A method for detecting the operating state of a decomposition furnace

By preprocessing and calculating the anomaly probability of the infrared thermal image of the decomposer, and combining cluster analysis of grayscale, gradient and spatial distance, the problem of accurate location of temperature anomaly diagnosis in the decomposer was solved, the accuracy and response speed of fault diagnosis were improved, and the quantitative assessment of the decomposer's operating status was realized.

CN118823027BActive Publication Date: 2024-11-29TONGCHUAN YAOBAI SPECIAL CEMENT CO LTD

Patent Information

Application Number
CN202411313137.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-20
Publication Date
2024-11-29
Estimated Expiration
2044-09-20

AI Technical Summary

Technical Problem

In existing technologies, the accuracy and response speed of temperature anomaly diagnosis in the operation of the decomposition furnace are low, and the fixed temperature threshold early warning mechanism cannot accurately locate the temperature anomaly area, resulting in low accuracy and response speed of fault diagnosis.

Method used

By preprocessing the infrared thermal image of the decomposition furnace, the anomaly probability of each pixel is calculated. The anomaly probability is estimated by using the grayscale difference and gradient difference of the pixels. Clustering is performed by combining the spatial distance and blur degree of the pixels to correct the anomaly probability. The entropy value of the abnormal area is quantified, and the operating status of the decomposition furnace is comprehensively evaluated.

Benefits of technology

It enables accurate location and clustering of abnormal temperature areas in the decomposer, improving the accuracy and response speed of fault diagnosis, providing a quantitative evaluation of the decomposer's operating status, and helping operators take timely maintenance measures.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118823027B_ABST
    Figure CN118823027B_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of image processing, and in particular to a method for detecting the operating state of a decomposition furnace. The method comprises: preprocessing a plurality of infrared thermal imaging images of the decomposition furnace into original images, quantifying the abnormal probability of each pixel point, determining the possibility that each pixel point and other pixel points belong to the same region according to the difference and distance between the abnormal probabilities of each pixel point and other pixel points, clustering all pixel points into a plurality of regions based on the possibility, calculating the abnormal degree of the region, calculating the abnormal probability of the region according to the abnormal degree and blur degree of the region, correcting the abnormal probability of the region by using the entropy value of the abnormal probability of the region corresponding to the region in all original images, screening the abnormal region according to the size of the corrected abnormal probability, and evaluating the operating state of the decomposition furnace according to the abnormal region. The method improves the accuracy and response speed of fault diagnosis, and can accurately locate the region with abnormal temperature.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method for detecting the operating state of a decomposition furnace. Background Art

[0002] In modern cement industry production, the decomposition furnace, as one of the key equipment in the clinker burning system, is responsible for completing important processes such as raw material preheating, fuel combustion and carbonate decomposition. Its operating status is directly related to the quality and production efficiency of the final clinker. The ideal decomposition furnace operating conditions require maintaining a uniform and stable temperature field distribution in the furnace, which not only helps to improve material conversion efficiency, but also helps to extend the service life of the equipment.

[0003] The current technical solution mainly relies on various temperature sensors arranged on the furnace body to collect temperature information in the furnace in real time, and transmit it to the central control system by wired or wireless means. The control system determines whether an abnormal situation occurs according to a preset temperature threshold, and triggers an alarm when necessary to remind the operator to take measures for maintenance. For example, the Chinese patent document with the existing authorization announcement number CN116612438B discloses a real-time monitoring system for the combustion state of a steam boiler based on thermal imaging, including: an image data acquisition module, which acquires a thermal imaging temperature image; an image preprocessing module, which obtains a temperature difference image through the thermal imaging temperature image; a real-time monitoring data calculation module, which clusters the coordinates of the pixel points in the thermal imaging temperature image and the temperature difference image and the temperature values ​​of the sample pixel points to obtain the first clustering results of the two images respectively; obtains the first importance and the second importance according to the first clustering results of the two images; obtains the effectiveness of each pixel point; and then obtains the effective temperature change data; obtains the real-time monitoring data of the combustion state of the steam boiler according to the effective temperature change data and the temperature change amount; and a real-time state monitoring module monitors the real-time combustion state of the steam boiler. The solution is to analyze the real-time monitoring data of each pixel to form a real-time monitoring image, which represents the temperature change of each pixel between the current moment and the previous moment, and then set a threshold. When it is detected that the temperature of a pixel in the image exceeds the threshold, an early warning is issued.

[0004] However, the amount of temperature data generated during the operation of the decomposition furnace is huge, and the temperature data in the temperature monitoring diagram changes complexly. Based on the fixed temperature threshold alarm mechanism, once it is detected that the temperature of a certain point in the furnace exceeds the preset range, an alarm is triggered immediately. However, this method fails to provide specific location information about the abnormal temperature area, making it difficult for maintenance personnel to quickly locate the source of the fault, resulting in low accuracy and response speed of fault diagnosis. Summary of the invention

[0005] In order to solve the problem that the fixed temperature threshold warning mechanism of the traditional solution cannot accurately locate the temperature abnormal area, resulting in low accuracy and response speed of fault diagnosis, the present invention proposes a decomposition furnace operation status detection method, comprising:

[0006] Preprocessing multiple infrared thermal images of the decomposition furnace to obtain multiple original images, and calculating the abnormal probability of each pixel point according to the grayscale difference and gradient difference between each pixel point and neighboring pixel points in the multiple original images, wherein the abnormal probability is positively correlated with the grayscale difference and the gradient difference;

[0007] Determine the possibility that each pixel point and other pixels point belong to the same region according to the difference and distance of the abnormal probability between each pixel point and other pixels point in any of the original images, and the possibility is negatively correlated with the difference and distance of the abnormal probability;

[0008] In the original image, all pixels are clustered into multiple regions based on the possibility, the mean of the abnormal probabilities of all pixels in any region is normalized as the abnormality degree of the region, and the abnormality probability of the region is calculated according to the abnormality degree of the region and the blur degree of the region, wherein the abnormality probability of the region is positively correlated with the blur degree and the abnormality degree;

[0009] The abnormal probability of the area is corrected by using the entropy value of the abnormal probability of the area corresponding to the area in all original images, and whether the area is an abnormal area is determined according to the size of the corrected abnormal probability, and the operating status of the decomposition furnace is evaluated according to the total number of all abnormal areas and the total number of pixels contained in the abnormal areas.

