CN110544261A - Blast furnace tuyere coal injection state detection method based on image processing - Google Patents
Blast furnace tuyere coal injection state detection method based on image processing Download PDFInfo
- Publication number
- CN110544261A CN110544261A CN201910834054.0A CN201910834054A CN110544261A CN 110544261 A CN110544261 A CN 110544261A CN 201910834054 A CN201910834054 A CN 201910834054A CN 110544261 A CN110544261 A CN 110544261A
- Authority
- CN
- China
- Prior art keywords
- image
- coal
- blast furnace
- adopting
- tuyere
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 239000003245 coal Substances 0.000 title claims abstract description 104
- 238000002347 injection Methods 0.000 title claims abstract description 34
- 239000007924 injection Substances 0.000 title claims abstract description 34
- 238000012545 processing Methods 0.000 title claims abstract description 33
- 238000001514 detection method Methods 0.000 title claims abstract description 21
- 239000004484 Briquette Substances 0.000 claims abstract description 31
- 238000001914 filtration Methods 0.000 claims abstract description 21
- 238000013528 artificial neural network Methods 0.000 claims abstract description 13
- 230000009466 transformation Effects 0.000 claims abstract description 12
- 238000000034 method Methods 0.000 claims description 40
- 230000001186 cumulative effect Effects 0.000 claims description 20
- 238000005315 distribution function Methods 0.000 claims description 10
- 125000001475 halogen functional group Chemical group 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 10
- 238000009826 distribution Methods 0.000 claims description 5
- 230000006870 function Effects 0.000 claims description 5
- CIWBSHSKHKDKBQ-JLAZNSOCSA-N Ascorbic acid Chemical compound OC[C@H](O)[C@H]1OC(=O)C(O)=C1O CIWBSHSKHKDKBQ-JLAZNSOCSA-N 0.000 claims description 3
- 230000000877 morphologic effect Effects 0.000 claims description 2
- 230000002159 abnormal effect Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 11
- 230000011218 segmentation Effects 0.000 description 10
- 238000005457 optimization Methods 0.000 description 6
- 238000007664 blowing Methods 0.000 description 5
- 238000013016 damping Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 239000000243 solution Substances 0.000 description 4
- 238000003723 Smelting Methods 0.000 description 3
- 230000000903 blocking effect Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 2
- 229910000831 Steel Inorganic materials 0.000 description 2
- 230000032683 aging Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000003708 edge detection Methods 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 239000000843 powder Substances 0.000 description 2
- 239000010959 steel Substances 0.000 description 2
- 241001270131 Agaricus moelleri Species 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 239000000571 coke Substances 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 229910052742 iron Inorganic materials 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000007790 solid phase Substances 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 238000002834 transmittance Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
the invention discloses a blast furnace tuyere coal injection state detection method based on image processing, which comprises the following steps: denoising the blast furnace tuyere image by adopting a mode of combining frequency domain filtering and spatial domain filtering to obtain a denoised image; processing the de-noised image by adopting an improved image enhancement algorithm combining self-adaptive gray stretching and power transformation to obtain a clear image; processing the clear image by adopting a binarization algorithm combining a global threshold and a local threshold to obtain a binarization image; fitting the tuyere region in the binary image by adopting a least square ellipse fitting mode, so as to avoid the phenomenon of adhesion between the coal briquette and the furnace wall caused by overlarge coal injection amount; segmenting a coal gun region in a clear image through an improved multi-scale full convolution neural network, and subtracting an integral connected domain of a coal gun and a coal briquette in a binary image from a coal gun region detected by the full convolution neural network to obtain a coal briquette region; and judging whether the branch pipe coal injection of the blast furnace is blocked in an abnormal state or not according to the obtained area information of the coal briquette area.
Description
Technical Field
the invention relates to the technical field of image processing, in particular to a blast furnace tuyere coal injection state detection method based on image processing.
Background
The problem of serious aging of the optical filter is considered when the blast furnace tuyere camera is in a high-temperature and dusty severe industrial environment for a long time, so that gray scales of all areas of a shot tuyere image are unbalanced and blurred. In addition, there is noise interference in the image generation, transmission and conversion processes, and the image needs to be analyzed and processed accordingly. The gray information of the coal guns, the coal briquette and the inner wall area in the tuyere image is similar, and the traditional threshold segmentation algorithm is difficult to apply. In the actual production process, the coal guns need to be disassembled for maintenance during the damping down of the blast furnace, and the positions of the coal guns in the images before and after damping down may be changed due to re-installation. The above characteristics bring difficulties to the extraction of the coal briquette region and require the detection method to have adaptivity.
in the existing technology for detecting the coal injection state of the blast furnace tuyere, a temperature difference method mainly detects according to the temperature change before and after blockage, but the temperature difference judgment criterion is difficult to accurately conclude due to the interference of factors such as seasons, wind directions and the like. The optical detection method requires that the detected area has good light transmittance and the detected fluid has proper solid-phase concentration, which becomes a main application obstacle. The fixed background template method based on image processing does not consider the potential position change of the coal gun in the wind repairing process, and influences the detection accuracy.
