Hearth flame detection device and detection method

Abstract

The invention relates to the field of hearth flame detection, and particularly discloses a hearth flame detection device and method, and the method comprises the following steps: a, capturing a flame image in a hearth through a high-resolution infrared thermal imager, and transmitting the flame image to an image processing unit; step b, executing the following operations in the image processing unit to evaluate and optimize the image quality: removing noise in the image by applying an adaptive median filtering algorithm, wherein the algorithm effect is quantified by calculating a flame image noise removal effect evaluation value Enoise (l); the flame state parameters are monitored in real time, once the parameters are lower than the preset threshold value, the safety pre-warning system is triggered immediately, the real-time monitoring and pre-warning mechanism helps to find and handle potential problems such as instability or extinguishment of flames in time, and therefore safety accidents are effectively prevented.

Description

Technical Field

[0001] The invention belongs to the field of furnace flame detection, in particular to a furnace flame detection device and a detection method. Background Art

[0002] The furnace is the core part of the boiler equipment, which is used to contain fuel and carry out combustion reactions. Flame is the bright and hot phenomenon produced when the fuel burns in the furnace, which accompanies the occurrence and continuation of the combustion process. The generation of furnace flame is closely related to the type of fuel, the design of the burner, the structure of the furnace and various parameters in the combustion process.

[0003] Traditional furnace flame detection technologies usually lack the characteristics of intelligence and automation. They can only provide simple monitoring data, but cannot conduct in-depth analysis and prediction of flame status. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a furnace flame detection method to solve the problem that the conventional furnace flame detection technology in the prior art usually lacks the characteristics of intelligence and automation. They can often only provide simple monitoring data, but cannot perform in-depth analysis and prediction of the flame state.

[0005] A furnace flame detection method comprises the following steps:

[0006] Step a: Capturing the flame image in the furnace through a high-resolution infrared thermal imager and transmitting the flame image to an image processing unit;

[0007] Step b: Perform the following operations in the image processing unit to evaluate and optimize image quality:

[0008] The adaptive median filter algorithm is used to remove the noise in the image, where the algorithm effect is evaluated by calculating the flame image noise removal effect evaluation value E noise (1) to quantify;

[0009] The histogram equalization technique is used to enhance the contrast of the flame area, and the contrast enhancement effect evaluation value E is used to evaluate the effect of the contrast enhancement. contrast (Ι) to measure the image quality after equalization;

[0010] Step c: In the preprocessed image, the edge contour of the flame is detected using the Hough transform algorithm, and the color features of the flame are extracted based on the color space transformation, including hue, saturation and brightness; the comprehensive effect of feature extraction is evaluated by the flame feature extraction evaluation value E feature (1) to evaluate;

[0011] Step d: inputting the extracted features into a trained support vector machine (SVM) classifier to identify the flame; the SVM classifier is trained with a large number of known flame and non-flame image samples, and the training process includes data preprocessing, feature extraction, parameter selection and model training;

[0012] Step e: According to the recognition results, the geometric parameters and dynamic parameters of the flame are calculated to evaluate the stability and intensity of the flame;

[0013] Step f: If the flame state parameter is lower than the preset threshold, the safety warning system is triggered to maintain the stable combustion of the flame in the furnace through sound and light alarm and automatic adjustment of the fuel supply; the safety warning system also includes real-time recording of the change history of the flame state parameters, automatically recording the image and parameter information at the abnormal moment, and providing a remote monitoring interface;

[0014] Step g: Based on big data analysis and machine learning technology, analyze the changing trends of flame state parameters in real time, predict potential fault risks, and provide targeted fault diagnosis suggestions.