[0010] The above technical solution estimates the probability of a pixel becoming an anomaly by calculating the grayscale and gradient differences of the pixel, and can quantitatively evaluate the possibility of each pixel becoming an anomaly. By analyzing the abnormal probability differences and spatial distances between pixels, it is determined whether they belong to the same temperature zone, which helps to distinguish different temperature zones. Pixels with similar temperature characteristics can be effectively clustered together to form different areas, which is convenient for further analysis of the temperature anomalies in these areas. The image is divided into multiple temperature zones through a clustering algorithm, and the degree of abnormality of each zone is quantified, which provides a basis for subsequent anomaly detection. The factor of regional boundary clarity is taken into account. By introducing the degree of fuzziness, the definition of the abnormal area is further refined, making the definition of the abnormal area more accurate, and through By considering the change of abnormal probability in the same area in multiple frames of images, the abnormal probability is corrected to enhance the robustness of the system. The corrected abnormal probability can better reflect the actual temperature abnormality and reduce the false alarm rate. Finally, by counting the number and area of ​​abnormal areas, the operating status of the decomposition furnace can be comprehensively evaluated, providing a method for quantitatively evaluating the operating status of the decomposition furnace, so that the operator can understand the health status of the decomposition furnace in time and take corresponding measures. In summary, the technical scheme analyzes and identifies the temperature anomaly of the decomposition furnace from multiple angles through a series of steps. Compared with the traditional fixed threshold warning mechanism, it can more accurately locate the temperature abnormality area and evaluate the operating status of the decomposition furnace accordingly, thereby improving the accuracy of fault diagnosis and the response speed.

[0011] Furthermore, the probability that each pixel belongs to the same region as other pixels satisfies the following relationship:

[0012]

[0013] In the formula, For the Pixels and The pixel at The probability that the original images belong to the same region, For the The abnormal probability of a pixel point is For the The abnormal probability of a pixel point is For the The pixel at The horizontal coordinate in the original image, For the The pixel at The horizontal coordinate in the original image, For the The pixel at The vertical coordinate in the original image, For the The pixel at The vertical coordinate in the original image, For the The maximum value of the horizontal coordinate in the original image. For the The maximum value of the vertical coordinate in the original image. is a natural exponential function.

[0014] The above technical solution combines the abnormal probability difference and spatial position difference between pixels, performs nonlinear transformation through natural exponential function, and effectively quantifies the similarity or correlation between pixels. If the abnormal probability of two pixels is very close, then the possibility that they belong to the same area is higher. If two pixels are close to each other in space, then the possibility that they belong to the same area is also relatively high. By considering the abnormal probability difference and spatial position difference between pixels, it is possible to more accurately determine which pixels should be classified as the same area.

[0015] Furthermore, the modified abnormal probability satisfies the following relationship:

[0016]

[0017] In the formula, It is The original image The corrected anomaly probability of a region is It is The original image The abnormal probability of each region before correction, It is The weight of the original image, and , It is The entropy value of the abnormal probability of a region in all original images, is a preset parameter used to avoid the denominator being 0. is the total number of original images.

[0018] The above technical scheme quantifies the fluctuation degree of abnormal probability of the same area in different original images through entropy value. The higher the entropy value, the greater the change of abnormal probability of the area in different original images, that is, the greater the change of abnormal probability of the area at different time points, and the higher the uncertainty of abnormal probability of the area; conversely, the lower the entropy value, the smaller the change of abnormal probability of the area, and the lower the uncertainty. The weight is calculated by entropy value. If the abnormal probability of an area does not change much, that is, the entropy value is low, then its weight will be high; conversely, if the abnormal probability of an area changes greatly, that is, the entropy value is high, then its weight will be low. The purpose of this design is to emphasize the importance of those areas where the abnormal probability changes relatively stably. By weighted average, the abnormal probability of the same area in multiple original images is comprehensively considered, and the corrected abnormal probability is finally obtained. The images with higher weights, that is, those images with smaller changes in regional abnormal probability, have a greater impact on the final result. The scheme introduces entropy value to measure the change of regional abnormal probability, and calculates weight accordingly, which can better identify those areas with continuous and stable temperature anomalies and improve the accuracy of temperature anomaly detection.

[0019] Furthermore, the abnormal probability of each pixel satisfies the following relationship:

[0020]

[0021] In the formula, It is The abnormal probability of a pixel point is is the normalization function, is the total number of original images, is the sequence number of the original image, The value of All integers in the range, For the The original image The gray value of a pixel, For the The original image The grayscale mean of the eight neighboring pixels of a pixel, For the The original image The gradient value of a pixel, For the The original image The mean gradient of the eight neighboring pixels of a pixel.

