A method for detecting waste gas leakage in a bag filter

Abstract

The present invention relates to the technical field of image processing, and particularly to a method for detecting exhaust gas leakage of a bag filter. The detection method includes: continuously collecting multiple frames of images of the flat plate of the flower plate, calculating the local information entropy of each bag area in each frame of image, determining the local information entropy sequence of each bag area according to multiple reference images of the current frame of image, and obtaining the entropy change stability through the first-order difference sequence of the local information entropy sequence; using the gradient of the local information entropy of each bag area in the reference image, and combining the spatial vectors of each bag area and adjacent bag areas, calculating the vector convergence degree of each bag; comprehensively obtaining the leakage possibility degree of each bag area based on the local information entropy, entropy change stability and vector convergence degree of each bag area, and determining the bag with leakage according to the leakage possibility degree, thereby improving the accuracy of exhaust gas leakage detection of the bag filter.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing. Specifically, it relates to a method for detecting exhaust gas leakage in a bag filter. Background Art

[0002] In the field of industrial waste gas purification, bag filters play a crucial role in reducing dust emissions, protecting the ecological environment, and ensuring human health due to their excellent performance in efficiently capturing solid particulate matter.

[0003] Implementing exhaust gas leakage detection for bag filters is a key link to ensure their stable and efficient operation. Because once exhaust gas leaks, a large amount of dust will escape, which will not only significantly reduce the dust removal efficiency of the dust collector but also increase the risk of environmental pollution.

[0004] At present, using image processing technology to locate the bags with exhaust gas leakage in a bag filter has become an important detection method adopted in the industry. Existing detection schemes mainly judge the bag areas with exhaust gas leakage by analyzing the change of local information entropy of the bag areas in consecutive multiple frames of images according to the magnitude of the difference in local information entropy.

[0005] However, in actual working conditions, when the exhaust gas leakage and diffusion speed is too fast, due to the rapid diffusion of exhaust gas in a short time, the image features in a large area of bag areas will change significantly. As a result, it is difficult for the detection algorithm based on the change of local information entropy to accurately distinguish which changes are from the real leakage source and which changes are the "false leakage" phenomena caused by the diffusion of exhaust gas. A large area of bag areas will be misjudged as leakage areas, reducing the accuracy of exhaust gas leakage detection and affecting the dust removal effect of the bag filter. Summary of the Invention

[0006] To solve the problem that the existing exhaust gas leakage detection method for bag filters is prone to misjudgment and reduces the accuracy of exhaust gas leakage detection. The present invention proposes a method for detecting exhaust gas leakage in a bag filter, including:

[0007] Continuously collect multiple frames of images of the platen plane of the bag filter, the platen plane includes multiple bag areas, and calculate the local information entropy of each bag area in each frame of image; taking any frame of image as the current frame image, determine the local information entropy sequence of each bag area of the current frame image according to multiple reference images of the current frame image;

[0008] Determine the entropy change stability of each bag area according to the first-order difference sequence of the local information entropy sequence of each bag area; calculate the vector convergence degree according to the gradient of the local information entropy of each bag area in each reference image and the spatial vector between each bag area and adjacent bag areas:

[0009] , is the vector convergence degree of the th cloth bag area, is the th cloth bag area's th adjacent cloth bag area's gradient of local information entropy in the th reference image, is the th cloth bag area's spatial vector with the th adjacent cloth bag area, is the total number of adjacent cloth bag areas of the th cloth bag area, is the total number of reference images, is for calculating the modulus length, is for taking the absolute value, is the natural exponential function, is the normalization function;

[0010] Taking the product of the local information entropy, entropy change stability and vector convergence degree of each cloth bag area as the possible degree of leakage of each cloth bag area, and performing waste gas leakage detection according to the possible degree of leakage.