[0015] Preferably, the adaptive median filtering algorithm in step b includes determining the size of the filtering window, and dynamically adjusting the window size according to the difference between the median value of the pixel value in the window and the pixel value at the center of the window until the condition is met; the flame image noise removal effect evaluation value E noise The calculation of (I) takes into account the difference in pixel values ​​before and after filtering and the influence of the noise standard deviation on the filtering effect;

[0016] The evaluation formula for flame image noise removal effect is as follows:

[0017]

[0018] in:

[0019] E noise (I) is the noise removal effect evaluation value;

[0020] Ι filtered (i) is the pixel value of the image after filtering;

[0021] Ι original is the original image pixel value;

[0022] N is the total number of image pixels;

[0023] median(Ι window (i)) is the median value of the pixel value in the window;

[0024] σ is the noise standard deviation, which is selected according to the actual situation;

[0025] Ι window(i) is the set of pixel values ​​within the filter window centered at pixel i;

[0026] E noise The smaller the value of (Ι), the better the noise removal effect is, and its value range is [0, +∞), where 0 means complete noise removal and a larger value means poor noise removal effect.

[0027] Preferably, the histogram equalization technique in step b is performed by evaluating the contrast enhancement effect value E contrast (I) is used to measure the image quality after equalization, which takes into account the statistical characteristics of the pixel values ​​in the flame area;

[0028] The evaluation formula for the contrast enhancement effect in the flame area is as follows:

[0029]

[0030] in:

[0031] E contrast (I) is the contrast enhancement effect evaluation value;

[0032] Ι ed (j) is the pixel value of the image after equalization;

[0033] M is the total number of pixels in the flame area;

[0034] mean(Ι ed (j)) and std(Ι ed (j)) are the mean and standard deviation of the flame area after equalization;

[0035] CDF(I ed (j)) is the cumulative distribution function value of pixel value j after equalization;

[0036] CDF max is the maximum value of the cumulative distribution function;

[0037] E contrast The larger the value of (Ι), the better the contrast enhancement effect is, and its value range is [0, +∞), where 0 means no contrast enhancement, and a larger value means a significant contrast enhancement effect.

[0038] Preferably, the comprehensive effect of the feature extraction in step c is calculated by the flame feature extraction evaluation value E feature (Ι) is used to evaluate, which combines the smoothness of edge contours and the accuracy of color features;

[0039] The comprehensive evaluation formula for flame feature extraction is as follows:

[0040]

[0041] in:

[0042] E feature (1) extracting a comprehensive evaluation value for flame characteristics;

[0043] α and β are weight coefficients, which are adjusted according to actual conditions;

[0044] P is the number of points on the edge contour;

[0045] E(k) is the coordinate of point k on the edge contour;

[0046] θ is the tangent direction of the edge profile;

[0047] C l is the color feature value, l = 1, 2, 3 correspond to hue, saturation and brightness respectively;

[0048] is the reference color feature value;

[0049] mean(C l ) and std(C l ) are the mean and standard deviation of the color feature values ​​respectively;

[0050] E feature The smaller the value of (Ι), the better the flame feature extraction effect is, and its value range is [0, +∞), where 0 means that the feature extraction is completely accurate, and a larger value means that there is an error in the feature extraction.

[0051] Preferably, the support vector machine (SVM) classifier training process in step d includes collecting a training data set, preprocessing the training data set, extracting image features, selecting optimal SVM parameters using cross-validation, and training the SVM model until a predetermined classification accuracy is achieved.

[0052] Preferably, the safety warning system in step f also includes real-time recording of the change history of flame state parameters for trend analysis; in the event of flame abnormality, automatically recording images and parameter information at the time of abnormality to facilitate subsequent troubleshooting; and providing a remote monitoring interface to allow operators to view the flame status and receive warning information in real time at a location away from the furnace.

[0053] Preferably, the intelligent early warning and fault diagnosis in step g also includes using big data technology to deeply mine flame state parameters and establish a fault prediction model; based on the output results of the fault prediction model, providing targeted fault diagnosis suggestions, including possible fault causes, impact range and repair suggestions; and automatically recording and analyzing images and parameter information at the moment of flame abnormality to assist subsequent fault troubleshooting.