[0022] The above technical solution considers the difference between the grayscale value of the pixel and the grayscale mean of the neighborhood, and the difference between the gradient value of the pixel and the gradient mean of the neighborhood pixels. A larger difference value indicates that the pixel is different from the surrounding pixels in grayscale and may be an anomaly. The gradient value of the pixel itself reflects the steepness of the temperature change around the pixel. The larger the difference, the greater the inconsistency in temperature change between the pixel and the surrounding points. The abnormal conditions of the same pixel in multiple original images are combined through the summation term, thereby enhancing the robustness of anomaly detection. The gradient value of the pixel itself is used as a weighting factor when calculating the anomaly probability, indicating that in anomaly detection, areas with steep temperature changes may be more worthy of attention.

[0023] Furthermore, the abnormal probability of any region in any of the original images satisfies the following relationship: ,in, It is The original image The abnormal probability of a region, For the The original image The blurriness of the area, For the The original image The degree of abnormality in the area , For the The grayscale blur degree of the area, For the The gradient blur degree of the region, is the normalization function.

[0024] In the above technical solution, the blur degree includes grayscale blur degree and gradient blur degree. The grayscale blur degree reflects the consistency or inconsistency of the grayscale value of the pixels inside the region, and the gradient blur degree reflects the consistency or inconsistency of the gradient value of the pixels inside the region. The greater the blur degree, the worse the internal consistency of the region, which may mean that the region contains more edge or texture details, or there is an uneven temperature distribution. The abnormality degree indicates that the The original image The degree of deviation of a region from the normal situation reflects whether the temperature in the region is abnormal. By multiplying the degree of fuzziness by the degree of abnormality, the internal consistency of the region and the degree of temperature abnormality can be comprehensively considered to accurately assess the abnormal probability of the region.

[0025] Furthermore, the method for determining whether the area is an abnormal area according to the size of the corrected abnormal probability is:

[0026] The corrected abnormal probability of the area in all original images is obtained, the abnormal probabilities are normalized and the average is calculated, and the average is used as the final abnormal probability of the area. If the final abnormal probability is greater than a preset probability threshold, the corresponding area in all original images is an abnormal area; if the final abnormal probability is not greater than the preset probability threshold, the corresponding area in all original images is not an abnormal area.

[0027] Furthermore, the grayscale blur degree and gradient blur degree of the region are obtained by:

[0028] Taking the center of gravity of the region as the starting point, extending lines from the center of gravity to the eight neighborhood directions are drawn respectively until the extension lines intersect the edge of the region, thereby obtaining eight neighborhood straight lines. The variance of the grayscale values ​​of all pixels passing through each neighborhood straight line is taken as the grayscale variance of the neighborhood straight line. Each two neighborhood straight lines symmetrical around the center are grouped together, thereby obtaining four groups of neighborhood straight lines.

[0029] According to the difference in grayscale variances of two neighborhood lines in each group of neighborhood lines, the difference mean of the grayscale variances corresponding to the four groups of neighborhood lines is determined, and the sum of the difference mean and the grayscale variance mean of the eight neighborhood lines is taken as the grayscale blur degree of the area; the gradient variance of all edge pixels in the area is taken as the gradient blur degree of the area.

[0030] Consideration of the degree of grayscale blur in the above technical solution: By calculating the grayscale variance in different directions, the consistency of the temperature distribution in the region can be reflected. If the temperature distribution in the region is uniform, the grayscale variance in each direction should be similar, and the difference mean is small; conversely, if the temperature distribution is uneven, the difference mean is large. Consideration of the degree of gradient blur: By calculating the gradient variance of edge pixels, the change of the temperature gradient at the boundary of the region can be reflected. If the temperature changes slowly at the edge, the gradient variance is small; conversely, if the temperature changes drastically, the gradient variance is large. Finally, through the comprehensive consideration of the grayscale blur and gradient blur, the grayscale and gradient changes inside and at the edge of the region are carefully analyzed, which can more accurately identify areas with uneven temperature distribution, and help improve the accuracy of temperature anomaly detection.

[0031] Furthermore, the size and position of each region of all the original images are the same.

[0032] Furthermore, the method for evaluating the operating state of the decomposition furnace according to the total number of all abnormal areas and the total number of pixels contained in the abnormal areas includes:

[0033] Calculate the abnormal indicators of the operating status of the decomposition furnace:

[0034]

[0035] In the formula, is the value of the abnormal index of the operating status of the decomposition furnace, is the normalization function, is the total number of abnormal regions, is the total number of all regions, is the total number of pixels contained in all abnormal areas, is the total number of pixels contained in all regions, is the mean of the final abnormal probability of all regions, , and are respectively a first preset weight, a second preset weight and a third preset weight;

[0036] The numerical value of the abnormal index is compared with a preset abnormal index threshold to determine whether the operating state of the decomposition furnace is abnormal.

[0037] The above technical scheme measures the proportion of abnormal areas to all areas. If the proportion is high, it indicates that there are more temperature abnormal areas. It measures the proportion of the number of pixels contained in the abnormal areas to the total number of pixels. If the proportion is high, it indicates that the range of temperature anomaly is wide. It measures the average abnormal probability of all areas. If the average abnormal probability is high, it indicates that the overall degree of temperature anomaly is serious. Finally, the above three indicators are combined by weighted average, so that the abnormal indicator can comprehensively reflect the operating status of the decomposition furnace.

[0038] Further, the method for determining whether the operating state of the decomposition furnace is abnormal is:

[0039] If the abnormal index is greater than or equal to a preset abnormal index threshold, the operating state of the decomposition furnace is abnormal; if the abnormal index is less than the preset abnormal index threshold, the operating state of the decomposition furnace is not abnormal.

