Fire detection method, system, device and medium based on flame visual blurring

Abstract

The present invention discloses a fire detection method, system, device, and medium based on flame visual blurring. The method comprises: obtaining a first video image of an area to be detected, performing color segmentation on the first video image to obtain a suspected flame area; determining the dispersion characteristics, similarity characteristics, and center of mass motion characteristics of the suspected flame area; extracting the flame visual blurring characteristics of the suspected flame area through binocular ranging and laser ranging; constructing an MES multi-expert decision system, and performing feature fusion and decision analysis on the dispersion characteristics, similarity characteristics, center of mass motion characteristics, and flame visual blurring characteristics through the MES multi-expert decision system to obtain a fire detection result. The present invention can accurately eliminate interference from light bulbs, reflections, red fire extinguishers, etc., reduce the false positive rate of fire detection, detect flames generated by various fuels, reduce the missed detection rate of fire detection, and improve the accuracy of fire detection. The invention can be widely used in the field of fire detection technology.

Description

Technical Field

[0001] The present invention relates to the field of fire detection technology, and in particular to a fire detection method, system, device and medium based on flame visual blurring. Background Art

[0002] The frequency and prevalence of fires threaten public safety and social development and are considered one of the most serious threats to our daily lives. The rapid spread of fires and the significant property losses they inflict make fire prevention and avoidance essential for protecting lives and property. Low-cost, reliable, and widely available fire alarm systems enable people to detect fires early and escape safely and quickly. Fire detection and alarm systems can also minimize fires and the damage they cause to people.

[0003] In industry, agriculture, and all aspects of social production and life, traditional fire detection systems no longer meet the practical needs of fire alarm systems in complex environments. Most automatic fire alarm systems use a single passive sensor, which inevitably presents some problems. For example, some devices using photosensitive detectors are affected by sunlight and light. Smoke detectors can be affected by various gases. As the number of sensors installed increases, the false alarm rate of multi-sensor flame detection technology based on simple algorithms will increase significantly, resulting in multiple missed or false alarms in the alarm system.

[0004] Video fire detection technology is a new technology that has only been applied in fire detection in recent decades. It is a non-contact method that is faster, more intelligent, and more reliable than traditional methods such as smoke detection and temperature detection. Conventional color cameras are typically used to capture scene video and extract unique fire characteristics such as color and shape as input to fire detection algorithms. However, image-based flame detection technology is not highly accurate and exhibits complex characteristics in complex environments. Therefore, research into flame detection technologies beyond sensors and image-based methods is crucial.

[0005] Explanation of terms:

[0006] Visual blurring of flame refers to the fact that the glowing, volatile portion of a flame is visible to the human eye and can be captured by cameras. However, special light sources like lasers penetrate the flame and are not reflected, resulting in the flame appearing blurred. Common interfering objects are generally solid and do not exhibit visual blurring.

[0007] Multi-feature fusion: Multi-feature fusion, also known as multi-sensor correlation, multi-sensor fusion, etc., refers to extracting information from different aspects of the same thing, using computers to comprehensively analyze various features based on certain criteria, and then accurately and comprehensively judge the same thing. Summary of the Invention

[0008] The purpose of the present invention is to solve one of the technical problems existing in the prior art to at least a certain extent.

[0009] To this end, an object of an embodiment of the present invention is to provide a fire detection method based on the visual blurring of flames. The method uses an RGB-HIS color segmentation model to extract suspected flame areas, and simultaneously captures the visual blurring characteristics of flames through binocular ranging and laser ranging. The method combines the dispersion characteristics, similarity characteristics, and center of mass motion characteristics of the flames through a MES multi-expert decision system to perform feature fusion and decision analysis to obtain fire detection results, thereby reducing the false positive rate and missed detection rate of fire detection and improving the accuracy of fire detection.

[0010] Another object of the embodiments of the present invention is to provide a fire detection system based on flame visual blurring.

[0011] In order to achieve the above technical objectives, the technical solutions adopted by the embodiments of the present invention include:

[0012] In a first aspect, an embodiment of the present invention provides a fire detection method based on flame visual blurring, comprising the following steps:

[0013] Acquire a first video image of the area to be detected, perform color segmentation on the first video image, and obtain a suspected flame area;

[0014] Determining the dispersion characteristics, similarity characteristics, and center of mass motion characteristics of the suspected flame area;

[0015] Extracting the flame visual blurring characteristics of the suspected flame area by binocular ranging and laser ranging;

[0016] An MES multi-expert decision system is constructed, and the MES multi-expert decision system performs feature fusion and decision analysis on the dispersion feature, the similarity feature, the center of mass motion feature, and the flame visual blurring feature to obtain a fire detection result.