[0054] A furnace flame detection device, comprising:

[0055] A high-resolution infrared thermal imager to capture images of flames inside the furnace;

[0056] An image processing unit, connected to the high-resolution infrared thermal imager, for receiving and processing flame images, and performing image processing operations such as adaptive median filtering, histogram equalization, edge contour detection, and color feature extraction;

[0057] A support vector machine (SVM) classifier, connected to the image processing unit, is used to receive processed image features and identify flames;

[0058] A parameter calculation module, connected to the SVM classifier, for calculating the geometric parameters and dynamic parameters of the flame according to the recognition result;

[0059] A safety warning system, connected to the parameter calculation module, is used to trigger a safety warning according to the flame state parameters, including sound and light alarms and automatic adjustment of the fuel supply, and records the change history of the flame state parameters, images and parameter information at abnormal times, and provides a remote monitoring interface;

[0060] The big data analysis and fault diagnosis module is connected to the safety warning system to analyze the changing trend of flame state parameters in real time, predict potential fault risks, and provide targeted fault diagnosis suggestions.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] The flame image is captured by a high-resolution infrared thermal imager, and combined with image processing technologies such as adaptive median filtering and histogram equalization, the quality of the flame image is effectively improved, making the edge contour and color characteristics of the flame clearer and more accurate. This provides a solid foundation for subsequent flame recognition and ensures high recognition accuracy;

[0063] By monitoring flame status parameters in real time, once the parameters are lower than the preset threshold, the safety warning system is immediately triggered. This real-time monitoring and warning mechanism helps to promptly detect and deal with potential flame instability or extinction problems, thereby effectively preventing the occurrence of safety accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] like Figure 1 As shown:

[0067] Embodiment 1: The present invention provides a furnace flame detection method, comprising the following steps:

[0068] Step a: Capturing the flame image in the furnace through a high-resolution infrared thermal imager and transmitting the flame image to an image processing unit;

[0069] Step b: Perform the following operations in the image processing unit to evaluate and optimize image quality:

[0070] The adaptive median filter algorithm is used to remove the noise in the image, where the algorithm effect is evaluated by calculating the flame image noise removal effect evaluation value E noise (1) to quantify;

[0071] The histogram equalization technique is used to enhance the contrast of the flame area, and the contrast enhancement effect evaluation value E is used to evaluate the effect of the contrast enhancement. contrast (Ι) to measure the image quality after equalization;

[0072] Step c: In the preprocessed image, the edge contour of the flame is detected using the Hough transform algorithm, and the color features of the flame are extracted based on the color space transformation, including hue, saturation and brightness; the comprehensive effect of feature extraction is evaluated by the flame feature extraction evaluation value E feature (1) to evaluate;

[0073] Step d: Input the extracted features into a trained support vector machine (SVM) classifier to identify the flame; the SVM classifier is trained with a large number of known flame and non-flame image samples, and the training process includes data preprocessing, feature extraction, parameter selection and model training;

[0074] Step e: According to the recognition results, the geometric parameters and dynamic parameters of the flame are calculated to evaluate the stability and intensity of the flame;

[0075] Step f: If the flame state parameter is lower than the preset threshold, the safety warning system is triggered to maintain the stable combustion of the flame in the furnace through sound and light alarm and automatic adjustment of the fuel supply; the safety warning system also includes real-time recording of the change history of the flame state parameters, automatically recording the image and parameter information at the abnormal moment, and providing a remote monitoring interface;

[0076] Step g: Based on big data analysis and machine learning technology, analyze the changing trends of flame state parameters in real time, predict potential fault risks, and provide targeted fault diagnosis suggestions.