[0040] The present invention has the following effects:

[0041] Through multi-angle analysis and quantitative evaluation, this technical solution can accurately locate the abnormal temperature areas in the decomposition furnace. By comprehensively considering the grayscale and gradient differences of pixel points, spatial distance and regional blur, it can achieve accurate positioning, identification and clustering of abnormal temperature areas. Finally, through statistics on the number and area of ​​abnormal areas, it provides a quantitative evaluation of the operating status of the decomposition furnace, and can accurately locate the area with abnormal temperature, thereby improving the accuracy and response speed of fault diagnosis and helping operators take maintenance measures in a timely manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0043] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0044] 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 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0045] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0046] Reference Figure 1 A method for detecting the operating status of a decomposition furnace includes steps S1 to S4, which are specifically as follows:

[0047] S1: Collect infrared thermal images of the decomposition furnace, preprocess to obtain the original image, and calculate the abnormal probability of each pixel.

[0048] Install an infrared thermal imager at a suitable position in front of the decomposition furnace to ensure that the infrared thermal imager can cover the entire decomposition furnace. Collect infrared thermal images of the decomposition furnace during stable operation. Set the acquisition frequency to one image per second and the acquisition time to seconds, and finally collected Infrared thermal imaging image, in this embodiment, Then, each infrared thermal image is preprocessed, specifically: grayscale processing. After grayscale processing, the higher the original temperature of the pixel, the larger the grayscale value corresponding to the pixel, and the lower the original temperature of the pixel, the lower the grayscale value corresponding to the pixel. Then, an adaptive filtering algorithm is used to remove noise, and histogram equalization is used to enhance the contrast to make the detail differences of the image clearer. Finally, 300 original images after preprocessing are obtained, and each original image has the same size as a set of original images.

[0049] In order to more effectively identify the abnormal temperature of the decomposition furnace from the infrared thermal imaging image, it is necessary to perform a detailed analysis of each pixel. In each original image, a rectangular coordinate system is first defined, with the lower left corner of the original image as the origin, the horizontal axis to the right as the horizontal axis, and the vertical axis upward as the vertical axis. Then, the position coordinates of each pixel are traversed, and the grayscale value and gradient value of each pixel are recorded. Assuming that the maximum width of the original image is X pixels and the height is Y pixels, the number of pixels in each original image is .

[0050] Each pixel here is determined based on the position, that is, the pixel at the same position in all original images, for example, The pixel at or all pixels in the original image located at In order to evaluate the probability of abnormality of each pixel, this step is to determine the probability of abnormality of the pixel according to the grayscale change and gradient change of the pixel in multiple consecutive original images.

[0051] Therefore, the pixels and Take the original image as an example, and get the The pixel at The grayscale value in the original image and the The pixel at The gradient value in the original image is In the original image, The difference in grayscale value and gradient value between a pixel and its eight neighboring pixels. Since each time point (each acquisition moment) corresponds to an original image, analyzing the change of the pixel in each original image is to analyze the change of the pixel at each time point.

[0052] Then, according to The grayscale difference and gradient difference between the pixel point and the eight neighboring pixels in all original images are calculated. The abnormal probability of a pixel point is calculated as follows:

[0053]

[0054] In the formula, It is The abnormal probability of a pixel point is is the normalization function, is the total number of original images, is the sequence number of the original image, The value of All integers in the range, For the The original image The gray value of a pixel, For the The original image The grayscale mean of the eight neighboring pixels of a pixel, For the The original image The gradient value of a pixel reflects the steepness of the temperature change around the pixel. The greater the difference, the greater the inconsistency between the temperature change of the pixel and the surrounding points. For the The original image Therefore, the greater the grayscale difference and gradient difference between a pixel and its neighboring pixels, the more likely the pixel is to have a temperature anomaly. It can be seen that the abnormal probability of a pixel is positively correlated with the grayscale difference and gradient difference.

[0055] In this formula, Indicates The original image The difference between the gray value of a pixel and the average gray value of the pixels in its eight neighborhoods. If the gray value of a pixel deviates significantly from the average of its surrounding pixels, it may be abnormal. Indicates The original image The difference between the gradient value of a pixel and the average gradient value of the pixels in its eight neighborhoods. The gradient value reflects the rate of change of the grayscale value between pixels, that is, the speed of temperature change. If the temperature change rate of a pixel is significantly different from that of its neighboring areas, it is considered that there may be an abnormality at this point. It is used as a weighting factor to weight the sum of grayscale difference and gradient difference, which means that anomalies in areas with drastic temperature changes (high gradient values) are more important because they may be more significant temperature anomalies. The function ensures the probability of anomaly Within a reasonable range, the abnormal probabilities of different pixels can be effectively compared.

[0056] In this formula, First, calculate the first The difference between the gray value of the pixel and the average gray value of the pixels in its eight neighborhoods, and The difference between the gradient value of the pixel and the average gradient value of the pixels in its eight neighborhoods, and then calculate the cumulative value of the two differences and the The product value of the gradient values ​​of the pixels is calculated, and then one original image obtains a product value. Finally, the average value of the product values ​​obtained from all original images is calculated. The smaller the average value is, the more The grayscale changes and gradient changes of each pixel are relatively stable. The abnormal probability of each pixel is also smaller, reflecting that the decomposition furnace is The temperature change at the pixel position is also relatively stable.