[0017] Furthermore, in one embodiment of the present invention, the step of performing color segmentation on the first video image to obtain the suspected flame area specifically includes:

[0018] Determine the red component threshold and saturation threshold;

[0019] Constructing an RGB-HIS color segmentation model according to the red component threshold and the saturation threshold;

[0020] Color segmentation is performed on the first video image using the RGB-HIS color segmentation model to obtain a suspected flame area.

[0021] Furthermore, in one embodiment of the present invention, the RGB-HIS color segmentation model is:

[0022] R>G>B

[0023] R>R T

[0024] S>(255-R)×S T / R T

[0025] Among them, R represents the red component of the target pixel, G represents the green component of the target pixel, B represents the blue component of the target pixel, and R T represents the red component threshold, S represents the saturation of the target pixel, S T Indicates the saturation threshold.

[0026] Furthermore, in one embodiment of the present invention, the step of determining the dispersion feature, similarity feature, and center of mass motion feature of the suspected flame area specifically includes:

[0027] performing image analysis on the suspected flame area to extract dispersion features of each part of the suspected flame area;

[0028] Comparing the suspected flame regions in consecutive frames of the first video image to obtain similarity features of the suspected flame regions;

[0029] The center of mass position of the suspected flame area is determined, and the center of mass motion feature of the first video image is determined based on the center of mass position.

[0030] Furthermore, in one embodiment of the present invention, the step of extracting the flame visual blurring feature of the suspected flame area by binocular ranging and laser ranging specifically includes:

[0031] Performing binocular ranging on the suspected flame area using a binocular camera to obtain first flame depth information of the suspected flame area;

[0032] performing laser ranging on the suspected flame area by a laser measurement system to obtain second flame depth information of the suspected flame area;

[0033] The flame visual blurring feature of the suspected flame area is determined according to the difference between the first flame depth information and the second flame depth information.

[0034] Furthermore, in one embodiment of the present invention, the step of constructing the MES multi-expert decision system specifically includes:

[0035] Obtain pre-trained dispersion classifier, similarity classifier, centroid motion classifier, and flame visual blurring classifier;

[0036] Determining a first weight of the dispersion classifier, a second weight of the similarity classifier, a third weight of the centroid motion classifier, and a fourth weight of the flame visual blurring classifier;

[0037] An MES multi-expert decision system is constructed according to the dispersion classifier, the similarity classifier, the centroid motion classifier, the flame visual blurring classifier, the first weight, the second weight, the third weight and the fourth weight.

[0038] Furthermore, in one embodiment of the present invention, the step of performing feature fusion and decision analysis on the dispersion feature, the similarity feature, the centroid motion feature, and the flame visual blurring feature by the MES multi-expert decision system to obtain a fire detection result specifically includes:

[0039] The dispersion feature, the similarity feature, the centroid motion feature, and the flame visual blur feature are respectively classified according to the dispersion classifier, the similarity classifier, the centroid motion classifier, and the flame visual blur feature to obtain a plurality of flame classification labels;

[0040] Feature fusion is performed based on the first weight, the second weight, the third weight, the fourth weight, and the flame classification label to obtain a first weighted sum, and then whether a fire occurs in the suspected flame area is determined based on the first weighted sum and a preset threshold.

[0041] In a second aspect, an embodiment of the present invention provides a fire detection system based on flame visual blurring, comprising:

[0042] A color segmentation module is used to obtain a first video image of the area to be detected, perform color segmentation on the first video image, and obtain a suspected flame area;

[0043] A first feature determination module is used to determine the dispersion feature, similarity feature and centroid motion feature of the suspected flame area;

[0044] A second feature determination module is used to extract the flame visual blurring feature of the suspected flame area through binocular ranging and laser ranging;

[0045] The feature fusion module is used to build an MES multi-expert decision system, through which the dispersion feature, the similarity feature, the center of mass motion feature and the flame visual blur feature are subjected to feature fusion and decision analysis to obtain a fire detection result.

[0046] In a third aspect, an embodiment of the present invention provides a fire detection device based on flame visual blurring, comprising:

[0047] at least one processor;

[0048] at least one memory for storing at least one program;

[0049] When the at least one program is executed by the at least one processor, the at least one processor implements the above-mentioned fire detection method based on flame visual blurring.

[0050] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium storing a program executable by a processor, wherein the program executable by the processor is used to execute the above-mentioned fire detection method based on flame visual blurring when executed by the processor.