[0077] As can be seen from the above, this method first uses a high-resolution infrared thermal imager to capture the flame image in the furnace and transmits it to the image processing unit for quality optimization. Through adaptive median filtering and histogram equalization technology, the image noise is effectively removed and the contrast of the flame area is enhanced. Subsequently, the edge contour and color features of the flame are extracted using the Hough transform algorithm and color space transformation, and the accuracy of feature extraction is ensured through quantitative evaluation. These features are input into the trained SVM classifier to achieve accurate recognition of the flame. According to the recognition results, the geometric and dynamic parameters of the flame are calculated, and its stability and intensity are evaluated. Once the flame state parameters are lower than the preset threshold, the safety warning system is immediately triggered and corresponding measures are taken to keep the flame stable. In addition, this method also uses big data analysis and machine learning technology to analyze the changing trend of flame state parameters in real time, predict potential faults, provide fault diagnosis suggestions, realize comprehensive flame monitoring and management, and ensure the safe operation of the furnace.

[0078] Embodiment 2: This embodiment is basically the same as the previous embodiment, except that it also includes: the adaptive median filtering algorithm in step b includes determining the size of the filtering window, and dynamically adjusting the window size according to the difference between the median value of the pixel value in the window and the pixel value at the center of the window until the condition is met; the flame image noise removal effect evaluation value E noise The calculation of (I) takes into account the difference in pixel values ​​before and after filtering and the influence of the noise standard deviation on the filtering effect;

[0079] The evaluation formula for flame image noise removal effect is as follows:

[0080]

[0081] in:

[0082] E noise (I) is the noise removal effect evaluation value;

[0083] Ι filtered (i) is the pixel value of the image after filtering;

[0084] Ι original is the original image pixel value;

[0085] N is the total number of image pixels;

[0086] median(Ι window (i)) is the median value of the pixel value in the window;

[0087] σ is the noise standard deviation, which is selected according to the actual situation;

[0088] Ι window (i) is the set of pixel values ​​within the filter window centered at pixel i;

[0089] E noise The smaller the value of (Ι), the better the noise removal effect is, and its value range is [0, +∞), where 0 means complete noise removal, and a larger value means poor noise removal effect;

[0090] The specific application process of the above formula is as follows:

[0091] Data preparation:

[0092] Get the original flame image data and record the original value of each pixel. original .

[0093] The image is filtered using an adaptive median filter algorithm to obtain filtered image data and record the filter value of each pixel. filtered (i).

[0094] Window Selection:

[0095] For each pixel i in the image, select a filter window Ι centered at that pixel window (i).

[0096] Calculate the median of the pixel values ​​in the window window (i)).

[0097] Noise standard deviation estimate:

[0098] According to actual conditions or prior knowledge, estimate the noise standard deviation σ in the image.

[0099] Calculation formula:

[0100] Use Formula E noise (Ι) Calculate the noise removal effect evaluation value.

[0101] Result analysis:

[0102] The E obtained by analysis noise (Ι) value, the smaller the value, the better the noise removal effect.

[0103] According to the value range [0,+∞), it is judged whether the noise removal effect reaches the expected result.

[0104] Specifically, the histogram equalization technique in step b is evaluated by the contrast enhancement effect value E contrast (I) is used to measure the image quality after equalization, which takes into account the statistical characteristics of the pixel values ​​in the flame area;

[0105] The evaluation formula for the contrast enhancement effect in the flame area is as follows:

[0106]

[0107] in:

[0108] E contrast (I) is the contrast enhancement effect evaluation value;

[0109] Ι ed (j) is the pixel value of the image after equalization;

[0110] M is the total number of pixels in the flame area;

[0111] mean(Ι ed (j)) and std(Ι ed (j)) are the mean and standard deviation of the flame area after equalization;

[0112] CDF(I ed (j)) is the cumulative distribution function value of pixel value j after equalization;

[0113] CDF max is the maximum value of the cumulative distribution function;

[0114] E contrast The larger the value of (Ι), the better the contrast enhancement effect is. Its value range is [0, +∞), where 0 means no contrast enhancement, and a larger value means a significant contrast enhancement effect.

[0115] The specific application process of the above formula is as follows:

[0116] Flame region segmentation:

[0117] Use an appropriate image segmentation algorithm (such as threshold segmentation, region growing, etc.) to segment the flame area from the image.