[0011] The above technical solution calculates the local information entropy of multiple frames of images, establishes the time series data of the state of the cloth bag area, which can be used to analyze the dynamic changes of the state of the cloth bag area over time. Compared with a single frame of image, it can comprehensively understand the changes of each cloth bag area. After selecting the current frame of image, the local information entropy sequence of each cloth bag area is determined by combining multiple reference images. By analyzing the sequence changes, the change trend of the local information entropy of the cloth bag area can be obtained. By analyzing the change trend, the abnormal entropy value fluctuations caused by waste gas leakage can be more accurately identified, avoiding misjudging the pseudo-leakage characteristics caused by waste gas diffusion as real leakage. And further, the change rate of the local information entropy sequence is reflected by the first-order difference sequence, and the entropy change stability can accurately measure whether the change of the local information entropy of the cloth bag area is stable. The gradient of the local information entropy reflects the change direction and degree of the local information entropy of the cloth bag area, and the spatial vector reflects the spatial position relationship between the cloth bag areas. By combining the two to calculate the vector convergence degree, the mutual relationship between the cloth bag areas can be comprehensively considered from the spatial correlation dimension and the time correlation dimension, and it can accurately reflect whether the waste gas flow and information change between adjacent cloth bag areas are regular, and then more accurately locate the leakage source. And further, by integrating the above three key indicators, the possibility of waste gas leakage in the cloth bag area can be comprehensively and comprehensively evaluated, avoiding the misjudgment problem caused by waste gas diffusion, and improving the accuracy of waste gas leakage detection.

[0012] Furthermore, the entropy change stability of each cloth bag area is determined based on the following formula:

[0013] ;

[0014] In the formula, is the entropy change stability of the th cloth bag area, is the th value of the first-order difference sequence of the local information entropy of the th cloth bag area, is the serial number corresponding to the first time the value is greater than 0 except for the starting position in the first-order difference sequence of the local information entropy of the th cloth bag area, is the total number of reference images, is the total number of cloth bag areas, is the linear normalization function, is the product symbol, is the absolute value symbol.

[0015] Through the above technical solution, by processing the first-order difference sequence of the local information entropy, the starting point and subsequent change trend of the change of the local information entropy of the cloth bag area can be captured, and then the entropy change stability of each cloth bag area can be calculated, and the change trend of the local information entropy of each cloth bag area in a continuous multi-frame image can be accurately quantified.

[0016] Furthermore, the gradient of the local information entropy of each cloth bag area in each reference image is determined based on the following method:

[0017] Obtain the local information entropy of each cloth bag area and its adjacent cloth bag areas in each reference image;

[0018] In each reference image:

[0019] The direction of the gradient of the local information entropy of each cloth bag area is: from the center of the cloth bag area to the adjacent cloth bag area with the largest local information entropy;

[0020] The value of the gradient of the local information entropy of each cloth bag area is: the normalized value of the difference between the information entropy of the cloth bag area and the information entropy of the adjacent cloth bag area with the largest local information entropy.

[0021] The above technical solution determines the gradient based on the local information entropy of each cloth bag area and its adjacent cloth bag areas, fully considering the mutual correlation between the cloth bag areas, and can more comprehensively capture the information transmission and interaction between the cloth bag areas, so as to more accurately evaluate the state of a single cloth bag area. The change of the information entropy of the adjacent cloth bag area can reflect the diffusion path and influence range, which helps to distinguish the true leakage source and the area affected by diffusion.

[0022] Further, the spatial vectors between each cloth bag area and its adjacent cloth bag areas are determined based on the following method:

[0023] For each cloth bag area and any one of its adjacent cloth bag areas, the direction of the spatial vector is from the center of the adjacent cloth bag area to the center of this cloth bag area; the value of the spatial vector is the normalized value of the Euclidean distance between the center of the adjacent cloth bag area and the center of this cloth bag area.

[0024] The above technical solution clearly constructs the spatial position association between cloth bag areas, can intuitively display the relative positions between cloth bag areas, and provides a basis for subsequent analysis of the interaction between cloth bag areas. When studying the exhaust gas diffusion path, the direction of the spatial vector can help determine from which adjacent areas the exhaust gas may flow into the target cloth bag area, which is helpful for understanding the gas flow pattern in the whole system. And through the value of the spatial vector, the quantification of the spatial distance between cloth bag areas is realized. The spatial distance will affect the diffusion speed and degree of the exhaust gas between cloth bags. When judging the influence range of exhaust gas leakage, the quantified spatial vector value can be used as an important reference. The adjacent cloth bag areas with a closer distance are more likely to be affected by the leaked exhaust gas, and the normalized value can more accurately measure the potential degree of this influence.

[0025] Further, the method for obtaining the local information entropy of each cloth bag area in each frame of image is as follows:

[0026]

[0027] In the formula, is the local information entropy of the th cloth bag area in the th frame of image, is the probability that the th gray value in the th frame of image appears in the th cloth bag area, is the total number of types of gray values in the th frame of image.

[0028] The above technical solution quantifies the degree of disorder of the gray information of the cloth bag area through the information entropy formula to obtain the local information entropy. The exhaust gas leakage may be a dynamically changing process. By continuously collecting multiple frames of images and calculating the local information entropy of the cloth bag area in each frame of image, the changing trend of the local information entropy over time can be observed.