disclosure of Invention
according to the problems in the prior art, the invention discloses a blast furnace tuyere coal injection state detection method based on image processing, which comprises the following specific processes:
Denoising the blast furnace tuyere image by adopting a mode of combining frequency domain filtering and spatial domain filtering to obtain a denoised image;
Processing the de-noised image by adopting an improved image enhancement algorithm combining self-adaptive gray stretching and power transformation to obtain a clear image;
Processing the clear image by adopting a binarization algorithm combining a global threshold and a local threshold to obtain a binarization image;
fitting the tuyere region in the binary image by adopting a least square ellipse fitting mode, so that the phenomenon of 'adhesion' between the tuyere edge and the coal briquette caused by overlarge coal injection amount can be avoided;
Segmenting a coal gun region in a clear image through an improved multi-scale full convolution neural network, and subtracting an integral connected domain of a coal gun and a coal briquette in a binary image from a coal gun region detected by the full convolution neural network to obtain a coal briquette region;
and judging whether the area information of the coal briquette area is smaller than a threshold value T or not according to the obtained area information of the coal briquette area, if the continuous k frames are smaller than the threshold value T, judging that the blast furnace branch pipe is in a blocked state, and if not, judging that the blast furnace branch pipe is still in a coal injection state.
further, when the blast furnace tuyere image is subjected to denoising processing, firstly, the noise type existing in the blast furnace tuyere image is analyzed, wherein the noise type comprises horizontal direction stripe noise and Gaussian noise.
Selecting a frequency domain filtering method to process the stripe noise in the image: firstly, Fourier transformation is carried out on a tuyere image to obtain a spectrogram F (u, v), cumulative distribution functions in the horizontal direction and the vertical direction are respectively calculated for the spectrogram F (u, v), whether a function value at a midpoint of the cumulative distribution function in the vertical direction is greater than a threshold value or not is judged, if so, stripe noise exists in the image, otherwise, the stripe noise does not exist;
If the blast furnace tuyere image has stripe noise, calculating the abscissa u0 of the quadratic peak frequency point of the cumulative distribution function in the horizontal direction, constructing a self-adaptive second-order Butterworth filter H (u, v) according to u0 for filtering, and performing inverse Fourier transform on the filtered spectrogram to obtain a restored image
Filtering out light spots and Gaussian noise in the image by adopting median filtering;
the digital label area in the image is eliminated by using a morphological open-close filter.
Further, when the de-noised image is subjected to enhancement processing, cumulative histogram distribution of the image is firstly obtained, gray values corresponding to the cumulative histogram when the frequency of the gray values meets a set threshold value are selected as a window bottom b and a window top c of gray stretching, and the image is subjected to gray stretching; analyzing the histogram characteristics of the image with the halo phenomenon, and eliminating the halo part in the image through power transformation.
Further, firstly, classifying pixels in the image into a high gray level region, a middle gray level region and a low gray level region; secondly, dividing the high-gray area and the low-gray area by adopting a global threshold maximum inter-class difference method; and judging the remaining middle gray level area through a Gaussian weighted local threshold algorithm to obtain a final binary image.
Furthermore, the multi-scale full convolution neural network model is improved on the basis of a full convolution neural network (FCN), and mainly comprises an encoding-decoding full convolution network and an optimization network, wherein in the encoding-decoding full convolution network, the shallow layer characteristics of the network mainly comprise texture, color and edge information, the deep layer characteristics of the network comprise high-level semantic information, and different layer networks are cascaded to predict the segmentation result of the coal gun; the coding-decoding full convolution network is also added with a pyramid pooling layer, and feature extraction is carried out from different scales;
The optimization network takes the prediction result of the coding-decoding full convolution network as input, wherein the optimization network is composed of six convolution layers, two of the convolution layers are hollow convolution, and the edge area can be optimized for the segmentation result.
By adopting the technical scheme, the method for detecting the coal injection state of the blast furnace tuyere based on image processing judges whether stripe noise exists in an image or not by analyzing the characteristics of a spectrogram, if the stripe noise exists, the stripe noise is removed by a self-adaptive Butterworth filter, the cut-off frequency of the filter selects the pixel number corresponding to the sub-peak point of the cumulative distribution function of the spectrogram in the horizontal direction, and the stripe noise in the image can be well removed by the method; aiming at the gray level unbalance of an image caused by the aging of the optical filter, the image is processed by a self-adaptive gray level stretching algorithm of the cumulative histogram, the gray levels with the cumulative content of 10 percent and 95 percent in the cumulative distribution histogram are respectively selected at the bottom and the top of the stretched window, and the contrast of each area in the image is obviously improved through the gray level stretching treatment, thereby laying a foundation for the subsequent treatment. In the image binarization processing, the processing result of the global threshold value method can well reflect the overall gray level distribution condition of the image, but the detail information of the image is easy to ignore. The local threshold can reflect the detail information of the image, but is easily affected by 'noise' during the segmentation, and the phenomenon of inaccurate segmentation is generated. A method combining a global threshold method and a local threshold method is provided, high-gray and low-gray pixel points are processed by using a global threshold, and the rest of pixel points are processed by using a local threshold, so that the method can well reflect detail information in an image from a processing result, is not influenced by 'noise', and has a good processing effect. Aiming at the phenomenon that the gray information of a coal gun and the gray information of a coal briquette are close and the traditional threshold segmentation algorithm is difficult to separate the coal gun and the coal briquette, the coal gun needs to be disassembled and overhauled in the process of repairing wind, and the position of the coal gun can be changed before and after damping down due to the fact that the coal gun is installed again. And automatically segmenting a coal gun area through a full convolution neural network, and subtracting the whole connected domain of the coal gun and the coal briquette in the binary image from the coal gun area detected by the full convolution neural network to obtain the coal briquette area. And judging whether the coal gun is in a blocking state or not according to the area information of the coal briquette area. 10000 coal gun sample image data sets are manufactured and used for training a network, and experiments prove that the coal briquette region extracted by the method is high in precision and adaptive.
drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the present application, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a flow chart of the detection of the injection state of a coal injection branch pipe at a blast furnace tuyere according to the present invention;
FIG. 2 is a flow chart of an image denoising algorithm;
FIG. 3 is a flowchart of an image binarization process;
FIG. 4 is a schematic view of a normal injection detection effect of a coal injection branch pipe;
FIG. 5 is a diagram illustrating the effect of detecting the blockage of the coal injection branch pipe;
FIG. 6(a) (b) (c) (d) is a schematic diagram of the tuyere image after denoising in the embodiment; (a) normal blowing tuyere images; (b) denoising the normal blowing air inlet image; (c) coal gun blocking tuyere image; (d) denoising an image of a coal gun blocking air port;
FIG. 7(a) (b) (c) (d) is a schematic diagram showing the enhanced front and rear tuyere image contrast in the example; (a) the method comprises the steps of (a) obtaining an image of a normal injection tuyere before image enhancement, (b) obtaining an image of a normal injection tuyere after image enhancement, (c) obtaining an image of a tuyere blocked by a coal gun before image enhancement, (b) obtaining an image of a tuyere blocked by a coal gun after image enhancement;
Fig. 8(a), (b), (c) and (d) are comparison diagrams before and after the tuyere image binarization processing, (a) normal blowing tuyere images before the binarization processing; (b) carrying out binarization treatment on the normal blowing air inlet image; (c) an air port image blocked by a coal gun before binarization processing; (d) binarizing the image of the tuyere blocked by the coal gun; a
FIG. 9(a) (b) is a graph showing the results of fitting the tuyere region;
FIG. 10(a) (b) is a schematic diagram of coal lance region detection;
FIG. 11(a) (b) is a schematic diagram showing the extraction result of coal briquette region; (a) is a normal injection schematic diagram, and (b) is a schematic diagram of coal gun blockage;
FIG. 12 is a schematic view of a coal injection flow rate characteristic curve; (a) is a schematic diagram of normal injection, and (b) is a schematic diagram of blockage of a coal gun.
Fig. 13 is a schematic diagram of a multi-scale FCN model structure.
Detailed Description
in order to make the technical solutions and advantages of the present invention clearer, the following describes the technical solutions in the embodiments of the present invention clearly and completely with reference to the drawings in the embodiments of the present invention:
As shown in fig. 1, the method for detecting the coal injection state of the blast furnace tuyere based on image processing specifically comprises the following steps:
step 1, researching a self-adaptive image denoising algorithm. Aiming at the problem of serious image noise, the noise type of the image is judged by analyzing the frequency spectrum information and the histogram information of the image, the noise type mainly comprises stripe noise and Gaussian noise, and an improved image denoising algorithm combining frequency domain filtering and space domain filtering is provided. On the frequency domain, a self-adaptive Butterworth filter is adopted to remove stripe noise; in the spatial domain, a median filter is adopted to remove Gaussian noise and light spots, and finally, a mathematical morphology operation is adopted to eliminate digital labels in the image.
Step 1-1: and selecting a frequency domain filtering method aiming at the stripe noise in the image. Fourier transformation is carried out on the tuyere image with the size of 240 x 192 to obtain a spectrogram F (u, v), cumulative distribution functions in the horizontal direction and the vertical direction are respectively calculated for the spectrogram according to a formula (1) and a formula (2), the function value of the midpoint of the cumulative distribution function in the vertical direction with stripe noise is obviously larger than the function value without stripe noise, and through setting a threshold value T, if the function value is larger than the T, the image has stripe noise, otherwise, the image does not have stripe noise;
Step 1-2: if the tuyere image has stripe noise, calculating an abscissa u0 of a horizontal cumulative distribution function sub-peak frequency point, constructing a self-adaptive second-order Butterworth filter H (u, v) according to u0, and performing filtering processing by using an equation (6). Finally, the filtered spectrogram is subjected to inverse Fourier transform to obtain a restored image
D(u,v)＝[(u-96-u)+(v-120-96)] (3)
D(u,v)＝[(u-96+u)+(v-120+96)] (4)
Step 1-3: and filtering out light spots and Gaussian noise in the image by adopting median filtering.