[0118] Histogram Equalization:

[0119] The histogram equalization technique is applied to the segmented flame area to obtain the equalized image data. ed (j).

[0120] Statistical calculations:

[0121] Calculate the mean value of the flame area after equalization. ed (j)) and standard deviation std(Ι ed (j)).

[0122] Calculate the cumulative distribution function value CDF(Ι of the pixel value j after equalization) ed(j)) and find the maximum value of the cumulative distribution function CDF max .

[0123] Calculation formula:

[0124] Use Formula E contrast (I) Calculate the contrast enhancement effect evaluation value.

[0125] Result analysis:

[0126] The E obtained by analysis contrast (Ι) value, the larger the value, the better the contrast enhancement effect.

[0127] According to the value range [0, +∞), it is judged whether the contrast enhancement effect reaches the expected value.

[0128] Specifically, the comprehensive effect of feature extraction in step c is evaluated by the flame feature extraction evaluation value E feature (Ι) is used to evaluate, which combines the smoothness of edge contours and the accuracy of color features;

[0129] The comprehensive evaluation formula for flame feature extraction is as follows:

[0130]

[0131] in:

[0132] E feature (1) extracting a comprehensive evaluation value for flame characteristics;

[0133] α and β are weight coefficients, which are adjusted according to actual conditions;

[0134] P is the number of points on the edge contour;

[0135] E(k) is the coordinate of point k on the edge contour;

[0136] θ is the tangent direction of the edge profile;

[0137] C l is the color feature value, l = 1, 2, 3 correspond to hue, saturation and brightness respectively;

[0138] is the reference color feature value;

[0139] mean(C l ) and std(C l ) are the mean and standard deviation of the color feature values ​​respectively;

[0140] E feature The smaller the value of (Ι), the better the flame feature extraction effect is, and its value range is [0, +∞), where 0 means that the feature extraction is completely accurate, and a larger value means that there is an error in the feature extraction;

[0141] The specific application process of the above formula is as follows:

[0142] Edge contour detection:

[0143] The improved Hough transform algorithm is used to detect the edge contour of the flame in the preprocessed image, and the number of points P on the edge contour and the coordinates E(k) of each point k are recorded.

[0144] Calculate the tangent direction θ of the edge contour.

[0145] Color feature extraction:

[0146] Convert the image from RGB color space to HSV color space and extract the hue, saturation and brightness feature values ​​of the flame C l (l=1,2,3).

[0147] Determine the reference color characteristic value C ref , which can be the median or average value of the flame characteristic value range obtained based on prior knowledge or experimental data.

[0148] Statistical calculations:

[0149] Calculate the mean value of the color feature value mean(C l ) and standard deviation std(C l ).

[0150] Calculation formula:

[0151] Use Formula E feature (Ι) Calculate the comprehensive evaluation value of flame feature extraction.

[0152] Note: Here It represents the rate of change of the tangent direction at point k on the edge contour, which can be obtained by calculating the angle between the tangent directions of adjacent points.

[0153] Result analysis:

[0154] The E obtained by analysis feature (Ι) value, the smaller the value, the better the flame feature extraction effect.

[0155] According to the value range [0, +∞), determine whether the feature extraction is accurate, and adjust the algorithm parameters or improve the algorithm accordingly.

[0156] As can be seen from the above, in step b of this method, the adaptive median filtering algorithm optimizes the noise removal effect by dynamically adjusting the filter window size, and uses the noise removal effect evaluation formula to quantify the filtering performance. The formula comprehensively considers the pixel value difference before and after filtering and the noise standard deviation to ensure that the noise is effectively and appropriately removed. In addition, the histogram equalization technology measures the image quality improvement through the contrast enhancement effect evaluation value. The evaluation value incorporates the statistical characteristics of the pixel values ​​in the flame area, making the contrast enhancement effect more significant and controllable. In step c, the comprehensive effect of feature extraction is comprehensively evaluated through the flame feature extraction evaluation value. The evaluation value combines the smoothness of the edge contour and the accuracy of the color feature to ensure the accuracy and reliability of feature extraction. These improvement measures jointly improve the accuracy and stability of furnace flame detection, and provide a more solid foundation for subsequent flame identification, parameter calculation and safety warning.