[0057] This step can capture the continuous fluctuation of the grayscale and gradient features of each pixel in the original image at multiple time points by calculating the grayscale and gradient features of each pixel. By analyzing the eight neighborhoods of the pixel, it can detect temperature mutations in a small range. The gradient value of the pixel itself is used as a weighting factor for weighted summation, highlighting that those areas with fast temperature changes may be more likely to have temperature anomalies. According to this step, the anomaly probability of all pixels (based on coordinate position) can be obtained.

[0058] S2: Select any original image, and determine the possibility that each pixel belongs to the same region as other pixels based on the difference and distance between the abnormal probabilities of each pixel and other pixels. Based on this possibility, cluster all pixels in the original image to obtain multiple regions, and use the regions in the original image as the standard to obtain the corresponding regions in the remaining original images.

[0059] Select an original image from all the original images, perform clustering operation on all the pixels in the image, obtain multiple regions composed of pixels, complete the regional division of all the pixels in the original image, and considering that the pixels need to be clustered, it is necessary to set a clustering standard, and then perform clustering operation based on the clustering standard.

[0060] Therefore, the custom clustering criterion in this step is: the probability that each pixel belongs to the same region as other pixels. As an example, in the original image Select the original image pixel and any other pixel, calculate the The probability that a pixel belongs to the same area as other pixels satisfies the following relationship:

[0061]

[0062] In the formula, For the Pixels and The pixel at The probability that the original images belong to the same region, For the The abnormal probability of a pixel point is For the The abnormal probability of a pixel point is For the The pixel at The horizontal coordinate in the original image, For the The pixel at The horizontal coordinate in the original image, For the The pixel at The vertical coordinate in the original image, For the The pixel at The vertical coordinate in the original image, For the The maximum value of the horizontal coordinate in the original image. For the The maximum value of the vertical coordinate in the original image. is a natural exponential function.

[0063] In this formula, Indicates The abnormal probability of the pixel is The difference between the abnormal probabilities of two pixels. If the difference between the abnormal probabilities of two pixels is small, it means that they are more consistent in temperature anomaly and are more likely to belong to the same area. Represents the Euclidean distance between the two pixels, the Euclidean distance divided by It means that the Euclidean distance is normalized to make it consistent with the dimension of the abnormal probability difference between the two pixels. If the distance between two pixels is small, it means that they are closer in space and more likely to belong to the same area.

[0064] The formula calculates the first Pixels and The possibility of whether the pixels belong to the same area can be determined based on the possibility, and which pixels are more likely to belong to the same temperature anomaly area can be identified.

[0065] By comprehensively considering the abnormal probability difference and spatial distance, and performing nonlinear transformation through exponential function, It can better reflect the possibility that two pixels belong to the same area. When the difference in the abnormal probability of the two pixels is small and the spatial distance is small, The larger the value of , the more likely they are to belong to the same region.

[0066] According to this formula, we can get the probability that any two pixels in the original image belong to the same area. Then we use the probability that the pixels belong to the same area as the clustering criterion to perform clustering operations:

[0067] The number of initial clusters is set to 2, and increased by 1 each time. After each increase, cluster center points are randomly selected in the original image according to the number of clusters. As many cluster center points as the number of clusters are randomly selected, each cluster center point is a pixel point, and then the possibility of each pixel point belonging to the same area as each cluster center point is calculated, and the pixel point is assigned to the cluster center point with the greatest possibility. In this way, all pixels in the original image are clustered into multiple clusters, and then the SSE (within cluster error variance) of the clustering result corresponding to the current number of clusters is analyzed, that is, the sum of squared errors of the possibility that the pixels in all clusters in the current clustering result and the cluster center points of their clusters belong to the same area.

[0068] The SSE of the clustering results corresponding to the number of each cluster is obtained in turn, and a coordinate system is constructed with the number of clusters as the horizontal axis and the SSE of the corresponding clustering results as the vertical axis to obtain the changing trend of SSE with the number of clusters. When the elbow point method is used to obtain the "elbow point" in the changing trend, the number of clusters is stopped from increasing, and the number of clusters corresponding to the "elbow point" is taken as the optimal number of clusters.

[0069] According to the optimal number of clusters, cluster center points are randomly selected in the original image, and clustering operations are performed on all pixels in the original image. All pixels are divided into multiple clusters, and the pixels in each cluster constitute a region. Finally, the original image is divided into multiple regions.

[0070] Taking the original image as a standard, the corresponding regions in the remaining original images are obtained, that is, the size and position of each region in all the original images are the same, and are one-to-one corresponding.

[0071] S3: Evaluate the abnormality degree of each region according to the abnormality probability of the pixel points in each region in the original image, and calculate the abnormality probability of each region according to the abnormality degree and blur degree of each region.

[0072] In order to more accurately evaluate the degree of abnormality in each area, the mean abnormal probability of all pixels in the area is calculated, and then the probability mean is normalized and the normalized value is used as the degree of abnormality in the area.

[0073] First The first Take a region as an example, and its abnormality degree is:

[0074]

[0075] In this formula, It is The first The degree of abnormality in the area It is The first The mean of the abnormal probability of all pixels in the region, represents normalization processing, Represents the inverse tangent function. When the average value of the abnormal probability of all pixels in the area is larger, the abnormal degree of the area is greater, and vice versa.

[0076] Since the degree of abnormality is only a preliminary and simple overall abnormality assessment of the degree of abnormality of the area based on the average abnormal probability of the pixels, the result is relatively rough. Therefore, in order to improve the accuracy of abnormality analysis, it is necessary to further analyze the edge features (gradient features) and internal features (grayscale features) of the area in detail to obtain more accurate abnormality assessment results.