[0029] Further, each cloth bag area and its adjacent cloth bag areas are determined based on the following method:

[0030] For each cloth bag area, calculate the distance between the center of this cloth bag area and the centers of other cloth bag areas, and use the other cloth bag areas with the distance less than or equal to the preset distance threshold as the adjacent cloth bag areas of this cloth bag area.

[0031] Further, the method for obtaining multiple cloth bag areas is as follows: use the Hough circle detection method or the row-column segmentation method to divide the flower plate plane of the bag filter into multiple cloth bag areas.

[0032] Further, the method for detecting waste gas leakage according to the possible degree of leakage is as follows:

[0033] If the possible degree of leakage of a certain cloth bag area is greater than the preset leakage possible degree threshold, the cloth bag corresponding to this cloth bag area is the cloth bag with waste gas leakage; if the possible degree of leakage of a certain cloth bag area is not greater than the preset leakage possible degree threshold, the cloth bag corresponding to this cloth bag area is the cloth bag without waste gas leakage.

[0034] Further, before calculating the local information entropy of each cloth bag area in each frame of image, perform the following operation: convert all the collected multiple frames of images into grayscale images.

[0035] Further, the multiple reference images of the current frame image are: the current frame image and a total of frames of images before it, which is a preset value.

[0036] The present invention has the following effects:

[0037] By continuously collecting multiple frames of images, calculating the local information entropy of each cloth bag area and comprehensively judging by combining multiple indicators such as entropy change stability and vector convergence degree, the present invention not only considers the change of the local information entropy of each cloth bag area from the time dimension, but also combines the spatial correlation between each cloth bag area and its adjacent cloth bag areas from the spatial dimension. Finally, the possible degree of leakage of each cloth bag area is obtained, and the cloth bags with leakage are determined based on the accurate possible degree of leakage, reducing the possibility of misjudgment when waste gas diffuses rapidly, improving the accuracy and reliability of waste gas leakage detection, and ensuring the dust removal effect of the bag filter. Brief Description of the Drawings

[0038] Figure 1 is a schematic flow chart of the method of the present invention;

[0039] Figure 2 is a schematic structural diagram of the bag filter of the present invention;

[0040] Figure 3 is a schematic diagram of the flower plate plane of the bag filter of the present invention. Detailed Embodiments

[0041] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention.

[0042] As Figure 1 shown, a method for detecting exhaust gas leakage of a bag filter provided by the present invention includes steps S1 - S6:

[0043] S1: Collect multiple frames of images of the bag filter, and each frame of image includes multiple bag regions.

[0044] Since the bag filter adopts an external filtration dust removal structure with air blown from bottom to top, the bags all have circular bag mouths, and the specific structure is as Figure 2 shown. After the exhaust gas enters from the bottom of the bag filter, it escapes upward and passes through the bags for filtration and then is discharged. The top of the bag filter is a perforated plate plane. In this step, an industrial camera is installed at the exact center of the top of the bag filter to ensure that the shooting angle is perpendicular to the perforated plate plane of the bag filter and all the bag mouths on the perforated plate plane are completely captured. The arrangement of the bag mouths on the perforated plate plane is as Figure 3 shown. In order to accurately obtain and analyze the exhaust gas leakage situation of each bag, it is necessary to accurately divide each bag region on the perforated plate plane.

[0045] Therefore, continuously collect multiple frames of images of the perforated plate plane of the bag filter. Each frame of image is as Figure 3 shown and contains multiple bag regions (the regions where the bag mouths of each bag are located). In the multiple frames of images, the size and dimensions of each bag region are fixed because the inherent structure of the perforated plate plane of the bag filter does not change. To reduce the computational complexity, each frame of image is grayscaled, and then all the bag regions in each frame of image are accurately divided.

[0046] In one embodiment, the perforated plate plane of the bag filter is divided into multiple bag regions by using the Hough circle detection method or the row - column segmentation method.

[0047] For the Hough circle detection method, the specific implementation process is as follows: First, use the Hough circle detection algorithm to accurately identify all the bag mouths on the perforated plate plane of the bag filter. For each detected circle corresponding to a bag mouth, extract its center coordinates. These center coordinates will be used as the centers of the subsequent divided bag regions respectively, and then decide to select a square frame or a circular frame to delimit the bag regions. If there are high requirements for computational efficiency and shape regularity, the square frame may be more advantageous in subsequent calculations; if more attention is paid to the degree of fitting with the actual shape of the bag, then the circular frame is selected.