[0051] The advantages and benefits of the present invention will be described in part in the following description and will become apparent from the following description or learned through practice of the present invention:

[0052] The embodiment of the present invention uses the RGB-HIS color segmentation model to extract suspected flame areas, and simultaneously captures the visual blurring characteristics of the flame through binocular ranging and laser ranging. The dispersion characteristics, similarity characteristics, and center of mass motion characteristics of the flame are combined through the MES multi-expert decision system to perform feature fusion and decision analysis to obtain fire detection results. On the one hand, it can accurately eliminate interference from light bulbs, reflections, red fire extinguishers, etc., reducing the false positive rate of fire detection. On the other hand, it can detect flames generated by various fuels, reducing the missed detection rate of fire detection, thereby improving the accuracy of fire detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following introduction is made to the drawings required for use in the embodiments of the present invention. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative work.

[0054] Figure 1 A flowchart of the steps of a fire detection method based on flame visual blurring provided by an embodiment of the present invention;

[0055] Figure 2 A schematic diagram illustrating the principle of binocular ranging for measuring flame depth information provided by an embodiment of the present invention;

[0056] Figure 3 A schematic diagram illustrating the principle of measuring flame depth information using laser ranging according to an embodiment of the present invention;

[0057] FIG4 is a schematic diagram showing a comparison of flame depth information obtained by binocular ranging and laser ranging according to an embodiment of the present invention;

[0058] Figure 5 A schematic diagram of the structure of the MES multi-expert decision-making system provided by an embodiment of the present invention;

[0059] Figure 6 A structural block diagram of a fire detection system based on flame visual blurring provided by an embodiment of the present invention;

[0060] Figure 7 This is a structural block diagram of a fire detection device based on flame visual blurring provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention and are not to be construed as limiting the present invention. The step numbers in the following embodiments are provided for ease of explanation only and do not limit the order of the steps. The order of execution of the steps in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0062] In the description of the present invention, "a plurality" means two or more. The terms "first" and "second" are used solely to distinguish technical features and are not to be construed as indicating or implying relative importance, or as implicitly indicating the number of the indicated technical features, or as implicitly indicating the order of the indicated technical features. Furthermore, unless otherwise defined, all technical and scientific terms used herein have the same meanings as commonly understood by those skilled in the art.

[0063] In the early days, traditional smoke, light, and temperature sensors were used for fire detection. However, these sensors wouldn't sound an alarm until smoke particles or heat had diffused to a certain level, resulting in significant delays in fire detection and preventing their application in large spaces and open areas. In recent years, with the maturity of computer image processing technology, video-based flame detection technology has garnered significant attention. Image-based flame detection overcomes the shortcomings of sensor-based fire detection systems, such as their single structure, poor real-time performance, and low accuracy. By extracting dynamic and static flame features and integrating multiple features to achieve flame detection, this technology represents a significant breakthrough in flame detection technology.

[0064] At present, video image flame detection technology is mainly analyzed from two aspects: static characteristics and dynamic characteristics of flames.

[0065] Static features: The most representative of these are flame color space features, which primarily include RGB, HIS, and YCbCr. Static features also include the flame's spectral information, regional structure, and geometric features. Spectral information encompasses color and saturation features, while flame regional structure includes texture and center of gravity height coefficients. Geometric features include circularity, rectangularity, and flame angles. Color information in static features is one of the most commonly used features in flame detection, effectively distinguishing objects of different colors. The high saturation of flames can distinguish flames from general red objects, but static features cannot effectively eliminate interference from noise and fire-like objects, such as light bulbs, red walls, and fire extinguishers.

[0066] Dynamic characteristics: Fire detection methods that utilize only a single characteristic have drawbacks such as low accuracy and high false alarm rates. Therefore, many studies have further analyzed the dynamic characteristics of flames and integrated both static and dynamic characteristics to reduce false alarm rates. Dynamic characteristics include overall motion and random motion. Overall flame motion characteristics include area change, overall movement, and similarity, while random motion includes flame stroboscopic characteristics and shape changes. The motion characteristics of flames can distinguish between light sources, reflectors, and other objects. However, fire detection methods based on dynamic flame characteristics cannot effectively eliminate interference from moving light sources, such as taillights and sunlight refraction, which could potentially be fire sources. Furthermore, they cannot detect flames produced by special fuels such as magnesium, phosphorus, and copper.