[0077] In a physical environment, heat usually follows the second law of thermodynamics and diffuses from high-temperature areas to low-temperature areas. This heat diffusion effect causes the edges of the heat distribution to gradually transition rather than suddenly change, making the edges blurred. When the temperature difference in a certain area is large (that is, there is a high temperature gradient), the edge of the area will be relatively clear because the temperature change is more rapid. On the contrary, when the temperature difference in the area is small (that is, the temperature gradient is low), the edge of the area will appear blurred because the temperature change is relatively gentle.

[0078] In thermal imaging images, the grayscale value usually reflects the temperature. The higher the grayscale value, the higher the temperature; the lower the grayscale value, the lower the temperature. Therefore, when we observe the change of grayscale value from the center to the edge of a high temperature area, and when we observe the change of grayscale value from the center to the edge of a low temperature area, different temperature change trends can be expected. In a high temperature area, if the temperature distribution in the center of the high temperature area is uniform, then the grayscale value in this area will present a platform area, that is, the grayscale value is basically unchanged. When traversing the pixels from the center point to the edge of the area, the change of grayscale value mainly depends on the size of the temperature gradient. Specifically, the change trend of grayscale value can be divided into two cases: when the temperature gradient is large (the edge of the area is clear) and when the temperature gradient is small (the edge of the area is fuzzy). When the temperature gradient is large, the temperature changes rapidly from the center to the edge, which means that when traversing the pixels from the center to the edge, the grayscale value decreases rapidly, which is specifically manifested as the grayscale value drops from the highest value to the lower value in a short time; when the temperature gradient is small, the temperature changes slowly from the center to the edge. This means that when traversing pixels from the center to the edge, the grayscale value will gradually decrease, which is manifested as the grayscale value slowly decreases with increasing distance.

[0079] Therefore, based on the above analysis, the calculation formula for the abnormal probability of each area is set as:

[0080]

[0081] In this formula, It is The original image The abnormal probability of a region, For the The original image The blurriness of the area, For the The original image The degree of abnormality in an area.

[0082] In this formula, , For the The grayscale blur degree of the area, For the The gradient blur degree of the region, is a normalized function, and the abnormal probability is positively correlated with the grayscale blur degree and the gradient blur degree. By multiplying the blur degree and the abnormal degree and then performing normalization, a comprehensive evaluation of the regional abnormal probability is achieved, taking into account the consistency (blur degree) and significance (abnormal degree) of temperature changes, making the calculation of abnormal probability more comprehensive. The grayscale blur degree in the formula reflects the change of grayscale values ​​in the region, and the gradient blur degree reflects the severity of temperature changes in the region. By adding the two, the blur degree of the region can be comprehensively evaluated.

[0083] Among them, the method for obtaining the grayscale blur degree of each area is:

[0084] Starting from the center of gravity of the region, draw extension lines from the center of gravity to the eight neighborhood directions, and stop when the extension lines intersect the edge of the region, to obtain eight neighborhood lines. The variance of the grayscale values ​​of all pixels passing through each neighborhood line is taken as the grayscale variance of the neighborhood line. Each two neighborhood lines symmetrical around the center are grouped together to obtain four groups of neighborhood lines.

[0085] According to the difference in grayscale variances of two neighborhood lines in each group of neighborhood lines, the difference mean of grayscale variances corresponding to the four groups of neighborhood lines is determined, and the sum of the difference mean and the grayscale variance mean of the eight neighborhood lines is taken as the grayscale blur degree of the region.

[0086] This method reflects the degree of change of the grayscale value on each neighborhood line through the variance of the grayscale value of the pixel point on the neighborhood line. The larger the grayscale variance, the greater the grayscale value change of the pixel point on the neighborhood line, that is, the more drastic the temperature change. By calculating the difference in the grayscale variance of two neighborhood lines in each group of neighborhood lines, the inconsistency of grayscale changes in different directions can be reflected. The larger the difference in grayscale variance, the more inconsistent the grayscale changes in different directions. The difference mean of the grayscale variance corresponding to the four groups of neighborhood lines is calculated, and the grayscale in different directions can be comprehensively evaluated. The consistency of the change, the larger the difference mean, the more inconsistent the grayscale changes in different directions, that is, the more uneven the grayscale distribution of the region, the more likely it is an abnormal region. The sum of the difference mean of the grayscale variance and the grayscale variance mean of the eight neighborhood lines is used as the grayscale fuzziness of the region, which can comprehensively consider the change of the grayscale value inside the region and the consistency of the grayscale change in different directions. The larger the grayscale fuzziness, the greater the grayscale value change inside the region, the more inconsistent the grayscale change in different directions, that is, the more uneven the temperature distribution of the region, the more likely it is an abnormal region. The calculation method of the grayscale fuzziness can more comprehensively reflect the temperature change inside the region, so as to better identify the temperature abnormality area.

[0087] Among them, the method for obtaining the gradient blur degree of each area is:

[0088] The gradient variance of all edge pixels in the region is used as the gradient blurring degree of the region. Edge pixels refer to pixels located on the boundary of the region. The gradient values ​​of these pixels are usually large because they are located at locations where the temperature changes more drastically. The variance of the gradient values ​​of all edge pixels is calculated. The gradient value reflects the severity of the temperature change, and the variance reflects the change in the gradient values ​​of these edge pixels. The gradient variance of all edge pixels is used as the gradient blurring degree of the region. The gradient blurring degree reflects the change in the gradient values ​​of these edge pixels. The greater the gradient blurring degree, the greater the change in the gradient values ​​of the edge pixels in the region, that is, the stronger the inconsistency of the temperature change. The greater the gradient blurring degree, the greater the change in the gradient values ​​of the edge pixels and the more blurred the edge. Conversely, the smaller the gradient blurring degree, the smaller the change in the gradient values ​​of the edge pixels and the clearer the edge. The gradient blurring degree helps to analyze the severity and distribution of temperature changes, more comprehensively reflect the temperature changes of edge pixels, and thus better identify temperature abnormality areas.

[0089] S4: Correct the abnormal probability of each area, filter the abnormal area according to the corrected abnormal probability, and evaluate the operation status of the decomposition furnace according to the total number of all abnormal areas and the total number of pixels contained in the abnormal areas.

[0090] The temperature field or material state in the decomposition furnace changes dynamically. This change may cause the captured thermal imaging edge information to be unstable, further increasing the possibility of edge blur. Therefore, when analyzing each original image, the abnormal probability of certain areas may be misjudged due to local instantaneous changes. Therefore, it is necessary to combine multiple original images to optimize the abnormal probability of each area in order to more accurately evaluate the abnormal probability of the area.

[0091] The specific optimization methods are:

[0092] Get all the original images Regions (with The original image The abnormal probability of the area corresponding to the area) is calculated by the method of calculating the abnormal probability in step S3. The original image The calculation method of abnormal probability for each region is the same.

[0093] Calculate the The entropy value of the abnormal probability of a region in all original images. If the entropy value is high, it means that the abnormal probability of the region varies greatly over time and the uncertainty is high. It is more necessary to smooth the abnormal probability of the region to reduce the fluctuation of the abnormal probability caused by time changes. On the contrary, if the entropy value is low, it means that the abnormal probability of the region is relatively stable and does not require too much correction.

[0094] Therefore, based on the above entropy value, the above abnormal probability is corrected, and the correction formula is:

[0095]

[0096] In the formula, It is The original image The corrected anomaly probability of a region is It is The original image The abnormal probability of each region before correction, It is The weight of the original image, and , It is The entropy value of the abnormal probability of a region in all original images, is a preset parameter used to avoid the denominator being 0. is the total number of original images.

[0097] The weight is calculated based on the entropy value. When the entropy value is high, the weight is small, indicating that the The original image The abnormal probability of the region needs more smoothing. When the entropy value is low, the weight is large, indicating that the first The original image The abnormal probability of each area has been relatively stable and does not require too much correction.

[0098] The significance of the revised abnormal probability: The revised abnormal probability takes into account the first The abnormal probability of a region in all original images is weighted by the weight. Because these original images are based on time sequence, each acquisition moment corresponds to an original image. Through this operation, the fluctuation of abnormal probability caused by time change can be reduced and the stability of abnormal probability can be improved.

[0099] According to the above method, the corrected abnormal probability of each region (the same region in all original images) is obtained.

[0100] The method for determining whether the area is an abnormal area according to the size of the corrected abnormal probability is as follows:

[0101] Get the corrected abnormality probability of the area in all original images, normalize these abnormality probabilities and calculate the average, and use the average as the final abnormality probability of the area, that is, the overall abnormality probability of the area within 300 seconds. The preset final abnormality probability is 0.6. If the final abnormality probability of the area is greater than 0.6, the area is an abnormal area; if the final abnormality probability of the area is not greater than 0.6, the area is not an abnormal area.

[0102] The method for evaluating the operation status of the decomposition furnace according to the total number of all abnormal areas and the total number of pixels contained in the abnormal areas includes:

[0103] Calculate the abnormal indicators of the operating status of the decomposition furnace:

[0104]

[0105] In the formula, is the value of the abnormal index of the operating status of the decomposition furnace, is the normalization function, is the total number of abnormal regions, is the total number of all regions, is the total number of pixels contained in all abnormal areas, is the total number of pixels contained in all regions, is the mean of the final abnormal probability of all regions, , and are respectively a first preset weight, a second preset weight and a third preset weight, .

[0106] In this formula, if the proportion of abnormal areas to all areas is high, it means that there are more temperature abnormal areas within the 300 seconds of operation of the decomposition furnace. By measuring the proportion of the number of pixels contained in the abnormal area to the total number of pixels, if the proportion is high, it means that the range of temperature abnormality is wider. The average abnormal probability of all areas is high, which means that the overall degree of temperature abnormality is more serious. Finally, the above three indicators are combined by weighted average, so that the abnormal indicators can fully reflect the operating status of the decomposition furnace.

[0107] Comparing the value of the abnormal index with the preset abnormal index threshold value to determine whether the operation state of the decomposition furnace is abnormal:

[0108] The preset abnormal index threshold is 0.5. If the abnormal index is greater than or equal to 0.5, there is an abnormality in the operation state of the decomposition furnace. The relevant operators are notified to make timely adjustments and take corresponding measures. The abnormal area can be accurately located according to this step. If the abnormal index is less than the preset abnormal index threshold, there is no abnormality in the operation state of the decomposition furnace.

[0109] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0110] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for detecting the operating status of a decomposition furnace, characterized in that: include: Preprocessing multiple infrared thermal images of the decomposition furnace to obtain multiple original images, and calculating the abnormal probability of each pixel point according to the grayscale difference and gradient difference between each pixel point and neighboring pixel points in the multiple original images, wherein the abnormal probability is positively correlated with the grayscale difference and the gradient difference; According to the difference and distance of the abnormal probability of each pixel point with other pixels in any of the original images, the possibility that each pixel point belongs to the same area as other pixels is determined, satisfying the following relationship: ; In the formula, For the Pixels and The pixel at The probability that the original images belong to the same region, For the The abnormal probability of a pixel point is For the The abnormal probability of a pixel point is For the The pixel at The horizontal coordinate in the original image, For the The pixel at The horizontal coordinate in the original image, For the The pixel at The vertical coordinate in the original image, For the The pixel at The vertical coordinate in the original image, For the The maximum value of the horizontal coordinate in the original image. For the The maximum value of the vertical coordinate in the original image. is the natural exponential function; In the original image, all pixels are clustered into multiple regions based on the possibility, the mean of the abnormal probabilities of all pixels in any region is normalized as the abnormality degree of the region, and the abnormality probability of the region is calculated according to the abnormality degree of the region and the blur degree of the region, wherein the abnormality probability of the region is positively correlated with the blur degree and the abnormality degree; The abnormal probability of the area is corrected by using the entropy value of the abnormal probability of the area corresponding to the area in all original images, and whether the area is an abnormal area is determined according to the size of the corrected abnormal probability, and the operating status of the decomposition furnace is evaluated according to the total number of all abnormal areas and the total number of pixels contained in the abnormal areas.

2. The method for detecting the operating status of a decomposition furnace according to claim 1, characterized in that: The modified abnormal probability satisfies the following relationship: ; In the formula, It is The original image The corrected anomaly probability of a region is It is The original image The abnormal probability of each region before correction, It is The weight of the original image, and , It is The entropy value of the abnormal probability of a region in all original images, is a preset parameter used to avoid the denominator being 0. is the total number of original images.

3. The method for detecting the operating status of a decomposition furnace according to claim 1, characterized in that: The abnormal probability of each pixel satisfies the following relationship: ; In the formula, It is The abnormal probability of a pixel point is is the normalization function, is the total number of original images, is the sequence number of the original image, The value of All integers in the range, For the The original image The gray value of a pixel, For the The original image The grayscale mean of the eight neighboring pixels of a pixel, For the The original image The gradient value of a pixel, For the The original image The mean gradient of the eight neighboring pixels of a pixel.

4. The method for detecting the operating status of a decomposition furnace according to claim 2, characterized in that: The abnormal probability of any region in any of the original images satisfies the following relationship: ,in, It is The original image The probability of anomaly in a region, For the The original image The blurriness of the area, For the The original image The degree of abnormality in the area , For the The grayscale blur degree of the area, For the The gradient blur degree of the region, is the normalization function.

5. The method for detecting the operating status of a decomposition furnace according to claim 2, characterized in that: The method for determining whether the area is an abnormal area according to the size of the corrected abnormal probability is: The corrected abnormal probability of the area in all original images is obtained, the abnormal probabilities are normalized and the average is calculated, and the average is used as the final abnormal probability of the area. If the final abnormal probability is greater than a preset probability threshold, the corresponding area in all original images is an abnormal area; if the final abnormal probability is not greater than the preset probability threshold, the corresponding area in all original images is not an abnormal area.

6. The method for detecting the operating status of a decomposition furnace according to claim 4, characterized in that: The method for obtaining the grayscale blur degree and gradient blur degree of the region is: Taking the center of gravity of the region as the starting point, extending lines from the center of gravity to the eight neighborhood directions are drawn respectively until the extension lines intersect the edge of the region, thereby obtaining eight neighborhood straight lines. The variance of the grayscale values ​​of all pixels passing through each neighborhood straight line is taken as the grayscale variance of the neighborhood straight line. Each two neighborhood straight lines symmetrical around the center are grouped together, thereby obtaining four groups of neighborhood straight lines. According to the difference in grayscale variances of two neighborhood lines in each group of neighborhood lines, the difference mean of the grayscale variances corresponding to the four groups of neighborhood lines is determined, and the sum of the difference mean and the grayscale variance mean of the eight neighborhood lines is taken as the grayscale blur degree of the area; the gradient variance of all edge pixels in the area is taken as the gradient blur degree of the area.

7. The method for detecting the operating status of a decomposition furnace according to claim 4, characterized in that: The size and position of each region of all the original images are the same.

8. The method for detecting the operating status of a decomposition furnace according to claim 5, characterized in that: The method for evaluating the operating state of the decomposition furnace according to the total number of all abnormal areas and the total number of pixels contained in the abnormal areas comprises: Calculate the abnormal indicators of the operating status of the decomposition furnace: ; In the formula, is the value of the abnormal index of the operating status of the decomposition furnace, is the normalization function, is the total number of abnormal regions, is the total number of all regions, is the total number of pixels contained in all abnormal areas, is the total number of pixels contained in all regions, is the mean of the final abnormal probability of all regions, , and are respectively a first preset weight, a second preset weight and a third preset weight; The numerical value of the abnormal index is compared with a preset abnormal index threshold to determine whether the operating state of the decomposition furnace is abnormal.

9. The method for detecting the operating status of a decomposition furnace according to claim 8, characterized in that: The method for determining whether the operating state of the decomposition furnace is abnormal is: If the abnormal index is greater than or equal to a preset abnormal index threshold, the operating state of the decomposition furnace is abnormal; if the abnormal index is less than the preset abnormal index threshold, the operating state of the decomposition furnace is not abnormal.

Citation Information

Patent Citations

  • Real-time monitoring system for combustion status of steam boiler based on thermal imaging

    CN116612438B

  • Copper pipe production defect detection method and system

    CN118037722A

Cited By

  • Motion robot double-light fusion furnace bottom non-stop temperature measurement method and system

    CN121977701A

  • A method and system for measuring temperature of a moving robot double-light fusion furnace bottom without stopping

    CN121977701B