step 1-4: and eliminating the digital label area by adopting a combined filtering algorithm of firstly opening operation and then closing operation, wherein the structural elements adopt a 3 multiplied by 3 rectangular convolution kernel.
and 2, researching a self-adaptive image enhancement algorithm. Aiming at the problem that the quality of a shot tuyere image is seriously reduced and the phenomenon of unbalanced and fuzzy gray scale exists in the image due to the fact that the tuyere camera is in a complex severe environment with much dust and high temperature for a long time. In addition, the diffuse reflection phenomenon of the furnace wall and coke particles causes halo phenomena with different degrees at the edge of the image, and an improved image enhancement algorithm combining self-adaptive gray scale stretching and power transformation is provided.
step 2-1: the cumulative histogram distribution of the image is obtained, the cumulative histogram is a continuous accumulation of the number of all pixels before the current pixel in the histogram, reflecting the proportion of pixels less than or equal to the current value, as shown in fig. 1. Selecting the gray values corresponding to the gray value frequency of 0.1 and 0.95 in the cumulative histogram as the window bottom b and the window top c of the gray stretching, and performing the gray stretching on the image as shown in the formula (8).
wherein s is an image before gray stretching, t is an image after stretching, b is a window bottom of gray stretching, and c is a window top.
step 2-2: analyzing the histogram characteristics of the halo phenomenon image, and eliminating halo through power transformation, wherein the power transformation principle is as follows:
s＝c×I (9)
Where γ >1, the image global gray scale value may be reduced. The halo phenomenon is caused by strong light rays, and is expressed by that the gray value is increased and the range is widened in the image, so that the halo phenomenon in the image can be restrained by processing a high gray area of the image through power transformation.
and step 3: and processing the image through a binarization algorithm combining a global threshold and a local threshold to obtain a binarization image.
Step 3-1: firstly, the overall threshold T1 of the image is calculated by adopting the maximum inter-class difference method of the overall threshold
step 3-2: and calculating a local threshold value for each pixel of the image by adopting a Gaussian weighted local threshold value algorithm.
Step 3-3: setting a constant value a, belonging to (0,1), traversing image pixels, and judging a high-gray-level region and a low-gray-level region (namely f (x, y) > (1+ a) xT 1 or f (x, y) < (1-a) xT 1, wherein f (x, y) is a certain pixel point in the image) in the image by using a maximum inter-class difference method of a global threshold method; and judging the residual region by a local threshold value method to obtain a binary image.
and 4, step 4: the air vent area is positioned by adopting least square ellipse fitting, the method can fit the air vent edge deletion phenomenon in the image, and the phenomenon that coal briquettes are adhered to the furnace wall when the coal injection quantity is large can be avoided.
Step 4-1: and adopting canny edge detection to obtain edge information of the tuyere image.
step 4-2: according to the obtained air port edge, the air port area is located through a least square ellipse fitting algorithm, which is specifically as follows:
(1) For non-standard equations of an ellipse, the following form can be written:
x+Axy+By+Cx+Dy+E＝0 (10)
(2) For equation (10), there are 5 unknowns in the equation, a, B, C, D, E, respectively. To solve for 5 unknowns, at least 5 sets of samples are required. And fitting an elliptic non-standard equation to the sampling points by a least square method. The formula is as follows:
min||x+Axy+By+Cx+Dy+E||＝0 (11)
Solving it, where N represents the number of sample points:
(3) Equation (13) is solved for example:
(4) then, placing the items containing A, B, C, D, E on the left and the others on the right, yields:
∑Axy+∑Bxy+∑Cxy+∑Dxy+∑Exy＝∑-xy (15)
written in matrix form as follows:
Similarly, the following matrix equation is obtained and expressed:
the simplified writing is:
Finally, the fitting ellipse parameters are obtained by the formula (18):
and 5: considering that the gray information of a coal gun, a coal briquette and an inner wall area in the tuyere image are close, the traditional threshold segmentation algorithm is difficult to apply. In addition, the blast furnace damping down needs to disassemble and overhaul the coal gun, and the re-installation may cause the position of the coal gun in the images before and after damping down to change. And (3) detecting the coal gun area through an improved multi-scale full convolution neural network, and subtracting the binarized image in the step (3) from the coal gun area detected by the full convolution neural network to obtain a coal briquette area.
further, as shown in fig. 13, the multi-scale full convolution neural network model is improved based on a full convolution neural network (FCN), and the network includes two parts, namely an encoding-decoding full convolution network and an optimization network. Firstly, in an encoding-decoding full convolution network, considering that a lighter network can extract detailed characteristics such as textures, colors, edges and the like, and a deep network can learn abstract characteristics such as high-level semantic information and the like, different layers of networks are subjected to cascading prediction and segmentation to obtain results, so that learning of edge detailed information can be enhanced; secondly, a pyramid pooling layer is added, and the prediction accuracy of the network can be improved by extracting features from different scales; and finally, the prediction result of the coding-decoding full convolution network is used as input by the optimization network, wherein the optimization network consists of six convolution layers, two of the convolution layers are hollow convolution, the edge region of the segmentation result can be optimized, the integral characteristic learning capability is improved, and the accurate segmentation result is obtained.