[0067] Multi-feature fusion: Extracting a single flame feature to detect fires is no longer sufficient. While the algorithm is simple, single feature extraction is not very accurate. Furthermore, in complex fire environments, flames exhibit a wide variety of characteristics. Therefore, it is necessary to extract and fuse multiple features. Common feature fusion methods include probabilistic statistics, support vector machines, Bayesian estimation, fuzzy logic, and artificial neural networks. Multi-feature fusion inevitably generates high-dimensional feature vectors. Processing high-dimensional feature vectors is a challenging problem in current research. In machine learning and pattern recognition, using a high-dimensional feature vector significantly increases the data required to train the classifier, avoiding over-specialization to achieve more reliable results. Furthermore, the variability of fires and the high level of noise in the data acquired in fire environments hinder the system's recognition rate.

[0068] To overcome the problems of flame misjudgment and missed detection in visual image fire systems, an embodiment of the present invention discloses a fire detection method based on flame visual blurring. By combining binocular ranging with laser ranging, the method can accurately eliminate interference from light bulbs, reflections, and red fire extinguishers, reducing the misjudgment rate. Furthermore, the method can detect flames generated by various fuels, reducing the missed detection rate. Finally, the flame visual blurring is integrated with the flame features of video images through an MES (Multi-expert system) to accurately detect flames. The system is also applicable to complex indoor and outdoor environments.

[0069] Reference Figure 1 The embodiment of the present invention provides a fire detection method based on flame visual blurring, which specifically includes the following steps:

[0070] S101: Acquire a first video image of a region to be detected, perform color segmentation on the first video image, and obtain a suspected flame region.

[0071] As an optional implementation, the step of performing color segmentation on the first video image to obtain the suspected flame area specifically includes:

[0072] A1. Determine the red component threshold and saturation threshold;

[0073] A2. Construct an RGB-HIS color segmentation model based on the red component threshold and saturation threshold;

[0074] A3. Perform color segmentation on the first video image using an RGB-HIS color segmentation model to obtain a suspected flame area.

[0075] As an optional implementation, the RGB-HIS color segmentation model is:

[0076] R>G>B

[0077] R>R T

[0078] S>(255-R)×S T / R T

[0079] Among them, R represents the red component of the target pixel, G represents the green component of the target pixel, B represents the blue component of the target pixel, and R T represents the red component threshold, S represents the saturation of the target pixel, S T Indicates the saturation threshold.

[0080] Specifically, in order to segment the suspected flame area, the flame color feature is used as the judgment condition. The flame burns from red to yellow, corresponding to the RGB color space, satisfying R>G>B. Since the red component feature is obvious when the flame burns, the red component threshold R is set. T Objects of other colors can be excluded; due to the high saturation of the flame, the flame saturation threshold S is set T It can avoid the influence of low-saturation background objects such as red pedestrians and red walls.

[0081] According to the above flame characteristics, the present invention first uses the RGB-HIS color model to segment the suspected flame area, namely:

[0082] R>G>B

[0083] R>R T

[0084] S>(255-R)×S T / R T

[0085] Among them, R represents the red component of the target pixel, G represents the green component of the target pixel, B represents the blue component of the target pixel, and R T represents the red component threshold, S represents the saturation of the target pixel, S T It is understood that the saturation will decrease as the R component increases. As the R component increases toward the maximum value of 255, the saturation S will decrease to zero. T and S T The ranges are 55-65 and 115-135 respectively. In the embodiment of the invention, R T =55,S T =125.

[0086] S102: Determine the dispersion characteristics, similarity characteristics, and centroid motion characteristics of the suspected flame area.

[0087] As an optional implementation, step S102 specifically includes the following steps:

[0088] S1021, performing image analysis on the suspected flame area to extract dispersion features of each part of the suspected flame area;

[0089] S1022, comparing the suspected flame regions in consecutive frames of the first video image to obtain similarity features of the suspected flame regions;

[0090] S1023: Determine the centroid position of the suspected flame area, and determine the centroid motion characteristics of the first video image based on the centroid position.

[0091] Specifically, the degree of combustion and temperature in different parts of the flame are different, which manifests as a certain degree of dispersion. Dispersion is defined by the standard deviation of the color component. The blue component of the flame is produced by the burning oxygen. The blue component varies greatly in different parts of the flame, so the standard deviation of the blue component is large. The blue component of non-flame objects is usually determined by light. There is no dispersion in a small range, and the standard deviation is very small. The flame has a flickering characteristic when burning. It looks irregular in a single frame, but shows a certain similarity in continuous frames. This is very different from other fast-moving light sources or interfering objects with flame color characteristics. Therefore, similarity can be used as a basis for flame identification. The flame flickers continuously within a certain period of time, and the center of mass moves in a repetitive manner. Therefore, the ratio of the total displacement of the center of mass in the flame area to the total distance the center of mass moves is less than a certain threshold. Common interference sources, such as a moving flashlight, show uniform movement of the center of mass in a short period of time, so the ratio is large.

