the trust level of the fire source in the visible light sensor detection results constitutes the second decision vector. The first decision vector and the second decision vector are normalized respectively, and the normalized first decision vector is used as the basic probability assignment of the infrared sensor, and the normalized second decision vector is used as the basic probability assignment of the visible light sensor. Calculate the conflict coefficient in the DS evidence theory based on the basic probability assignments of infrared sensors and visible light sensors. If the conflict coefficient is not equal to 1, the combination rules of DS evidence theory are used to fuse the basic probability assignments of the infrared sensor and the basic probability assignments of the visible light sensor to obtain the multi-sensor decision-level fusion result. If the conflict coefficient is equal to 1, the basic probability assignments of the infrared sensor and the visible light sensor are fused using the improved combination rule of DS evidence theory to obtain the multi-sensor decision-level fusion result; the improved combination rule introduces a conflict allocation factor. Based on the multi-sensor decision-level fusion results, the danger sources in the fire threat area are identified, specifically including: identifying the threat area with the highest confidence in the multi-sensor decision-level fusion results as the danger source; The method for determining the conflict allocation factor includes: The degree of difference in evidence between two pieces of evidence is defined as: In the formula, d BPA (m e ,m f ) represents the degree of difference between the evidence, m e and m f Indicates evidence e and evidence f, m e (A p ) indicates that for the Ath p The e-th basic probability assignment function of a proposition, m f (A q ) indicates that for the Ath q The f-th basic probability assignment function of a proposition; D pq A represents p A q Similarity between them |·| represents the cardinality function, 2 N This represents the number of all subsets of the recognition frame Θ; Based on the aforementioned evidence dissimilarity, the formula Sim(m) is used. e ,m f )=1-d BPA (m e ,m f Sim(m) determines the similarity between two pieces of evidence; where Sim(m) e ,m f ) represents two pieces of evidence m e and m f The degree of similarity between them; Based on the aforementioned similarity, the formula is used. Determine the support level of one of the pieces of evidence; where Sup(m) e ) indicates evidence m e The support level, where n represents the number of pieces of evidence; Based on the stated support level, using the formula Determine the credibility of one of the pieces of evidence; where Cred(m) e ) indicates evidence m e Credibility; Sup(m g ) indicates evidence m g Support level; Based on the stated credibility, using the formula Determine the conflict allocation factor; where q(A,m) represents the conflict allocation factor, and m e (A) represents the e-th basic probability assignment function for the A-th proposition; When the relative credibility of one of the pieces of evidence meets the following conditions When the conflict allocation factor is calculated, the formula is: In the formula, Rcred(m e ) indicates evidence m e The relative credibility.

2. The fire detection method based on multi-sensor fusion in complex environments according to claim 1, characterized in that, The formula for normalizing the first decision vector is: In the formula, m1(θ) u ) represents the normalized value of the u-th temperature value in the first decision vector, r u r v Let u and v represent the u-th and v-th temperature values ​​in the first decision vector, respectively. The formula for normalizing the second decision vector is: In the formula, m2(θ) u′ ) represents the normalized value of the u'-th trust level in the second decision vector, c u′ c v′ Let u' and v' represent the trust levels in the second decision vector, respectively.

3. The fire detection method based on multi-sensor fusion in complex environments according to claim 1, characterized in that, The formula for calculating the conflict coefficient is: In the formula, K1 represents the conflict coefficient, m() represents the basic probability assignment function, and A i Let B represent the i-th proposition. j Z represents the j-th proposition. k Let m1(A) represent the k-th proposition. i Let m2(B) represent the first basic probability assignment function for the i-th proposition. j Let m denote the second basic probability assignment function for the j-th proposition. n (Z k Let ) denote the nth basic probability assignment function for the kth proposition. This represents the empty set.

4. The fire detection method using multi-sensor fusion in complex environments according to claim 3, characterized in that, The combination rule of the DS evidence theory is as follows: In the formula, C represents the fused proposition, and m(C) represents the degree of support of the evidence for the fused proposition C.

5. The fire detection method using multi-sensor fusion in complex environments according to claim 3, characterized in that, The improved combination rule of the DS evidence theory is as follows: In the formula, m(A) represents the degree of support of the evidence for the fused proposition A, and A a Let m represent the a-th proposition. b (A a Let ) denote the probability assignment function of the b-th basic proposition for the a-th proposition. Let represent the probability assignment function for the empty set; q(A,m) represents the conflict assignment factor. 2 Θ This represents the power set of the recognition frame Θ.

6. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the fire detection method for multi-sensor fusion in complex environments as described in any one of claims 1-5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the fire detection method for multi-sensor fusion in complex environments as described in any one of claims 1-5.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the fire detection method for multi-sensor fusion in complex environments as described in any one of claims 1-5.

Images