step 6: and (5) subtracting the binary image from the coal gun detection result in the step (5) to extract a coal briquette area. And calculating the area information of the obtained coal briquette area as a characteristic value, if the characteristic value extracted from the continuous k frames is smaller than a threshold value T, judging that the coal briquette is blocked, otherwise, still keeping the coal gun in a coal injection state.
Example (b):
The blast furnace tuyere raceway is an important reaction area for blast furnace operation, can be called as the heart of the blast furnace, can provide heat and energy required by smelting for the blast furnace by blowing coal powder to the blast furnace tuyere, and plays an important role in the stable and smooth operation of the blast furnace smelting process. In recent years, with the increase in the amount of coal injection into blast furnaces, there has been an increasing demand for safety and stability of coal injection. When the coal injection branch pipe is in a blocked state, discontinuous coal injection can cause local non-uniformity of load, and the service life of the blast furnace and the smelting quality are seriously influenced. If the blockage of the injection branch pipe can be detected in time, the blast furnace operator can take corresponding measures in advance, and the quality of the blast furnace molten iron is effectively improved and stabilized.
Although the tuyere camera has been installed and used in blast furnaces of various large steel mills, a new way is provided for tuyere state detection, but the tuyere image shot by the tuyere camera is blurred due to dusty and high-temperature severe environment, and the problem of abnormal state missing judgment of a manual detection method is serious. We propose a blast furnace tuyere branch pipe blockage detection method based on image processing, and verify the accuracy of the algorithm through images acquired by a 2500 m3 blast furnace in a certain steel mill. In the following, we will explain the implementation steps by combining the specific processes:
step 1: and denoising the acquired tuyere image to obtain a denoised tuyere image, as shown in fig. 6.
step 2: the contrast of each region is improved through a self-adaptive gray stretching algorithm, and the halo phenomenon existing at the edge of the image is removed by adopting power change, so that a good foundation is laid for subsequent processing work, and the result is shown in fig. 7.
And step 3: a binarization processing method combining the difference method between the maximum classes of the global threshold value and the gaussian weighted local threshold value algorithm is adopted to process the areas with high gray scale and low gray scale by adopting the difference method between the maximum classes of the global threshold value, and the residual pixel areas are processed by adopting the gaussian weighted local threshold value algorithm, so that a binarization image is obtained, and the result is shown in fig. 8.
And 4, step 4: consider an approximately elliptical area in the image where the ostia are located. In the partial image, the coal briquette target and the tuyere furnace wall are adhered due to large coal powder injection amount, the edge of the tuyere area is obtained by canny edge detection, and the tuyere area is fitted by a least square method, and the result is shown in fig. 9.
And 5: the coal gun area in the tuyere image is segmented by adopting a trained full convolution neural network, the network structure is shown in fig. 13, and the detection result is shown in fig. 10.
Step 6: and (5) subtracting the binarized tuyere image from the coal gun detection result in the step 5 to extract a coal briquette area, as shown in fig. 11. And taking the extracted area information of the coal briquette area as a characteristic value, judging that the coal briquette is blocked if the characteristic value extracted by continuous k frames is smaller than a threshold value T, otherwise, still keeping the coal gun in a coal injection state.
The detection method is used for detection, in 50 sections of tuyere videos acquired on site, coal guns in 20 sections of videos are blocked, the detection results are all correct, the real-time requirement of an industrial site can be met when GTX-1080GPU processing is used, and the coal injection flow characteristic curve is shown in figure 12.
the above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the scope of the present invention.
Claims (4)
1. A blast furnace tuyere coal injection state detection method based on image processing is characterized by comprising the following steps:
denoising the blast furnace tuyere image by adopting a mode of combining frequency domain filtering and spatial domain filtering to obtain a denoised image;
processing the de-noised image by adopting an improved image enhancement algorithm combining self-adaptive gray stretching and power transformation to obtain a clear image;
Processing the clear image by adopting a binarization algorithm combining a global threshold and a local threshold to obtain a binarization image;
fitting the tuyere region in the binary image by adopting a least square ellipse fitting mode;
inputting the enhanced clear image into an improved multi-scale full-convolution neural network, segmenting a coal gun region in the image, and subtracting the whole connected domain of the coal gun and the coal briquette in the binary image from the coal gun region detected by the full-convolution neural network to obtain a coal briquette region;
and judging whether the area information of the coal briquette area is smaller than a threshold value T or not according to the obtained area information of the coal briquette area, if the continuous k frames are smaller than the threshold value T, judging that the blast furnace branch pipe is in a blocked state, and if not, judging that the blast furnace branch pipe is still in a coal injection state.