[0092] S103. Extracting flame visual blur characteristics of the suspected flame area through binocular ranging and laser ranging.

[0093] Specifically, due to the interference of complex environments and different burning materials, only using video image features such as flame color and movement as the basis for flame detection still has a high false detection rate. In order to improve the accuracy of flame detection, the embodiment of the present invention adds a new flame feature, flame visual blurring, and combines binocular ranging with laser ranging to detect the flame visual blurring feature. Step S103 specifically includes the following steps:

[0094] S1031, performing binocular ranging on the suspected flame area using a binocular camera to obtain first flame depth information of the suspected flame area;

[0095] S1032. Perform laser ranging on the suspected flame area using a laser measurement system to obtain second flame depth information of the suspected flame area;

[0096] S1033: Determine a flame visual blurring feature of the suspected flame area according to a difference between the first flame depth information and the second flame depth information.

[0097] Specifically, if Figure 2 As shown in the figure, a pre-calibrated variable baseline binocular camera (HNY-CV-002) is first used to measure the depth information of the first flame. According to the binocular measurement principle: two cameras observe the flame from two different positions at the same time, and the position deviation between the image pixels is calculated using the triangle geometry principle, that is, the disparity d, to obtain the three-dimensional information of the measured target, that is, the three-dimensional coordinates of point P. P (X0, Y0, Z0) is the pixel point of the measured target flame, O l and O r They are the optical centers of the left and right cameras respectively. l (x l ,y l ) and p r (x r ,y r ) are the projection points of point P on the imaging planes of the left and right cameras respectively.

[0098] According to the triangle similarity principle, the first flame depth information measured by the binocular ranging method can be expressed as

[0099] The principle of laser ranging is as follows Figure 3 As shown in the figure, the distance observation value L is obtained by the time or phase difference of the pulse scanning object. According to the horizontal and vertical scanning angle observation values ​​α and β, any measured point P' is obtained and a three-dimensional point cloud is generated. The X-axis is located in the horizontal scanning plane, the Y-axis is perpendicular to the X-axis in the horizontal scanning plane, and the Z-axis is perpendicular to the horizontal scanning plane. When measuring the flame, the laser penetrates the flame pixel point P and shoots to a point P' on the background object such as the wall or the ground.

[0100] The laser measurement system measures the second flame depth information as depth′=L.

[0101] In order to analyze and verify the visual blurring of flames, the embodiment of the present invention establishes a schematic diagram of flame depth information comparison obtained by binocular ranging and laser ranging as shown in FIG4 , and analyzes the following situations:

[0102] Figure 4(a) shows a schematic diagram comparing the flame depth information obtained by binocular ranging and laser ranging when the ranging system and the flame are at the same horizontal position. When the angle α between the line connecting the ranging system and the flame and the horizontal line is 0° and there are no background obstacles, the laser passes through the flame and shoots to infinity. The difference in the flame depth information measured by the two is τ→∞.

[0103] Figure 4(b) shows a schematic diagram comparing the flame depth information obtained by binocular ranging and laser ranging when the ranging system and the flame are at different horizontal positions. The system measures the angle α∈(0,90°), the system suspension height is fixed at h in the actual scene, and the horizontal distance between the flame and the system is s.

[0104] According to the triangle principle:

[0105] According to the trapezoidal principle: where h' = h - stanα, therefore,

[0106] If the distance difference between the two systems is τ, then The value of τ changes with the angle α. The larger α is, the larger τ is, and the smaller α is, the smaller τ is.

[0107] Figure 4(c) shows a schematic diagram comparing the depth information of interference objects obtained by binocular ranging and laser ranging. Regardless of the angle and position from which the ranging system measures interference objects such as light bulbs and pedestrians, the laser will not penetrate the interference objects. The distance values ​​measured by the system are theoretically consistent without considering the physical differences of the equipment. Therefore, τ = 0.

[0108] In real-world environments, the camera is typically suspended at an angle between 30 and 45 degrees. Based on the principle that binocular and laser ranging can reconstruct the three-dimensional surface of an object, the depth information of characteristic pixels within a suspected flame region is measured using binocular and laser ranging. Pixels with τ ≈ 0 represent interference, while τ within the flame region is either finite or infinite. Therefore, the difference between the flame depth information obtained by binocular and laser ranging can be calculated, and this difference can be used to characterize the visual blurring characteristics of the flame.