[0157] Embodiment 3: This embodiment is basically the same as the previous embodiment, except that it also includes the support vector machine (SVM) classifier training process in step d, including collecting training data sets, preprocessing the training data sets, extracting image features, using cross-validation to select the best SVM parameters, and training the SVM model until a predetermined classification accuracy is reached.

[0158] Specifically, the safety warning system in step f also includes real-time recording of the change history of flame state parameters for trend analysis; in the case of flame abnormality, automatically recording the image and parameter information at the abnormal moment to facilitate subsequent troubleshooting; and providing a remote monitoring interface, allowing operators to view the flame status and receive warning information in real time at a location away from the furnace.

[0159] Specifically, the intelligent early warning and fault diagnosis in step g also includes using big data technology to deeply mine flame state parameters and establish a fault prediction model; based on the output results of the fault prediction model, providing targeted fault diagnosis suggestions, including possible fault causes, impact range and repair suggestions; and automatically recording and analyzing images and parameter information at the moment of flame abnormality to assist in subsequent fault troubleshooting.

[0160] It can be seen from the above that this embodiment further refines the training process of the SVM classifier to ensure that the classifier can achieve the predetermined classification accuracy. At the same time, the safety warning system has been enhanced, which can not only record the change history of flame state parameters in real time to facilitate trend analysis, but also automatically record key information when the flame is abnormal, providing strong support for troubleshooting. In addition, by providing a remote monitoring interface, operators can grasp the furnace flame status anytime and anywhere and receive warning information in time. In terms of intelligent warning and fault diagnosis, this embodiment uses big data technology to deeply mine flame state parameters and establish a fault prediction model, which can detect potential faults in advance and provide targeted repair suggestions, thereby effectively improving the intelligence level of furnace flame monitoring and fault response capabilities.

[0161] Embodiment 4: This embodiment is basically the same as the previous embodiment, except that it further comprises a furnace flame detection device, including:

[0162] A high-resolution infrared thermal imager to capture images of flames inside the furnace;

[0163] An image processing unit, connected to the high-resolution infrared thermal imager, is used to receive and process the flame image and perform image processing operations such as adaptive median filtering, histogram equalization, edge contour detection, and color feature extraction;

[0164] A support vector machine (SVM) classifier is connected to the image processing unit and is used to receive processed image features and identify flames;

[0165] A parameter calculation module, connected to the SVM classifier, is used to calculate the geometric parameters and dynamic parameters of the flame according to the recognition results;

[0166] The safety warning system is connected to the parameter calculation module and is used to trigger safety warnings according to the flame state parameters, including sound and light alarms and automatic adjustment of fuel supply, and records the change history of flame state parameters, images and parameter information at abnormal times, and provides a remote monitoring interface;

[0167] The big data analysis and fault diagnosis module is connected to the safety warning system to analyze the changing trends of flame state parameters in real time, predict potential fault risks, and provide targeted fault diagnosis suggestions.

[0168] As can be seen from the above, this device integrates a high-resolution infrared thermal imager, an image processing unit, an SVM classifier, a parameter calculation module, a safety warning system, and a big data analysis and fault diagnosis module. The high-resolution infrared thermal imager is responsible for capturing the flame image in the furnace. The image processing unit then processes the image through adaptive median filtering, histogram equalization, edge contour detection, and color feature extraction. The SVM classifier receives the processed image features and accurately identifies the flame. The parameter calculation module calculates the geometric and dynamic parameters of the flame based on the recognition results. The safety warning system triggers warnings based on these parameters, including sound and light alarms, adjusting fuel supply, and recording relevant information. Finally, the big data analysis and fault diagnosis module analyzes the flame state in real time, predicts fault risks, and provides targeted fault diagnosis suggestions. The design of this device has greatly improved the intelligence and safety of furnace flame detection.

[0169] The standard parts used in the present invention can all be purchased from the market, and the special-shaped parts can be customized according to the description and the drawings. The specific connection methods of each part adopt conventional means such as mature bolts, rivets, welding, etc. in the prior art. The machinery, parts and equipment all adopt conventional models in the prior art, and the circuit connection adopts the conventional connection method in the prior art, which will not be described in detail here. The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field.

[0170] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. "Multiple" means two or more, unless otherwise clearly and specifically defined.

[0171] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0172] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may mean that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. Moreover, a first feature being "above", "above" or "above" a second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. A first feature being "below", "below" or "below" a second feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is lower in level than the second feature.

[0173] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" etc. 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 representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0174] In the drawings of the embodiments disclosed in the present invention, only the structures related to the embodiments disclosed in the present invention are involved, and other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of the present invention can be combined with each other.

[0175] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A furnace flame detection method, characterized in that: The following steps are involved: Step a: Capturing the flame image in the furnace through a high-resolution infrared thermal imager and transmitting the flame image to an image processing unit; Step b: Perform the following operations in the image processing unit to evaluate and optimize image quality: The adaptive median filter algorithm is used to remove the noise in the image, where the algorithm effect is evaluated by calculating the flame image noise removal effect evaluation value E noise (1) to quantify; The histogram equalization technique is used to enhance the contrast of the flame area, and the contrast enhancement effect evaluation value E is used to evaluate the effect of the contrast enhancement. contrast (Ι) to measure the image quality after equalization; Step c: In the preprocessed image, the edge contour of the flame is detected using the Hough transform algorithm, and the color features of the flame are extracted based on the color space transformation, including hue, saturation and brightness; the comprehensive effect of feature extraction is evaluated by the flame feature extraction evaluation value E feature (1) to evaluate; Step d: inputting the extracted features into a trained support vector machine (SVM) classifier to identify the flame; the SVM classifier is trained with a large number of known flame and non-flame image samples, and the training process includes data preprocessing, feature extraction, parameter selection and model training; Step e: According to the recognition results, the geometric parameters and dynamic parameters of the flame are calculated to evaluate the stability and intensity of the flame; Step f: If the flame state parameter is lower than the preset threshold, the safety warning system is triggered to maintain the stable combustion of the flame in the furnace through sound and light alarm and automatic adjustment of the fuel supply; the safety warning system also includes real-time recording of the change history of the flame state parameters, automatically recording the image and parameter information at the abnormal moment, and providing a remote monitoring interface; Step g: Based on big data analysis and machine learning technology, analyze the changing trends of flame state parameters in real time, predict potential fault risks, and provide targeted fault diagnosis suggestions.

2. A furnace flame detection method as claimed in claim 1, characterized in that: The adaptive median filtering algorithm in step b includes determining the size of the filtering window and dynamically adjusting the window size according to the difference between the median value of the pixel value in the window and the pixel value at the center of the window until the condition is met; the flame image noise removal effect evaluation value E noise The calculation of (I) takes into account the difference in pixel values ​​before and after filtering and the influence of the noise standard deviation on the filtering effect; The evaluation formula for flame image noise removal effect is as follows: in: E noise (I) is the noise removal effect evaluation value; Ι filtered (i) is the pixel value of the image after filtering; Ι original is the original image pixel value; N is the total number of image pixels; median(Ι window (i)) is the median value of the pixel value in the window; σ is the noise standard deviation, which is selected according to the actual situation; Ι window (i) is the set of pixel values ​​within the filter window centered at pixel i; E noise The smaller the value of (Ι), the better the noise removal effect is, and its value range is [0, +∞), where 0 means complete noise removal, and a larger value means poor noise removal effect.

3. A furnace flame detection method as claimed in claim 2, characterized in that: The histogram equalization technique in step b evaluates the contrast enhancement effect E contrast (I) is used to measure the image quality after equalization, which takes into account the statistical characteristics of the pixel values ​​in the flame area; The evaluation formula for the contrast enhancement effect in the flame area is as follows: in: E contrast (I) is the contrast enhancement effect evaluation value; Ι ed (j) is the pixel value of the image after equalization; M is the total number of pixels in the flame area; mean(Ι ed (j)) and std(Ι ed (j)) are the mean and standard deviation of the flame area after equalization; CDF(I ed (j)) is the cumulative distribution function value of pixel value j after equalization; CDF max is the maximum value of the cumulative distribution function; E contrast The larger the value of (Ι), the better the contrast enhancement effect is, and its value range is [0, +∞), where 0 means no contrast enhancement, and a larger value means a significant contrast enhancement effect.

4. A furnace flame detection method as claimed in claim 3, characterized in that: The comprehensive effect of feature extraction in step c is evaluated by the flame feature extraction evaluation value E feature (Ι) is used to evaluate, which combines the smoothness of edge contours and the accuracy of color features; The comprehensive evaluation formula for flame feature extraction is as follows: in: E feature (1) extracting a comprehensive evaluation value for flame characteristics; α and β are weight coefficients, which are adjusted according to actual conditions; P is the number of points on the edge contour; E(k) is the coordinate of point k on the edge contour; θ is the tangent direction of the edge profile; C l is the color feature value, l = 1, 2, 3 correspond to hue, saturation and brightness respectively; is the reference color feature value; mean(C l ) and std(C l ) are the mean and standard deviation of the color feature values ​​respectively; E feature The smaller the value of (Ι), the better the flame feature extraction effect is, and its value range is [0, +∞), where 0 means that the feature extraction is completely accurate, and a larger value means that there is an error in the feature extraction.

5. A furnace flame detection method as claimed in claim 4, characterized in that: The support vector machine (SVM) classifier training process in step d includes collecting a training data set, preprocessing the training data set, extracting image features, selecting optimal SVM parameters using cross-validation, and training the SVM model until a predetermined classification accuracy is achieved.

6. A furnace flame detection method as claimed in claim 1, characterized in that: The safety warning system in step f also includes real-time recording of the change history of flame state parameters for trend analysis; in the event of flame abnormality, automatically recording images and parameter information at the time of abnormality to facilitate subsequent troubleshooting; and providing a remote monitoring interface to allow operators to view the flame status and receive warning information in real time at a location away from the furnace.

7. A furnace flame detection method as claimed in claim 1, characterized in that: The intelligent early warning and fault diagnosis in step g also includes using big data technology to deeply mine flame state parameters and establish a fault prediction model; based on the output results of the fault prediction model, provide targeted fault diagnosis suggestions, including possible fault causes, impact range and repair suggestions; and automatically record and analyze images and parameter information at the moment of flame abnormality to assist in subsequent fault troubleshooting.

8. A furnace flame detection device, characterized in that: include: A high-resolution infrared thermal imager to capture images of flames inside the furnace; An image processing unit, connected to the high-resolution infrared thermal imager, for receiving and processing flame images, and performing image processing operations such as adaptive median filtering, histogram equalization, edge contour detection, and color feature extraction; A support vector machine (SVM) classifier, connected to the image processing unit, is used to receive processed image features and identify flames; A parameter calculation module, connected to the SVM classifier, for calculating the geometric parameters and dynamic parameters of the flame according to the recognition result; A safety warning system, connected to the parameter calculation module, is used to trigger a safety warning according to the flame state parameters, including sound and light alarms and automatic adjustment of the fuel supply, and records the change history of the flame state parameters, images and parameter information at abnormal times, and provides a remote monitoring interface; The big data analysis and fault diagnosis module is connected to the safety warning system to analyze the changing trend of flame state parameters in real time, predict potential fault risks, and provide targeted fault diagnosis suggestions.

Images