2. The method of claim 1, further characterized by: when the blast furnace tuyere image is subjected to denoising processing, firstly analyzing the noise type existing in the blast furnace tuyere image, wherein the noise type comprises horizontal direction stripe noise and Gaussian noise;
Selecting a frequency domain filtering method to process the stripe noise in the image: firstly, Fourier transformation is carried out on a tuyere image to obtain a spectrogram F (u, v), cumulative distribution functions in the horizontal direction and the vertical direction of the spectrogram are respectively calculated for the spectrogram F (u, v), whether a function value at the midpoint of the cumulative distribution function in the vertical direction is greater than a threshold value or not is judged, if so, stripe noise exists in the image, otherwise, the stripe noise does not exist in the image;
if the blast furnace tuyere image has stripe noise, calculating the abscissa u0 of the quadratic peak frequency point of the cumulative distribution function in the horizontal direction, constructing a self-adaptive second-order Butterworth filter H (u, v) according to u0 for filtering, and performing inverse Fourier transform on the filtered spectrogram to obtain a restored image
Filtering out light spots and Gaussian noise in the image by adopting median filtering;
the digital label area in the image is eliminated by using a morphological open-close filter.
3. The method of claim 1, further characterized by: when the de-noised image is subjected to enhancement processing, firstly, the cumulative histogram distribution of the image is obtained, the corresponding gray values when the frequency of the gray values in the cumulative histogram meets the set threshold value are selected as the window bottom b and the window top c of gray stretching, and the image is subjected to gray stretching; analyzing the histogram characteristics of the image with the halo phenomenon, and eliminating the halo part in the image through power transformation.
4. the method of claim 1, further characterized by: the binary image is obtained by adopting the following method:
firstly, classifying pixels in an image into a high gray level region, a middle gray level region and a low gray level region; secondly, dividing the high-gray area and the low-gray area by adopting a global threshold maximum inter-class difference method; and judging the remaining middle gray level area through a Gaussian weighted local threshold algorithm to obtain a final binary image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910834054.0A CN110544261B (en) | 2019-09-04 | 2019-09-04 | Method for detecting coal injection state of blast furnace tuyere based on image processing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910834054.0A CN110544261B (en) | 2019-09-04 | 2019-09-04 | Method for detecting coal injection state of blast furnace tuyere based on image processing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110544261A true CN110544261A (en) | 2019-12-06 |
| CN110544261B CN110544261B (en) | 2023-08-29 |
Family
ID=68711207
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910834054.0A Active CN110544261B (en) | 2019-09-04 | 2019-09-04 | Method for detecting coal injection state of blast furnace tuyere based on image processing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110544261B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111353432A (en) * | 2020-02-28 | 2020-06-30 | 安徽华润金蟾药业股份有限公司 | Rapid honeysuckle medicinal material cleaning method and system based on convolutional neural network |
| CN111776500A (en) * | 2020-07-14 | 2020-10-16 | 长沙有色冶金设计研究院有限公司 | Method for clearing blockage of material bin |
| CN112017157A (en) * | 2020-07-21 | 2020-12-01 | 中国科学院西安光学精密机械研究所 | Method for identifying damage point in optical element laser damage threshold test |
| CN112633049A (en) * | 2020-08-19 | 2021-04-09 | 中冶南方工程技术有限公司 | System for monitoring working state of blast furnace tuyere |
| CN112819802A (en) * | 2021-02-09 | 2021-05-18 | 东北大学 | Method for monitoring and predicting abnormal condition of blast furnace based on deep learning of tuyere information |
| WO2021182220A1 (en) * | 2020-03-11 | 2021-09-16 | Ｊｆｅスチール株式会社 | Blast-furnace furnace condition learning method, furnace condition learning device, abnormality detection method, abnormality detection device, and operation method |
| CN113822158A (en) * | 2021-08-19 | 2021-12-21 | 鹏城实验室 | Blast furnace condition prediction method based on charge level flame image and related equipment |
| CN115049662A (en) * | 2022-08-16 | 2022-09-13 | 山东大拇指喷雾设备有限公司 | Sprayer flow control method based on image processing |
| CN116740384A (en) * | 2023-06-15 | 2023-09-12 | 东莞市商斯迈智能科技有限公司 | Intelligent control method and system of floor washing machine |