[0109] S104. Construct an MES multi-expert decision system, and perform feature fusion and decision analysis on dispersion features, similarity features, center of mass motion features, and flame visual blurring features through the MES multi-expert decision system to obtain fire detection results.

[0110] As an optional implementation method, the step of building an MES multi-expert decision system specifically includes:

[0111] B1. Obtain pre-trained dispersion classifier, similarity classifier, centroid motion classifier, and flame visual blurring classifier;

[0112] B2, determining a first weight of the dispersion classifier, a second weight of the similarity classifier, a third weight of the centroid motion classifier, and a fourth weight of the flame visual blurring classifier;

[0113] B3. Construct an MES multi-expert decision system based on the dispersion classifier, similarity classifier, centroid motion classifier, flame visual blurring classifier, first weight, second weight, third weight and fourth weight.

[0114] As an optional implementation, the step of obtaining fire detection results by performing feature fusion and decision analysis on dispersion features, similarity features, centroid motion features, and flame visual blurring features through the MES multi-expert decision system specifically includes:

[0115] C1. Classify the dispersion features, similarity features, centroid motion features, and flame visual blur features according to the dispersion classifier, similarity classifier, centroid motion classifier, and flame visual blur feature classifier to obtain multiple flame classification labels;

[0116] C2. Perform feature fusion based on the first weight, the second weight, the third weight, the fourth weight, and the flame classification label to obtain a first weighted sum, and then determine whether a fire has occurred in the suspected flame area based on the first weighted sum and a preset threshold.

[0117] Specifically, the MES (Multi-expert system) mechanism is widely used in the field of image recognition. While there are many approaches to multi-feature fusion, weighted classification is one of the most effective. MES splits the feature vector and employs a set of classifiers, each tailored to a feature set, which are then trained to become experts in a specific feature space. By combining the results of different single classifiers to make decisions, the MES multi-expert decision system often outperforms a single optimal classifier.

[0118] Therefore, the embodiment of the present invention adopts the MES multi-expert decision-making mechanism to perform multi-feature fusion and establishes Figure 5 The MES system based on the flame visual blurring feature is shown in the figure, where DE represents the color expert, SE represents the similarity expert, VE represents the centroid motion expert, and IE represents the visual blurring expert. The combination of experts and decision rules determine the performance of MES. For k experts, k∈{DE, SE, VE, IE}, the category c selected in the label k (b) The blob assigned to the input (F represents fire, represents non-fire); the vote for class i can be expressed as σ ik (b) If the output corresponds to the category, the voting result is 1, otherwise the result is 0. Weight ω k (i) Dynamic evaluation using the Bayesian formula yields the following maximum recognition rate for MES:

[0119]

[0120] The final decision is to identify a specific class by maximizing the reliability of the entire MES. The reliability of a blob belonging to class i is calculated by voting weights: The decision of c is ultimately made by maximizing the reliability of different classes:

[0121] Each expert in the MES is evaluated and the input blob is classified as flame if it meets the given threshold or interval condition, otherwise it is classified as non-flame. DE (F), W SE (F), W VE (F) and W IE (F) are the recognition accuracy of flame samples based on the color, similarity, centroid motion and flame visual blur of the training data, respectively. and are the recognition accuracy of experts on interference samples.

[0122] Based on the concept of accuracy, the embodiment of the present invention defines the true positive rate, true negative rate, false positive rate, and false negative rate as follows:

[0123] (1) True positive rate, true positive rate = number of true positives / (number of true positives + number of false positives), where a true positive indicates that a real fire is a fire.

[0124] (2) True negative rate, true negative rate = number of true negatives / (number of true negatives + number of false negatives), where true negative refers to the number of interferences detected as non-flame.

[0125] (3) False positive rate, false positive rate = 1-true positive rate, false positive means the interference object is mistakenly judged as flame.

[0126] (4) False negative rate, false negative rate = 1-true negative rate, false negative rate means misjudging the flame as an interference.

[0127] The multi-expert system was trained using random videos from 20% of the dataset. The training results are shown in Table 1. For sample calculation, the average accuracy is defined as accuracy = (true positive rate + true negative rate) / (true positive rate + true negative rate + false positive rate + false negative rate).

[0128]

[0129] Table 1

[0130] The accuracy, false positive rate, and false negative rate of the detection of the embodiment of the present invention are compared with those of other methods. The results are shown in Table 2 below.

[0131]

[0132] Table 2

[0133] As shown in Table 2, the false negative rate of the detection method of the embodiment of the present invention is 0.46%, and the average accuracy rate (95.86%) and the false positive rate (7.82%) are both optimal.

[0134] Compared with the traditional method based on a combination of features such as color, motion, and shape, the embodiment of the present invention has the following improvements:

[0135] First is the false positive rate, which is the misjudgment rate. In outdoor environments and under strong light interference, the multi-feature fusion algorithm based on video image flame features has a high false positive rate, and will mistakenly judge strong light sources and flashing car lights as flames, because the two-way car lights on the road will flash, forming a bright area similar to a flame. However, the multi-feature fusion algorithm based on the visual blurring of flames in the present invention can accurately detect physical interference and eliminate it; second is the false negative rate, which is the missed detection rate. Although the latest visual image flame detection has a low false negative rate, it will miss the white flame when magnesium phosphorus burns and the green flame when copper burns, and it will miss the flame based only on the color model and dispersion component. However, the flame visual blurring of the present invention is not affected by color and motion conditions, and can detect blurred flames produced by various fuels, further reducing the false negative rate, thereby greatly improving the accuracy of fire detection.

[0136] The above describes the method steps of the embodiment of the present invention. It is understandable that the embodiment of the present invention uses the RGB-HIS color segmentation model to extract the suspected flame area, and at the same time captures the visual blurring characteristics of the flame through binocular ranging and laser ranging, and combines the dispersion characteristics, similarity characteristics and centroid motion characteristics of the flame through the MES multi-expert decision system to perform feature fusion and decision analysis to obtain the fire detection results. On the one hand, it can accurately eliminate interference such as light bulbs, reflections, and red fire extinguishers, reducing the false positive rate of fire detection. On the other hand, it can detect flames generated by various fuels, reducing the missed detection rate of fire detection, thereby improving the accuracy of fire detection.

[0137] Reference Figure 6 The embodiment of the present invention provides a fire detection system based on flame visual blurring, comprising:

[0138] A color segmentation module is used to obtain a first video image of the area to be detected, perform color segmentation on the first video image, and obtain a suspected flame area;

[0139] The first feature determination module is used to determine the dispersion feature, similarity feature and centroid motion feature of the suspected flame area;

[0140] The second feature determination module is used to extract the flame visual blurring characteristics of the suspected flame area through binocular ranging and laser ranging;

[0141] The feature fusion module is used to build an MES multi-expert decision system. The MES multi-expert decision system performs feature fusion and decision analysis on dispersion features, similarity features, center of mass motion features, and flame visual blurring features to obtain fire detection results.

[0142] The contents of the above method embodiments are all applicable to the present system embodiments. The functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0143] Reference Figure 7 The embodiment of the present invention provides a fire detection device based on flame visual blurring, comprising:

[0144] at least one processor;

[0145] at least one memory for storing at least one program;

[0146] When the at least one program is executed by the at least one processor, the at least one processor implements the fire detection method based on flame visual blurring.

[0147] The contents of the above method embodiments are all applicable to the present device embodiments. The functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0148] An embodiment of the present invention further provides a computer-readable storage medium storing a program executable by a processor. When the program is executed by the processor, it is used to perform the above-mentioned fire detection method based on flame visual blurring.

[0149] A computer-readable storage medium according to an embodiment of the present invention can execute a fire detection method based on flame visual blurring provided by an embodiment of the present invention, can execute any combination of implementation steps of the method embodiment, and has the corresponding functions and beneficial effects of the method.

[0150] The embodiment of the present invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs Figure 1 The method shown.

[0151] In some optional embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the above-mentioned boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided in an exemplary manner for the purpose of providing a more comprehensive understanding of the technology. The disclosed method is not limited to the operation and logic flow presented herein. Optional embodiments are contemplated in which the order of the various operations is changed and the sub-operations described as a part of a larger operation are performed independently.

[0152] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the above-mentioned functions and / or features can be integrated into a single physical device and / or software module, or one or more functions and / or features can be implemented in separate physical devices or software modules. It is also understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present invention. More specifically, given the properties, functions, and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the module will be understood within the routine skills of an engineer. Therefore, a person skilled in the art can implement the present invention set forth in the claims using ordinary skills without undue experimentation. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0153] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the above methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0154] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0155] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable media on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0156] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0157] In the above description of this specification, reference to the terms "one embodiment / example," "another embodiment / example," or "certain embodiments / examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0158] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

[0159] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A fire detection method based on flame visual blurring, characterized in that: The following steps are involved: Acquire a first video image of the area to be detected, perform color segmentation on the first video image, and obtain a suspected flame area; Determining the dispersion characteristics, similarity characteristics, and center of mass motion characteristics of the suspected flame area; Extracting the flame visual blurring characteristics of the suspected flame area by binocular ranging and laser ranging; Constructing an MES multi-expert decision system, and performing feature fusion and decision analysis on the dispersion feature, the similarity feature, the center of mass motion feature, and the flame visual blurring feature through the MES multi-expert decision system to obtain a fire detection result; The step of extracting the flame visual blurring feature of the suspected flame area by binocular ranging and laser ranging specifically includes: Performing binocular ranging on the suspected flame area using a binocular camera to obtain first flame depth information of the suspected flame area; performing laser ranging on the suspected flame area by a laser measurement system to obtain second flame depth information of the suspected flame area; The flame visual blurring feature of the suspected flame area is determined according to the difference between the first flame depth information and the second flame depth information.

2. A fire detection method based on flame visual blurring according to claim 1, characterized in that: The step of performing color segmentation on the first video image to obtain a suspected flame area specifically includes: Determine the red component threshold and saturation threshold; Constructing an RGB-HIS color segmentation model according to the red component threshold and the saturation threshold; Color segmentation is performed on the first video image using the RGB-HIS color segmentation model to obtain a suspected flame area.

3. The fire detection method based on flame visual blurring according to claim 2, characterized in that: The RGB-HIS color segmentation model is: R>G>B R>R T S>(255-R)×S T / R T Among them, R represents the red component of the target pixel, G represents the green component of the target pixel, B represents the blue component of the target pixel, and R T represents the red component threshold, S represents the saturation of the target pixel, S T Indicates the saturation threshold.

4. The fire detection method based on flame visual blurring according to claim 1, characterized in that: The step of determining the dispersion feature, similarity feature, and center of mass motion feature of the suspected flame area specifically includes: performing image analysis on the suspected flame area to extract dispersion features of each part of the suspected flame area; Comparing the suspected flame regions in consecutive frames of the first video image to obtain similarity features of the suspected flame regions; The center of mass position of the suspected flame area is determined, and the center of mass motion feature of the first video image is determined based on the center of mass position.

5. A fire detection method based on flame visual blurring according to any one of claims 1 to 4, characterized in that: The step of building the MES multi-expert decision-making system specifically includes: Obtain pre-trained dispersion classifier, similarity classifier, centroid motion classifier, and flame visual blurring classifier; Determining a first weight of the dispersion classifier, a second weight of the similarity classifier, a third weight of the centroid motion classifier, and a fourth weight of the flame visual blurring classifier; An MES multi-expert decision system is constructed according to the dispersion classifier, the similarity classifier, the centroid motion classifier, the flame visual blurring classifier, the first weight, the second weight, the third weight and the fourth weight.

6. The fire detection method based on flame visual blurring according to claim 5, characterized in that: The step of performing feature fusion and decision analysis on the dispersion feature, the similarity feature, the center of mass motion feature, and the flame visual blur feature by the MES multi-expert decision system to obtain a fire detection result specifically includes: The dispersion feature, the similarity feature, the centroid motion feature, and the flame visual blur feature are respectively classified according to the dispersion classifier, the similarity classifier, the centroid motion classifier, and the flame visual blur feature to obtain a plurality of flame classification labels; Feature fusion is performed based on the first weight, the second weight, the third weight, the fourth weight, and the flame classification label to obtain a first weighted sum, and then whether a fire occurs in the suspected flame area is determined based on the first weighted sum and a preset threshold.

7. A fire detection system based on flame visual blurring, characterized in that: include: A color segmentation module is used to obtain a first video image of the area to be detected, perform color segmentation on the first video image, and obtain a suspected flame area; A first feature determination module is used to determine the dispersion feature, similarity feature and centroid motion feature of the suspected flame area; A second feature determination module is used to extract the flame visual blurring feature of the suspected flame area through binocular ranging and laser ranging; A feature fusion module is used to build an MES multi-expert decision system, and the MES multi-expert decision system performs feature fusion and decision analysis on the dispersion feature, the similarity feature, the center of mass motion feature, and the flame visual blur feature to obtain a fire detection result; The second feature determination module is specifically configured to: Performing binocular ranging on the suspected flame area using a binocular camera to obtain first flame depth information of the suspected flame area; performing laser ranging on the suspected flame area by a laser measurement system to obtain second flame depth information of the suspected flame area; The flame visual blurring feature of the suspected flame area is determined according to the difference between the first flame depth information and the second flame depth information.

8. A fire detection device based on flame visual blurring, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the fire detection method based on flame visual blurring according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to execute the fire detection method based on flame visual blurring as claimed in any one of claims 1 to 6 when executed by the processor.

Images