[0048] In this step, a circular frame is selected to delimit the bag area. For each center, starting from a smaller initial radius of 1, the radius is gradually increased by a step of 1 each time for expansion. It is checked whether the expanded circular frame overlaps with other circular frames until the largest non-overlapping radius is found. Based on each center and the largest non-overlapping radius, after obtaining the preliminary circular frames for each bag area, statistical analysis is performed on the sizes of the circular frames, and the average radius of all circular frames is calculated. For circular frames with sizes deviating from the average value, they are adjusted according to the difference between the size and the average radius, while ensuring that the adjusted circular frames still do not overlap with other circular frames. Each adjusted circular frame is taken as a bag area. In this way, the diaphragm plane of the bag filter in each frame of the image is divided into multiple bag areas.

[0049] For the row-column segmentation method, the specific implementation process is as follows:

[0050] An edge detection algorithm is used to perform edge detection on each frame of the image to identify the edges of each bag mouth in each frame of the image, and the bag mouth contour is further strengthened through morphological operations (such as dilation and erosion) to facilitate accurate identification of the position of the bag mouth, and the central coordinates of the closed area surrounded by the edges of each bag mouth are obtained, and these central coordinates represent the position of each bag mouth in the image.

[0051] The central coordinates of all the identified bag mouths are analyzed. First, the central coordinates of all the bag mouths are sorted according to the abscissa (horizontal direction), and the bag mouths with similar abscissas are grouped into the same row. For each row, the Euclidean distance between the central coordinates of adjacent bag mouths is calculated, and these distance values are statistically analyzed. The distance value with the highest frequency of occurrence is taken as the row spacing of this row, because the bags are usually arranged evenly on the diaphragm and the row spacing is relatively consistent. Similarly, the central coordinates of all the bag mouths are sorted according to the ordinate (vertical direction), columns are divided, and the column spacing is calculated.

[0052] According to the calculated row spacing and column spacing, starting from the central coordinates of the first bag mouth in the first row, along the horizontal direction, with the row spacing as the interval, horizontal lines are fitted through this starting point and subsequent equally spaced points. These horizontal lines cover the entire horizontal range of the image to complete the drawing of the row segmentation lines. Similarly, starting from the central coordinates of the first bag mouth in the first column, along the vertical direction, with the column spacing as the interval, vertical lines are fitted to cover the entire vertical range of the image to complete the drawing of the column segmentation lines. Since the drawn horizontal segmentation lines and vertical segmentation lines intersect with each other to form grid areas, each grid area is a bag area. In this way, the diaphragm plane of the bag filter in each frame of the image is divided into multiple bag areas.

[0053] S2: Obtain the local information entropy sequence of each bag area in the current frame of the image.

[0054] Under normal circumstances, the grayscale of all cloth bag areas is relatively uniform and stable, and the local information entropy calculated based on the grayscale values of pixel points in each cloth bag area is stable. When there is waste gas leakage in a certain cloth bag area, the waste gas will cause the local information entropy of that cloth bag area to increase. Therefore, analyzing the changes in the local information entropy of each cloth bag area in consecutive frame images is of great significance for waste gas leakage detection.

[0055] In one embodiment, the local information entropy of each cloth bag area in each frame of image is calculated according to the information entropy calculation formula:

[0056]

[0057] In this formula, is the local information entropy of the th cloth bag area in the th frame of image, is the probability that the th gray value in the th frame of image appears in the th cloth bag area, is the total number of types of gray values in the th frame of image.

[0058] Taking any frame of image as the current frame image, after obtaining the local information entropy of each cloth bag area in each frame of image, in order to more accurately analyze the change trend of the local information entropy of each cloth bag area in the time dimension entropy, the current frame image and a total of frames of images before it are used as the reference images of the current frame image (including the current frame image itself), is a preset value, and the empirical value is 10. If the number of images before the current frame is less than frames, the current frame image is postponed frame by frame backward until frames are satisfied, and then the current frame image is reset, and the reference images of the current frame image are determined according to the same method.

[0059] In one embodiment, the method for obtaining the local information entropy sequence of each cloth bag area in the current frame image is as follows: Obtain the local information entropy of the cloth bag area in all reference images of the current frame image, there are 10 in total, and they are sequentially formed into a local information entropy sequence in chronological order, and this local information entropy sequence is used as the local information entropy sequence of the cloth bag area in the current frame image.