| CN118362207A (en) * | 2024-06-19 | 2024-07-19 | 开信(大连)互联网服务有限公司 | Animal body temperature detection system for livestock and poultry |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030136318A1 (en) * | 2002-01-22 | 2003-07-24 | Brown Robert K. | Method and apparatus for clearing a powder accumulation in a powder delivery tube |
| CN1929552A (en) * | 2006-09-30 | 2007-03-14 | 四川长虹电器股份有限公司 | Spatial domain pixel data processing method |
| CN101540039A (en) * | 2008-03-21 | 2009-09-23 | 李金宗 | Method for super resolution of single-frame images |
| CN103578114A (en) * | 2013-11-25 | 2014-02-12 | 中冶南方工程技术有限公司 | Coal injection branch pipe blocking measuring and preventing method based on image processing |
| JP2015052148A (en) * | 2013-09-06 | 2015-03-19 | 新日鐵住金株式会社 | Method for controlling blast furnace by determining operational situation thereof |
| CN108537735A (en) * | 2018-04-16 | 2018-09-14 | 电子科技大学 | A kind of image split-joint method of focal plane terahertz imaging |
| CN108986124A (en) * | 2018-06-20 | 2018-12-11 | 天津大学 | In conjunction with Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method |
| US20190050667A1 (en) * | 2017-03-10 | 2019-02-14 | TuSimple | System and method for occluding contour detection |
| CN109426825A (en) * | 2017-08-31 | 2019-03-05 | 北京图森未来科技有限公司 | A kind of detection method and device of object closed outline |
-
2019
- 2019-09-04 CN CN201910834054.0A patent/CN110544261B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030136318A1 (en) * | 2002-01-22 | 2003-07-24 | Brown Robert K. | Method and apparatus for clearing a powder accumulation in a powder delivery tube |
| CN1929552A (en) * | 2006-09-30 | 2007-03-14 | 四川长虹电器股份有限公司 | Spatial domain pixel data processing method |
| CN101540039A (en) * | 2008-03-21 | 2009-09-23 | 李金宗 | Method for super resolution of single-frame images |
| JP2015052148A (en) * | 2013-09-06 | 2015-03-19 | 新日鐵住金株式会社 | Method for controlling blast furnace by determining operational situation thereof |
| CN103578114A (en) * | 2013-11-25 | 2014-02-12 | 中冶南方工程技术有限公司 | Coal injection branch pipe blocking measuring and preventing method based on image processing |
| US20190050667A1 (en) * | 2017-03-10 | 2019-02-14 | TuSimple | System and method for occluding contour detection |
| CN109426825A (en) * | 2017-08-31 | 2019-03-05 | 北京图森未来科技有限公司 | A kind of detection method and device of object closed outline |
| CN108537735A (en) * | 2018-04-16 | 2018-09-14 | 电子科技大学 | A kind of image split-joint method of focal plane terahertz imaging |
| CN108986124A (en) * | 2018-06-20 | 2018-12-11 | 天津大学 | In conjunction with Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method |
Non-Patent Citations (16)
| Title |
|---|
| 77喵~: "最小二乘法椭圆拟合", pages 1 - 2, Retrieved from the Internet <URL:https://blog.csdn.net/qq_30282427/article/details/82117365> * |
| AI视觉网奇: "利用python opencv实现图像自适应二值化", pages 1 - 2, Retrieved from the Internet <URL:https://blog.csdn.net/jacke121/article/details/75444872> * |
| ZAIWANG GU等: "DeepDisc: Optic Disc Segmentation Based on Atrous Convolution and Spatial Pyramid Pooling", 《COMPAY 2018: COMPUTATIONAL PATHOLOGY AND OPHTHALMIC MEDICAL IMAGE ANALYSIS》, pages 253 - 260 * |
| 吴晨;易本顺;章云港;黄松;冯雨;: "基于改进卷积神经网络的视网膜血管图像分割", 光学学报, no. 11, pages 133 - 139 * |
| 吴维聪: "《计算机图像处理》", 31 March 1989 * |
| 孙海川: "基于全卷积网络的图像语义分割算法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 孙海川: "基于全卷积网络的图像语义分割算法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》, 15 January 2019 (2019-01-15), pages 9 - 11 * |
| 张健: "基于深度学习的图像语义分割方法", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 张健: "基于深度学习的图像语义分割方法", 《中国优秀硕士学位论文全文数据库信息科技辑》, 15 August 2018 (2018-08-15) * |
| 秦亭亭: "基于视频图像的高炉风口喷煤状态检测方法研", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 秦亭亭: "基于视频图像的高炉风口喷煤状态检测方法研", 《中国优秀硕士学位论文全文数据库信息科技辑》, 15 July 2014 (2014-07-15), pages 17 - 24 * |
| 秦亭亭: "高炉风口喷煤流量检测方法研究", 《流量计量》 * |
| 秦亭亭: "高炉风口喷煤流量检测方法研究", 《流量计量》, 26 March 2015 (2015-03-26) * |
| 胡德勇: "《遥感图像处理原理和方法》", 30 November 2014 * |
| 邹园园: "THz图像条纹噪声消除方法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 邹园园: "THz图像条纹噪声消除方法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》, 15 October 2009 (2009-10-15), pages 9 - 17 * |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111353432B (en) * | 2020-02-28 | 2023-08-01 | 安徽华润金蟾药业股份有限公司 | Rapid clean selection method and system for honeysuckle medicinal materials based on convolutional neural network |
| CN111353432A (en) * | 2020-02-28 | 2020-06-30 | 安徽华润金蟾药业股份有限公司 | Rapid honeysuckle medicinal material cleaning method and system based on convolutional neural network |
| WO2021182220A1 (en) * | 2020-03-11 | 2021-09-16 | Ｊｆｅスチール株式会社 | Blast-furnace furnace condition learning method, furnace condition learning device, abnormality detection method, abnormality detection device, and operation method |