[0060] S3: Determine the entropy change stability of each cloth bag area in the current frame image.

[0061] In the current frame image, it is of great significance to determine the entropy change stability of each cloth bag area. When the exhaust gas escape speed is too fast, a special situation will occur: if there is exhaust gas leakage in a certain cloth bag, the cloth bags without exhaust gas leakage near this cloth bag will be affected by this cloth bag, resulting in an increase in the local information entropy of the cloth bags without exhaust gas leakage nearby, and then being misidentified as having exhaust gas leakage.

[0062] For the cloth bag with actual exhaust gas leakage, after the local information entropy increases, it will maintain at a stable high level. For the cloth bag without exhaust gas leakage, due to the continuous dilution of the escaped exhaust gas, its local information entropy is slightly lower than that of the cloth bag with exhaust gas leakage and the change is unstable.

[0063] Based on the above analysis, by calculating the entropy change stability of each cloth bag area in the current frame image, the change of the local information entropy of each cloth bag area can be deeply analyzed. The entropy change stability can be used as a key indicator to reflect the possibility of exhaust gas leakage in the cloth bag area. Specifically:

[0064] When the entropy change stability of a certain cloth bag area is low, it means that the change of its local information entropy is unstable, and it is very likely to be affected by the diffusion of the cloth bag area with exhaust gas leakage around it, rather than having its own leakage; when the entropy change stability of a certain cloth bag area is high, it indicates that the change of its local information entropy is relatively stable and at a high level, then the possibility of exhaust gas leakage in this cloth bag area is greater.

[0065] In one embodiment, the entropy change stability of each cloth bag area in the current frame image is obtained based on the local information entropy sequence of each cloth bag area. Specifically, first obtain the first-order difference sequence of the local information entropy sequence of each cloth bag area. The first-order difference sequence is a sequence obtained by subtracting the former of every two adjacent local information entropies in the original sequence from the latter. For example, the local information entropy sequence of a certain cloth bag area is: 2.5, 2.7, 2.6, 2.8, 2.9, then the first-order difference sequence is 0.2, -0.1, 0.2, 0.1. Then record the serial number corresponding to the value when the value greater than 0 first appears in the first-order difference sequence except for the starting position, and at the same time, this value corresponds to two consecutive frame images in the original sequence.

[0066] For example, in the first-order difference sequence 0.2, -0.1, 0.2, 0.1, when the value greater than 0 first appears except for the starting position, the corresponding serial number is 2, that is, the serial number of -0.1 in the first-order difference sequence, and -0.1 is calculated from the two local information entropies 2.6 and 2.7.

[0067] When a value greater than 0 appears for the first time in the first-order difference sequence of a bag area except for the starting position, the entropy of the bag area increases in two consecutive frame images (corresponding to the appearance of a value greater than 0). The local information entropy of the bag area shows a significant difference in these two consecutive frame images, and it can be determined that the entropy change stability of the bag area is low.

[0068] In one embodiment, the entropy change stability of each bag region is determined based on the following formula:

[0069]

[0070] In this formula, For the The entropy change stability of the bag region, For the The first-order difference sequence of the local information entropy of the bag region values, For the The number corresponding to the first time the value of the first-order difference sequence of the local information entropy of the bag region is greater than 0 except for the starting position, which marks the first The starting point where the local information entropy of the bag area begins to show an upward trend, that is, the starting point of entropy increase. This starting point is crucial for analyzing the change of the local information entropy of the bag area. is the total number of reference images of the current frame image, is the total number of bag areas, is a linear normalization function, The value of is mapped to between 0 and 1. is the multiplication symbol, is the absolute value symbol.

[0071] In this formula, is the first The absolute entropy change stability of the bag area. It is The first-order difference sequence of the local information entropy of the bag region is The absolute values ​​of all the values ​​are then multiplied together to comprehensively consider the cumulative effect of the entropy change in the bag area during this period. The use of absolute values ​​ensures that whether the entropy change is positive or negative, it can be reasonably included in the calculation, reflecting the overall situation of the magnitude of the entropy change. If the local information entropy of the bag area changes more dramatically during this period, then the multiplication result will be larger, otherwise it will be smaller. It reflects the starting point of entropy increase. To The time span of a reference image. This time span, combined with the cumulative effect of local information entropy, jointly affects the stability of absolute entropy change. A longer time span results in a larger cumulative entropy change, increasing the value of the stability of absolute entropy change, indicating that the entropy change of the cloth bag area is more complex during this period and may be in an unstable state; conversely, a shorter time span or a smaller cumulative entropy change indicates that the entropy change of the cloth bag is relatively stable.

[0072] In this formula, represents the mean value of the stability of absolute entropy change for all cloth bag areas in the current frame image. This mean value serves as a reference benchmark to measure the gap between the stability of absolute entropy change of the

[0073] th cloth bag area and the overall level. The overall formula is the difference between the stability of absolute entropy change of the th cloth bag image in the current frame image and the mean value of the stability of absolute entropy change for all cloth bag areas. If the stability of absolute entropy change of the th cloth bag area is much higher than the mean value, the larger this difference, the

[0074] larger the value, indicating that the entropy change of this cloth bag area is more prominent in the whole and is more likely to be in a relatively unstable state, meaning that this cloth bag area is more likely to have waste gas leakage, and vice versa. S4: Calculate the degree of vector convergence for each cloth bag area in the current frame image.

[0075] Since the local information entropy of the cloth bag area reflects the degree of disorder of the gray-scale information in the cloth bag area. When the cloth bag is intact and the waste gas is stably diluted, the gas distribution is relatively more regular and orderly, and the local information entropy of the cloth bag area corresponding to the intact cloth bag is lower. For a cloth bag with waste gas leakage, the internal gas flow and mixing are more complex and chaotic, so the local information entropy of the cloth bag area corresponding to the cloth bag with waste gas leakage is higher.

[0076] In the current frame image, when the waste gas escapes to an intact cloth bag, the dilution speed of the waste gas remains stable. In this case, the overall stability of entropy change for all cloth bag areas in the entire dust collector is still at a relatively high level. However, the local information entropy of the cloth bag area corresponding to the intact cloth bag is slightly lower than that of the cloth bag area corresponding to the cloth bag with waste gas leakage. It is difficult to accurately distinguish between intact cloth bags and cloth bags with waste gas leakage solely based on the magnitude of local information entropy.

[0077] Therefore, the following judgment method is set in this step:

[0078] During the operation of the bag filter, the state of the bag can be judged based on the gradient characteristics of the local information entropy around it. When a certain bag leaks, it serves as a leakage source, and the waste gas inside the bag will gradually disperse to the surrounding adjacent bags. From the opposite perspective, it is the waste gas converging from the surrounding adjacent bags to the bag where the leakage source is located. Therefore, it can be observed from consecutive frame images that if the gradients of the local information entropy of the adjacent bag regions around a certain bag region show a highly ordered and regular distribution, specifically manifested as the gradients of the local information entropy of the surrounding adjacent bag regions converging towards the center of this bag region, it indicates that the bag corresponding to this bag region is very likely to be the leakage source. Because this regular convergence trend means that the process of waste gas diffusion from the leaking bag to the surrounding area presents a specific pattern. In this specific bag region, the gradient value of its local information entropy is greater than that of the surrounding bag regions, highlighting the characteristic changes caused by the waste gas diffusion generated in this bag region.

[0079] Conversely, if in consecutive frame images, the gradients of the local information entropy of the adjacent bag regions around a certain bag region show a chaotic distribution without any pattern, or although there is a divergent trend, it does not converge regularly towards the center of a specific bag region but diverges randomly in all directions without obvious regular changes, then the bag corresponding to this bag region is more likely to be a bag without leakage, indicating that the flow of waste gas between these bags is in a normal, disordered state and is not interfered by bag leakage.

[0080] Therefore, in this step, the gradient of the local information entropy of each bag region in each reference image of the current frame image is determined, as well as the spatial vector between each bag region and its adjacent bag regions, and the vector convergence degree of each bag region is determined based on the gradient and the spatial vector.

[0081] In one embodiment, the method for determining the adjacent bag regions of each bag region is as follows:

[0082] Set the distance threshold to 200 (empirical value). For each bag region, calculate the Euclidean distance between the center of other bag regions and the center of this bag region, and take the other bag regions with the Euclidean distance less than or equal to 200 as all the adjacent bag regions of this bag region.

[0083] In one embodiment, the method for determining the gradient of the local information entropy of each bag region in each reference image of the current frame image is as follows:

[0084] Obtain the local information entropy of each bag region and its adjacent bag regions in each reference image;

[0085] In each reference image, the direction of the gradient of the local information entropy of each cloth bag area is: from the center of the cloth bag area to the adjacent cloth bag area with the largest local information entropy; the value of the gradient of the local information entropy of each cloth bag area is: the normalized value of the difference between the information entropy of the cloth bag area and the information entropy of the adjacent cloth bag area with the largest local information entropy. By this method, the gradient of the local information entropy of each cloth bag area in each reference image of the current frame image is obtained, and the gradient is a vector.

[0086] In one embodiment, the spatial vector between each cloth bag area and its adjacent cloth bag areas is determined as follows:

[0087] For each cloth bag area and any one of its adjacent cloth bag areas, the direction of the spatial vector is: from the center of the adjacent cloth bag area to the center of the cloth bag area; the value of the spatial vector is: the normalized value of the Euclidean distance between the center of the adjacent cloth bag area and the center of the cloth bag area. For example, the th cloth bag area and its th adjacent cloth bag area, the direction of the spatial vector is: from the center of the th adjacent cloth bag area to the center of the th cloth bag area, and the magnitude of the spatial vector is the normalized value of the Euclidean distance between the center of the th cloth bag area and the center of its th adjacent cloth bag area.

[0088] In one embodiment, the vector convergence degree of each cloth bag area is calculated based on the following formula:

[0089]

[0090] In this formula, is the vector convergence degree of the th cloth bag area, is the gradient of the local information entropy of the th cloth bag area in the th reference image of the th adjacent cloth bag area, is the spatial vector between the th cloth bag area and the th adjacent cloth bag area, is the total number of adjacent cloth bag areas of the th cloth bag area, is the total number of reference images, is for calculating the modulus length, is for taking the absolute value, is the natural exponential function, is the normalization function.

[0091] In this formula, is and the degree of direction consistency of these two vectors (both the gradient and the spatial vector are vectors). The greater the degree of direction consistency, the closer the gradient direction of the local information entropy is to the spatial vector direction, which means that for the th cloth bag area and its th adjacent cloth bag area, the higher the direction consistency between the change trend of the exhaust gas state and the spatial position, showing regularity, indicating that the adjacent cloth bag areas around the th cloth bag area show an orderly converging state in terms of direction, and the cloth bag corresponding to the th cloth bag area is more likely to be the leakage source.

[0092] In this formula, is and the numerical difference between these two vectors. The smaller this difference, the more likely it is that the cloth bag corresponding to the th cloth bag area has a leakage.

[0093] When there is exhaust gas leakage from the cloth bag, the gradient of the local information entropy and the spatial vector not only show regularity in direction but also in numerical value. This close correlation conforms to the physical process of exhaust gas diffusion when the cloth bag leaks, that is, with the leaking cloth bag as the center, the local information entropy changes rapidly within the surrounding limited space, making the gradient value representing the change of the exhaust gas state tend to match the vector value representing the spatial position. Therefore, by the negative exponential function in performs reverse weighting on

[0094]

[0095] the higher the degree of vector convergence of the

[0095] S5: Comprehensively evaluate the possible degree of leakage of each cloth bag area in the current frame image.

[0096] The leakage probability of each cloth bag area in the current frame image is jointly determined by the local information entropy, entropy change stability, and vector convergence degree of each cloth bag area. If the local information entropy, entropy change stability, and vector convergence degree of a cloth bag area are higher, then the probability that the cloth bag in this cloth bag area has waste gas leakage is higher.

[0097] Therefore, the product of the local information entropy, entropy change stability, and vector convergence degree of each cloth bag area is used as the leakage probability of each cloth bag area. For the th cloth bag area in the current frame image, the leakage probability satisfies the following relational expression:

[0098]

[0099] In the formula, is the leakage probability of the th cloth bag area in the current frame image, is the local information entropy (value after linear normalization) of the th cloth bag area in the current frame image, is the entropy change stability of the th cloth bag area in the current frame image, is the vector convergence degree of the th cloth bag area in the current frame image.

[0100] S6: Perform waste gas leakage detection according to the magnitude of the leakage probability of each cloth bag area.

[0101] Set the threshold of the leakage probability as (empirical value). If the leakage probability of a certain cloth bag area in the current frame image is greater than 0.5, waste gas leakage occurs in this cloth bag area; if the leakage probability of a certain cloth bag area is not greater than 0.5, waste gas leakage does not occur in this cloth bag area. Further, mark the cloth bag areas with waste gas leakage to timely remind the maintenance personnel to check the bag filter, and replace the cloth bags with waste gas leakage in a timely manner according to the marks.

[0102] Although this specification has shown and described multiple embodiments of the present invention, it is obvious to those skilled in the art that such embodiments are provided only by way of example. Those skilled in the art will think of many changes, alterations, and alternative ways without departing from the spirit and idea of the present invention. It should be understood that various alternative solutions to the embodiments of the present invention described herein can be adopted in the process of practicing the present invention.

Claims

1. A method for detecting exhaust gas leakage in a bag filter, characterized in that: include: Continuously collect multiple frames of images of the flower plate plane of the bag filter, wherein the flower plate plane includes multiple bag areas, and calculate the local information entropy of each bag area in each frame of the image; Taking any frame image as the current frame image, determining the local information entropy sequence of each bag area of ​​the current frame image according to multiple reference images of the current frame image; The entropy change stability of each bag region is determined according to the first-order difference sequence of the local information entropy sequence of each bag region; the degree of vector convergence is calculated according to the gradient of the local information entropy of each bag region in each reference image and the spatial vector of each bag region and the adjacent bag region: , For the The degree of vector convergence in the bag area, For the The first The adjacent bag area is in The gradient of the local information entropy of the reference image, For the The bag area and The space vector of adjacent bag regions, For the The total number of adjacent bag areas of a bag area, is the total number of reference images, To find the model length, To find the absolute value, is the natural exponential function, is the normalization function; The product of the local information entropy, entropy change stability and vector convergence degree of each bag area is used as the leakage possibility of each bag area, and exhaust gas leakage detection is performed according to the leakage possibility; The gradient of the local information entropy of each bag region in each reference image is determined as follows: Obtain the local information entropy of each bag region and its adjacent bag regions in each reference image; In each reference image: The direction of the gradient of the local information entropy of each bag area is: the center of the bag area points to the adjacent bag area with the largest local information entropy; The value of the gradient of the local information entropy of each bag area is: the normalized value of the difference between the information entropy of the bag area and the information entropy of the adjacent bag area with the maximum local information entropy.

2. The exhaust gas leakage detection method for bag filter according to claim 1, characterized in that: The entropy stability of each bag area is determined based on the following formula: ； In the formula, For the The entropy change stability of the bag region, For the The first-order difference sequence of the local information entropy of the bag region values, For the The number corresponding to the first time the value of the first-order difference sequence of the local information entropy of the bag region is greater than 0 except for the starting position, is the total number of reference images, is the total number of bag areas, is the linear normalization function, is the multiplication symbol, is the absolute value symbol.

3. The exhaust gas leakage detection method for bag filter according to claim 1, characterized in that: The space vectors of each bag area and the adjacent bag areas are determined based on the following method: For each bag area and any of its adjacent bag areas, the direction of the space vector is: the center of the adjacent bag area points to the center of the bag area; the value of the space vector is: the normalized value of the Euclidean distance between the center of the adjacent bag area and the center of the bag area.

4. The exhaust gas leakage detection method for bag filter according to claim 1, characterized in that: The method for obtaining the local information entropy of each bag area in each frame image is: ; In the formula, For the The bag area is in The local information entropy in the frame image, For the The first The gray value is The probability of appearing in the bag area is For the The total number of grayscale value types of the frame image.

5. The exhaust gas leakage detection method for bag filter according to claim 1, characterized in that: Each bag area and its adjacent bag areas are determined based on the following method: For each bag area, the distance between the center of the bag area and the centers of other bag areas is calculated, and other bag areas whose distances are less than or equal to a preset distance threshold are taken as adjacent bag areas of the bag area.

6. The exhaust gas leakage detection method for bag filter according to claim 1, characterized in that: The method for obtaining multiple bag areas is: using the Hough circle detection method or the row-column segmentation method to divide the flower plate plane of the bag filter into multiple bag areas.

7. The exhaust gas leakage detection method for bag filter according to claim 1, characterized in that: The methods for exhaust gas leak detection according to the possible degree of leakage are: If the possibility of leakage in a certain bag area is greater than the preset possibility of leakage threshold, the bag corresponding to the bag area is a bag with exhaust gas leakage; if the possibility of leakage in a certain bag area is not greater than the preset possibility of leakage threshold, the bag corresponding to the bag area is a bag with no exhaust gas leakage.

8. The exhaust gas leakage detection method for bag filter according to claim 1, characterized in that: Before calculating the local information entropy of each bag region in each frame of image, the following operations are performed: all the acquired multiple frames of images are converted into grayscale images.

9. The exhaust gas leakage detection method for bag filter according to claim 1, characterized in that: The multiple reference images of the current frame image are: the current frame image and the total number of images before it Frame image, is the default value.

Images