| JP2021143362A (en) * | 2020-03-11 | 2021-09-24 | Ｊｆｅスチール株式会社 | Furnace condition learning method for blast furnace, furnace condition learning device, abnormality detection method, abnormality detection device and operation method |
| CN111776500A (en) * | 2020-07-14 | 2020-10-16 | 长沙有色冶金设计研究院有限公司 | Method for clearing blockage of material bin |
| CN112017157A (en) * | 2020-07-21 | 2020-12-01 | 中国科学院西安光学精密机械研究所 | Method for identifying damage point in optical element laser damage threshold test |
| CN112017157B (en) * | 2020-07-21 | 2023-04-11 | 中国科学院西安光学精密机械研究所 | Method for identifying damage point in optical element laser damage threshold test |
| CN112633049A (en) * | 2020-08-19 | 2021-04-09 | 中冶南方工程技术有限公司 | System for monitoring working state of blast furnace tuyere |
| CN112819802A (en) * | 2021-02-09 | 2021-05-18 | 东北大学 | Method for monitoring and predicting abnormal condition of blast furnace based on deep learning of tuyere information |
| CN112819802B (en) * | 2021-02-09 | 2024-03-05 | 东北大学 | Method for supervising and predicting blast furnace condition abnormality based on tuyere information deep learning |
| CN113822158A (en) * | 2021-08-19 | 2021-12-21 | 鹏城实验室 | Blast furnace condition prediction method based on charge level flame image and related equipment |
| CN115049662A (en) * | 2022-08-16 | 2022-09-13 | 山东大拇指喷雾设备有限公司 | Sprayer flow control method based on image processing |
| CN115049662B (en) * | 2022-08-16 | 2022-11-08 | 山东大拇指喷雾设备有限公司 | Sprayer flow control method based on image processing |
| CN116740384A (en) * | 2023-06-15 | 2023-09-12 | 东莞市商斯迈智能科技有限公司 | Intelligent control method and system of floor washing machine |
| CN116740384B (en) * | 2023-06-15 | 2024-03-29 | 东莞市商斯迈智能科技有限公司 | Intelligent control method and system of floor washing machine |
| CN118362207A (en) * | 2024-06-19 | 2024-07-19 | 开信(大连)互联网服务有限公司 | Animal body temperature detection system for livestock and poultry |
| CN118362207B (en) * | 2024-06-19 | 2024-08-30 | 开信(大连)互联网服务有限公司 | Animal body temperature detection system for livestock and poultry |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110544261B (en) | 2023-08-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110544261B (en) | Method for detecting coal injection state of blast furnace tuyere based on image processing | |
| CN110268190B (en) | Underground pipe gallery leakage detection method based on static infrared thermography processing | |
| CN112819802B (en) | Method for supervising and predicting blast furnace condition abnormality based on tuyere information deep learning | |
| CN104616290A (en) | Target detection algorithm in combination of statistical matrix model and adaptive threshold | |
| CN109461141A (en) | A kind of workpiece starved detection method | |
| CN113658131B (en) | Machine vision-based tour ring spinning broken yarn detection method | |
| CN114926407A (en) | Steel surface defect detection system based on deep learning | |
| CN103509901B (en) | Blast furnace raw fuel particle size detection method | |
| CN107944354B (en) | Vehicle detection method based on deep learning | |
| CN113989257A (en) | Electric power comprehensive pipe gallery settlement crack identification method based on artificial intelligence technology | |
| Li et al. | Detection algorithm of defects on polyethylene gas pipe using image recognition | |
| CN112085651B (en) | Automatic shock wave detection and tracking algorithm based on image self-adaptive threshold and feature extraction | |
| CN108389216A (en) | Local auto-adaptive threshold segmentation method towards on-line ferrograph image wear Particles Recognition | |
| CN101996307A (en) | Intelligent video human body identification method | |
| CN116883408B (en) | Integrating instrument shell defect detection method based on artificial intelligence | |
| CN115456956A (en) | Method and device for detecting scratches of liquid crystal display and storage medium | |
| CN113487563B (en) | EL image-based self-adaptive detection method for hidden cracks of photovoltaic module | |
| CN111985314B (en) | Smoke detection method based on ViBe and improved LBP | |
| CN113096103A (en) | Intelligent smoke image sensing method for emptying torch | |
| CN110264434A (en) | A kind of removing rain based on single image method based on low-rank matrix completion | |
| CN207319290U (en) | One kind interference far field beam test system | |
| CN116524397A (en) | Intelligent detection and analysis method for pipeline defects | |
| CN115797327A (en) | Defect detection method and device, terminal device and storage medium | |
| CN114820383A (en) | Pollution grade evaluation method based on insulator visible light image | |
| Niu et al. | Application of CEM algorithm in the field of tunnel crack identification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |