A method and system for detecting water leakage in the secondary cooling chamber of continuous casting of copper rods

Abstract

This invention relates to the field of leakage detection technology, and in particular to a method and system for detecting leakage in the secondary cooling chamber of continuous casting of copper rods. The method includes: acquiring two infrared images of the acquisition area at various center wavelengths and two orthogonal polarization directions at regular intervals, and recording them as polarization images; after each image acquisition, converting all images into a temperature distribution matrix using a polarization conversion model and a temperature conversion model; using multiple previous temperature distribution matrices, establishing a feature model for each pixel in the acquisition area, and then comparing each pixel in the current temperature distribution matrix with its corresponding feature model to determine whether each pixel is a normal point or an abnormal point; calculating the comprehensive abnormality index based on the distribution of abnormal points, the degree of temperature deviation between abnormal points and normal points, and the persistence of abnormal points, and determining whether a leakage problem has occurred. This invention can effectively improve the sensitivity and accuracy of leakage detection.

Description

Technical Field

[0001] This invention relates to the field of leakage detection technology, and in particular to a method and system for detecting leakage in the secondary cooling chamber of continuous casting of copper rods. Background Technology

[0002] In the continuous casting production of copper rods, the uniformity of the secondary cooling process directly affects the uniformity of the solidification structure of the billet, thus impacting the machinability of the copper rod and leading to poor final product quality. Secondary cooling typically uses nozzles to evenly spray cooling water onto the billet. When leaks occur in the secondary cooling chamber, such as in the water pipes or at the nozzle connections, large water droplets can fall onto the copper rod, causing uneven cooling of the billet. Therefore, real-time monitoring of leaks is necessary, and alarms should be issued promptly upon detection. Existing leak detection systems typically detect leaks by measuring the water pressure within the secondary cooling chamber pipes. However, if the leak is small, even though water droplets may have formed on the copper rod, the change in water pressure within the pipes may not be significant. Therefore, existing detection methods have low sensitivity and poor accuracy in detecting leaks. Summary of the Invention

[0003] This invention provides a method and system for detecting water leakage in the secondary cooling chamber of continuous casting of copper rods, which can effectively solve the problems in the background art.

[0004] This invention provides a method for detecting water leakage in the secondary cooling chamber of a copper rod continuous casting plant, comprising the following steps:

[0005] Set the image acquisition area in the secondary cooling chamber and set two different center wavelengths L1 and L2.

[0006] At regular intervals, two infrared images of the acquisition area at each center wavelength and in two orthogonal polarization directions are acquired and recorded as polarization images.

[0007] After each image acquisition, the two polarization images for each center wavelength are converted into a radiation image using a polarization conversion model; then, the two radiation images are converted into a temperature distribution matrix using a temperature conversion model.

[0008] Using multiple previous temperature distribution matrices, a feature model is built for each pixel in the acquisition area. Then, each pixel in the current temperature distribution matrix is ​​compared with its corresponding feature model to determine whether each pixel is a normal point or an abnormal point.

[0009] The comprehensive anomaly index is calculated based on the distribution of anomalies, the degree of temperature deviation between anomalies and normal points, and the duration of anomalies. If the comprehensive anomaly index exceeds the alarm threshold, it is determined that a water leakage problem has occurred, an alarm is triggered, and the leakage area is calculated based on the anomalies.

[0010] Furthermore, the polarization conversion model is as follows:

[0011] Let Ip(L1) and Is(L1) be the two polarization images corresponding to the center wavelength L1;

[0012] Let Ip(L2) and Is(L2) be the two polarization images corresponding to the center wavelength L2;

[0013] The ambient humidity inside the secondary cooling chamber was RH when the image was acquired.

[0014] Calculate the polarization degree of each pixel:

[0015] p(Lq,n)=[ip(Lq,n)-is(Lq,n)] / [ip(Lq,n)+is(Lq,n)];

[0016] Where p(Lq,n) is the polarization degree of the nth pixel;

[0017] ip(Lq,n) is the value of the nth pixel in the polarization image Ip(Lq) corresponding to the center wavelength Lq;

[0018] is(Lq,n) is the value of the nth pixel in the polarization image Is(Lq) corresponding to the center wavelength Lq;

[0019] The radiation pattern at center wavelength Lq is Iaw(Lq) = [ip(Lq,n) + is(Lq,n) - path(p(Lq,n),RH)] / exp[-t(p(Lq,n),RH)];

[0020] Where path(p(Lq,n),RH) and t(p(Lq,n),RH) are the path radiation value and optical thickness corresponding to the nth pixel in the polarization image Is(Lq) corresponding to the center wavelength Lq, respectively, and the values ​​are obtained by looking up a table.

[0021] Furthermore, when acquiring polarization images Ip(L1), Is(L1), Ip(L2), and Is(L2), timestamps are added to each polarization image, and when calculating the radiation image, it is ensured that the timestamps of each polarization image are the same.

[0022] Furthermore, the temperature transformation model is as follows:

[0023] Let the radiation patterns for the two center wavelengths be Iaw(L1) and Iaw(L2);

[0024] Calculate the radiometric ratio of each pixel:

[0025] Rcor(n)=iaw(L1,n) / iaw(L2,n);

[0026] Where iaw(L1,n) is the value of the nth pixel in the radiation image Iaw(L1) corresponding to the center wavelength L1;

[0027] iaw(L1,n) is the value of the nth pixel in the radiation image Iaw(L2) corresponding to the center wavelength L2.

[0028] Calculate the temperature of each pixel:

[0029] Tn=C·(1 / L2-1 / L1) / [ln(Rcor(n))-5ln(L2 / L1)-ln(K)-ln(E(L1,Tn) / E(L2,Tn))];

[0030] Where Tn is the temperature corresponding to the nth pixel;

[0031] C is the radiation constant;

[0032] K is the equipment factor;

[0033] E(L1,Tn) and E(L2,Tn) are the emissivity of the center wavelengths of L1 and L2, respectively, and are both functions related to Tn.

[0034] Furthermore, when calculating the temperature of each pixel, the guessed value of Tn is continuously adjusted until the difference between the left and right sides of the temperature model formula is less than a set threshold, thereby solving for Tn.

[0035] Furthermore, the feature model for each pixel within the acquisition area is established as follows:

[0036] Let this be the Tth image acquisition. Calculate the horizontal temperature change rate of the nth pixel as Grax(n,T); calculate the vertical temperature change rate of the nth pixel as Gray(n,T); and let the temperature corresponding to the nth pixel be Tn.

[0037] Set the sliding window length K;

[0038] Calculate the three dimensions of the feature model corresponding to the nth pixel:

[0039] Historical rate of change μGrax(n) = [Grax(n,T-1) + Grax(n,T-2) + ... + Grax(n,TK)] / K;

[0040] Historical vertical change rate μGray(n) = [Gray(n,T-1) + Gray(n,T-2) + ... + Gray(n,TK)] / K;

[0041] Historical temperature μTn=[T(n-1)+T(n-2)+……+T(nK)] / K.

[0042] Furthermore, the specific process for determining normal or abnormal points is as follows:

[0043] Calculate the overall deviation of the nth pixel D(n,T)=[0.2·(Grax(n,T)-μGrax(n))] 2 +0.2·(Gray(n,T)-μGray(n)) 2 +0.6·(Tn-μTn) 2 ] 1 / 2 ;

[0044] Then, the dynamic threshold Dmax(n,T) is calculated as μD(n,T) + 2.5·σD(n,T);

[0045] Where μD(n,T) is the average value of D(n,T-1), D(n,T-2), ..., D(n,TK);

[0046] σD(n,T) is the standard deviation of D(n,T-1), D(n,T-2), ..., D(n,TK);

[0047] If the overall deviation D(n,T) of the nth pixel is greater than the dynamic threshold Dmax, then the pixel is recorded as an abnormal pixel; otherwise, it is recorded as a normal pixel.

[0048] Furthermore, the comprehensive anomaly index is calculated as follows:

[0049] Divide outliers into multiple groups, with each outlier in each group being adjacent to at least one other outlier; count the number of outlier groups Ngrp; calculate the minimum bounding rectangle of the region formed by each group, and denot the average aspect ratio of all minimum bounding rectangles as Ara; calculate the dispersion Sspa=Ngrp·(1+Ara).

[0050] Calculate the average temperature of all outliers as μTa; calculate the average temperature of all pixels as μTglo; calculate the temperature standard deviation of all pixels as σTglo; calculate the temperature deviation Sthe = |μTa - μTglo| / σTglo;

[0051] Compare the anomalies identified in the current image with those identified in the previous image, and denote the number of overlapping anomalies as Nsig; denote the total number of anomalies identified in the current image as Nglo; calculate the anomaly persistence Stem = Nsig / Nglo;

[0052] The comprehensive anomaly index I = 0.2·Sspa + 0.45·Sthe + 0.35·Stem.

[0053] Furthermore, when grouping abnormal points, the intersection of all the smallest bounding rectangles is calculated, and the area with the largest area is taken as the suspected leakage area. If the comprehensive abnormal indicators exceed the alarm threshold, the suspected leakage area is taken as the leakage area for this time.

[0054] The present invention also provides a leakage detection system for the secondary cooling chamber of continuous casting of copper rods, comprising an infrared camera, a polarizing filter assembly, a storage device, and a processor; the infrared camera is used to acquire images; the polarizing filter assembly is disposed in front of the lens of the infrared camera; the storage device is used to store one or more program instructions; the processor is used to run one or more program instructions to perform the steps of the above-described leakage detection method for the secondary cooling chamber of continuous casting of copper rods.

[0055] The technical solution of this invention can achieve the following technical effects:

[0056] This method indirectly detects leaks by using the surface temperature of a copper rod. It is better suited to the environment of the secondary cooling chamber for leak detection and can quickly detect leaks in their early stages, effectively improving the sensitivity and accuracy of leak detection. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 A flowchart illustrating the method for detecting water leakage in the secondary cooling chamber of continuous casting of copper rods;

[0059] Figure 2 This is a schematic diagram of the structural principle of a leakage detection system for the secondary cooling chamber of a copper rod continuous casting system.

[0060] Reference numerals: 1. Infrared camera; 2. Polarizing mirror assembly. Detailed Implementation

[0061] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0063] This invention relates to a method for detecting water leakage in the secondary cooling chamber of a copper rod continuous casting process, such as... Figure 1 As shown, the main steps include hardware parameter setting, image acquisition, image processing, and leak detection. The specific process of each step is as follows:

[0064] Hardware parameter settings: A dual-band infrared camera is set up in the secondary cooling chamber, with the copper rod located in the middle of the camera's field of view. The camera's field of view is the image acquisition area. Since the camera is fixed, each part in the acquisition area can be mapped to the pixel of the image captured by the camera. These small blocks always have a one-to-one correspondence with the coordinates of the pixel.

[0065] A polarizing mirror assembly is installed in front of a dual-band infrared camera. The assembly contains two polarizing mirrors, and the polarization directions of the two polarizing mirrors are orthogonal.

[0066] The dual-band infrared camera has two different center wavelengths, L1 and L2, which can identify the amount of thermal radiation emitted by the copper rod near these two center wavelengths and form an image. It's important to note that this image is not a traditional picture; typically, the thermal radiation image is a matrix, where the value in the x-th row and y-th column corresponds to the amount of thermal radiation at pixel (x, y). Of course, if needed, the value of the thermal radiation at pixel (x, y) can be converted into a corresponding color to make it a visible image, but in this scheme, all subsequent calculations are performed in matrix form.

[0067] Image acquisition: Images are acquired periodically using a dual-band infrared camera. Each acquisition session captures four images, which include:

[0068] Two infrared images of the acquisition area at the center wavelength L1 in two orthogonal polarization directions are denoted as polarization images Ip(L1) and Is(L1), respectively.

[0069] Two infrared images of the acquisition area are obtained in two orthogonal polarization directions at the center wavelength L2. These two infrared images are denoted as polarization images Ip(L2) and Is(L2), respectively.

[0070] Image Processing: After acquiring the four images mentioned above, they need to be immediately converted into a final temperature distribution matrix. This matrix represents the actual temperature at each pixel in the infrared camera image, corresponding to its location within the cooling chamber. For example, pixel (1,2) corresponds to the first row and second column of the temperature distribution matrix, and the data recorded at this location is the temperature value at pixel (1,2) in the infrared camera image. The specific conversion process is as follows:

[0071] The two polarization images Ip(L1) and Is(L1) of the center wavelength L1 are transformed into a single radiation image Iaw(L1) using a polarization transformation model.

[0072] The two polarization images Ip(L2) and Is(L2) of the center wavelength L2 are transformed into a single radiation image Iaw(L2) using a polarization transformation model.

[0073] The radiation images Iaw(L1) and Iaw(L2) show the actual thermal radiation distribution of the copper rod after polarization correction.

[0074] The two radiation images Iaw(L1) and Iaw(L2) are then transformed into a temperature distribution matrix Tn using a temperature transformation model.

[0075] Leakage detection: Using multiple previous temperature distribution matrices, a feature model is built for each pixel in the acquisition area. The feature model is mainly used to quantify the temperature difference between each pixel and its surrounding pixels in the previous temperature distribution matrix. Then, each pixel in the current temperature distribution matrix is ​​compared with its corresponding feature model. Pixels without problems are recorded as normal points. If a pixel in the current temperature distribution matrix shows a significant difference in temperature difference with its surrounding pixels, then the location of this pixel is likely to be problematic. For example, a water leakage problem may cause the local temperature of the copper rod to be too low. Of course, it may also be due to interference from detection noise. Therefore, the pixel with the problem is first recorded as an abnormal point.

[0076] After identifying all the anomalies, a comprehensive judgment is made based on the distribution of the anomalies, the degree of temperature deviation between the anomalies and the normal points, and the persistence of the anomalies, to determine whether these anomalies are leak points.

[0077] To enable the equipment to make automatic judgments, this method quantifies abnormal situations by calculating a comprehensive abnormality index based on the distribution of abnormal points and the degree of temperature deviation between abnormal and normal points. If the comprehensive abnormality index exceeds the alarm threshold, it is determined that a water leakage problem has occurred, an alarm is triggered, and the leakage area is calculated based on the abnormal points.

[0078] Under normal circumstances, radiation from an object's surface is unpolarized or weakly polarized, allowing for direct measurement. However, in the environment of a secondary cooling chamber, water mist is sprayed to cool the copper rod. The tiny spherical water droplets cause scattering of the copper rod's radiation, resulting in polarized thermal radiation from its surface. Therefore, this method uses a polarizing mirror to first acquire radiation in two orthogonal polarization directions. By comparing the two polarized radiation quantities, the interference intensity of the current environment on the radiation at each pixel location can be determined, and the radiation quantity can be reconstructed to obtain a highly accurate radiation image. When calculating the temperature at each pixel, the radiation images of the two center wavelength bands are converted into a temperature distribution matrix using colorimetric thermometry. The temperature distribution within this matrix can then be used to determine whether leakage has occurred.

[0079] Preferably, the polarization conversion model is as follows:

[0080] Let Ip(L1) and Is(L1) be the two polarization images corresponding to the center wavelength L1;

[0081] Let Ip(L2) and Is(L2) be the two polarization images corresponding to the center wavelength L2;

[0082] The ambient humidity inside the secondary cooling chamber was RH when the image was acquired.

[0083] Calculate the polarization degree of each pixel:

[0084] p(Lq,n)=[ip(Lq,n)-is(Lq,n)] / [ip(Lq,n)+is(Lq,n)];

[0085] Where p(Lq,n) is the polarization degree of the nth pixel, which directly reflects the intensity of the water mist scattered light at that pixel. The higher the value of p(Lq,n), the more severe the scattering interference that the pixel is subjected to.

[0086] ip(Lq,n) is the value of the nth pixel in the polarization image Ip(Lq) corresponding to the center wavelength Lq;

[0087] is(Lq,n) is the value of the nth pixel in the polarization image Is(Lq) corresponding to the center wavelength Lq.

[0088] After knowing the scattering interference intensity of each pixel, the radiation image of each center wavelength can be corrected and then calculated, resulting in the radiation image of center wavelength Lq: Iaw(Lq)=[ip(Lq,n)+is(Lq,n)-path(p(Lq,n),RH)] / exp[-t(p(Lq,n),RH)];

[0089] Wherein, path means the intensity of the additional radiation generated by the water mist itself due to thermal radiation and scattering; path(p(Lq,n),RH) is the path radiation value corresponding to the nth pixel in the polarization image Is(Lq) corresponding to the center wavelength Lq.

[0090] The meaning of t is the degree to which radiation is attenuated when passing through water mist. The larger the value of t(p(Lq,n),RH), the more severe the signal attenuation. t(p(Lq,n),RH) are the optical thicknesses corresponding to the nth pixel in the polarization image Is(Lq) corresponding to the center wavelength Lq.

[0091] exp[] represents a function with base e and powers within the brackets, for example, exp[2] is e. 2 .

[0092] The values ​​of path and t cannot be directly measured or calculated. A lookup table needs to be pre-established using scattering theory simulation. The inputs to the lookup table are the polarization degree p and the ambient humidity RH, and the outputs are path and t. When calculating the value of the nth pixel at the center wavelength Lq, the polarization degree p(Lq,n) and ambient humidity RH corresponding to the nth pixel are input into the lookup table to obtain path(p(Lq,n),RH) and t(p(Lq,n),RH) for the nth pixel.

[0093] Preferably, when acquiring polarization images Ip(L1), Is(L1), Ip(L2), and Is(L2), a timestamp is added to each polarization image. When calculating the radiation image of each center wavelength, it is ensured that the timestamps of each polarization image are the same to prevent large deviations in calculation caused by image mixing.

[0094] The preferred temperature conversion model is as follows:

[0095] The radiation patterns for the two center wavelengths have been calculated as Iaw(L1) and Iaw(L2), respectively.

[0096] Calculate the radiometric ratio of each pixel:

[0097] Rcor(n)=iaw(L1,n) / iaw(L2,n);

[0098] Where iaw(L1,n) is the value of the nth pixel in the radiation image Iaw(L1) corresponding to the center wavelength L1;

[0099] iaw(L1,n) is the value of the nth pixel in the radiation image Iaw(L2) corresponding to the center wavelength L2.

[0100] Calculate the temperature of each pixel:

[0101] Tn=C·(1 / L2-1 / L1) / [ln(Rcor(n))-5ln(L2 / L1)-ln(K)-ln(E(L1,Tn) / E(L2,Tn))];

[0102] Where Tn is the temperature corresponding to the nth pixel;

[0103] C is the radiation constant, which is a fixed value;

[0104] K is the device factor, which is related to the sensitivity of the detection device;

[0105] E(L1,Tn) and E(L2,Tn) are the emissivity of the center wavelengths of L1 and L2, respectively, and are both functions related to Tn. These functions are formed by fitting laboratory test data.

[0106] To facilitate the calculation of Tn, especially since E(L1,Tn) and E(L2,Tn) are functions related to Tn and their expressions may differ under different detection environments, making it impossible to simplify the above equation, this method continuously adjusts the guessed value of Tn when calculating the temperature of each pixel until the difference between the left and right sides of the temperature model formula is less than a set threshold, thereby solving for Tn.

[0107] Preferably, the feature model for each pixel within the acquisition area is established as follows:

[0108] Let this be the Tth image acquisition. Calculate the horizontal temperature change rate of the nth pixel as Grax(n,T); calculate the vertical temperature change rate of the nth pixel as Gray(n,T); and let the temperature corresponding to the nth pixel be Tn.

[0109] Grax(n,T) is the sum of the horizontal convolutions of the 3×3 neighborhood centered on the nth pixel in the temperature distribution matrix. Assuming the coordinates of the nth pixel are (x,y) and the corresponding temperature is represented as T(x,y), Grax(n,T) = [T(x+1,y-1) + 2·T(x+1,y) + T(x+1,y+1)] - [T(x-1,y-1) + 2·T(x-1,y) + T(x-1,y+1)];

[0110] Gray(n,T) is the sum of the vertical convolutions of the 3×3 neighborhood centered on the nth pixel in the temperature distribution matrix. Assuming the coordinates of the nth pixel are (x,y) and the corresponding temperature is represented as T(x,y), Gray(n,T)=[T(x-1,y+1)+2·T(x,y+1)+T(x+1,y+1)]-[T(x+1,y-1)+2·T(x,y-1)+T(x-1,y-1)].

[0111] Set the sliding window length K. The length of the sliding window is the number of data points corresponding to the images acquired before the current image acquisition. This sliding window sampling mode can ensure that the data in the feature model is always the latest generated data, thereby ensuring the accuracy of the feature model.

[0112] Now, based on historical data, we calculate the three dimensions of the feature model corresponding to the nth pixel:

[0113] The historical horizontal change rate μGrax(n) = [Grax(n,T-1) + Grax(n,T-2) + ... + Grax(n,TK)] / K reflects the average horizontal change gradient quantization parameter of the nth pixel within the sampling window. Its physical meaning is the degree of numerical change between this pixel and the pixels on its two sides.

[0114] The historical vertical change rate μGray(n) = [Gray(n,T-1) + Gray(n,T-2) + ... + Gray(n,TK)] / K reflects the average vertical change gradient quantization parameter of the nth pixel within the sampling window. Its physical meaning is the degree of numerical change between this pixel and the pixels on its vertical sides.

[0115] Historical temperature μTn=[T(n-1)+T(n-2)+……+T(nK)] / K reflects the average temperature of the nth pixel within the sampling window.

[0116] After obtaining the above feature model, the specific determination of normal or abnormal points can be carried out. The specific process is as follows:

[0117] Calculate the overall deviation of the nth pixel D(n,T)=[0.2·(Grax(n,T)-μGrax(n))] 2 +0.2·(Gray(n,T)-μGray(n)) 2 +0.6·(Tn-μTn) 2 ] 1 / 2 ;

[0118] Grax(n,T), Gray(n,T) and Tn are all values ​​calculated after the image is acquired this time, which are used to reflect the characteristics of the pixels when the image is acquired this time, and then compared with the corresponding historical average values.

[0119] The two values ​​of 0.2 and 0.6 are weighting coefficients, and their values ​​can be changed according to specific requirements.

[0120] Then, the dynamic threshold Dmax(n,T) is calculated as μD(n,T) + 2.5·σD(n,T);

[0121] Where μD(n,T) is the average value of D(n,T-1), D(n,T-2), ..., D(n,TK);

[0122] σD(n,T) is the standard deviation of D(n,T-1), D(n,T-2), ..., D(n,TK);

[0123] If the overall deviation D(n,T) of the nth pixel is greater than the dynamic threshold Dmax, then the pixel is recorded as an outlier; otherwise, it is recorded as a normal pixel. The dynamic threshold Dmax allows the evaluation criteria for outliers to be updated according to the changes in the image, thereby enabling more accurate identification of outliers.

[0124] Preferably, the comprehensive anomaly index is obtained by combining the degree of dispersion, the degree of temperature deviation, and the degree of anomaly persistence, and is specifically calculated as follows:

[0125] Dispersion degree: The outliers are divided into multiple groups, and each outlier in each group is adjacent to at least one other outlier; the number of outlier groups Ngrp is counted to reflect the dispersion degree of the outliers.

[0126] Calculate the minimum bounding rectangle of the area formed by each group. When water leaks, it will form a linear or strip-shaped water flow on the copper rod (i.e., the aspect ratio is relatively high). The average aspect ratio of all minimum bounding rectangles is denoted as Ara, which is used to reflect the situation that the anomaly point may be a water flow area.

[0127] The degree of dispersion is calculated as Sspa = Ngrp·(1+Ara), which combines the degree of dispersion and the degree of shape to reflect the degree of dispersion.

[0128] Temperature deviation: Calculate the average temperature of all outliers as μTa;

[0129] Calculate the average temperature μTglo of all pixels;

[0130] Calculate the temperature standard deviation σTglo for all pixels;

[0131] When water drips onto the copper rod, it causes the temperature of a local area of ​​the copper rod to be significantly lower than that of the surrounding area. The temperature deviation Sthe=|μTa-μTglo| / σTglo is calculated to reflect the temperature difference between the abnormal point and the normal point.

[0132] Persistence of anomalies: The number of anomalies identified in the current image acquisition is compared with the number of anomalies identified in the previous image acquisition, and the number of overlapping anomalies is recorded as Nsig.

[0133] Let Nglo be the total number of outliers identified in this image acquisition.

[0134] The anomaly persistence level Stem = Nsig / Nglo is calculated to reflect the number of persistent anomalies. Since water leakage is a continuous phenomenon, if an anomaly is persistent, then it is very likely that a water leakage has occurred. However, if an anomaly appears suddenly and does not overlap in the image before and after the anomaly, then this anomaly is very likely noise or other interference.

[0135] After calculating the indicators of the three dimensions, the comprehensive anomaly index I can be calculated as I = 0.2·Sspa + 0.45·Sthe + 0.35·Stem;

[0136] Among them, 0.2, 0.45 and 0.35 are all weighting coefficients, and the specific values ​​can be adjusted according to actual needs.

[0137] If a leak occurs, the water will be concentrated in a relatively concentrated area on the copper rod, reflected in the minimum bounding rectangle of the anomaly point group. These minimum bounding rectangles of the leak points will intersect each other. When grouping anomalies, the intersection of all minimum bounding rectangles can be calculated, and this intersection can be merged into a large area. However, some noise points will be far away from the leak points, and the minimum bounding rectangles formed cannot be merged into this large area. The area with the largest area obtained in this way is the suspected leak area. If the comprehensive abnormal indicators exceed the alarm threshold (that is, it is determined that a leak has occurred), the suspected leak area is taken as the current leak area.

[0138] The present invention also relates to a leakage detection system for the secondary cooling chamber of continuous casting of copper rods, comprising an infrared camera 1, a polarizing filter assembly 2, a storage device, and a processor; the infrared camera 1 is used to acquire images; the polarizing filter assembly 2 is disposed in front of the lens of the infrared camera 1; the storage device is used to store one or more program instructions; the processor is used to run one or more program instructions to perform the various steps of the above-described leakage detection method for the secondary cooling chamber of continuous casting of copper rods.

[0139] Similarly, the above-mentioned optimization schemes for the system can also achieve the optimization effects corresponding to the methods in Embodiment 1, which will not be repeated here.

[0140] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for detecting water leakage in the secondary cooling chamber of a copper rod continuous casting machine, characterized in that the steps include... include: Set the image acquisition area in the secondary cooling chamber and set two different center wavelengths L1 and L2. At regular intervals, two infrared images of the acquisition area at each center wavelength and in two orthogonal polarization directions are acquired and recorded as polarization images. After each image acquisition, the two polarization images for each center wavelength are converted into a radiation image using a polarization conversion model; then, the two radiation images are converted into a temperature distribution matrix using a temperature conversion model. Using multiple previous temperature distribution matrices, a feature model is built for each pixel in the acquisition area. Then, each pixel in the current temperature distribution matrix is ​​compared with its corresponding feature model to determine whether each pixel is a normal point or an abnormal point. The comprehensive anomaly index is calculated based on the distribution of anomalies, the degree of temperature deviation between anomalies and normal points, and the duration of anomalies. If the comprehensive anomaly index exceeds the alarm threshold, it is determined that a water leakage problem has occurred, an alarm is triggered, and the leakage area is calculated based on the anomalies. The polarization conversion model is as follows: Let Ip(L1) and Is(L1) be the two polarization images corresponding to the center wavelength L1; Let Ip(L2) and Is(L2) be the two polarization images corresponding to the center wavelength L2; The ambient humidity inside the secondary cooling chamber was RH when the image was acquired. Calculate the polarization degree of each pixel: p(Lq,n)=[ip(Lq,n)-is(Lq,n)] / [ip(Lq,n)+is(Lq,n)]; Where p(Lq,n) is the polarization degree of the nth pixel; ip(Lq,n) is the value of the nth pixel in the polarization image Ip(Lq) corresponding to the center wavelength Lq; is(Lq,n) is the value of the nth pixel in the polarization image Is(Lq) corresponding to the center wavelength Lq; The radiation pattern at center wavelength Lq is Iaw(Lq) = [ip(Lq,n) + is(Lq,n) - path(p(Lq,n),RH)] / exp[-t(p(Lq,n),RH)]; Where path(p(Lq,n),RH) and t(p(Lq,n),RH) are respectively the path radiation value and optical thickness of the nth pixel in the polarization image Is(Lq) corresponding to the center wavelength Lq, and the values ​​are obtained by looking up a table; The temperature conversion model is as follows: Let the radiation patterns for the two center wavelengths be Iaw(L1) and Iaw(L2); Calculate the radiometric ratio of each pixel: Rcor(n)=iaw(L1,n) / iaw(L2,n); Where iaw(L1,n) is the value of the nth pixel in the radiation image Iaw(L1) corresponding to the center wavelength L1; iaw(L1,n) is the value of the nth pixel in the radiation image Iaw(L2) corresponding to the center wavelength L2. Calculate the temperature of each pixel: Tn=C·(1 / L2-1 / L1) / [ln(Rcor(n))-5ln(L2 / L1)-ln(K)-ln(E(L1,Tn) / E(L2,Tn))]; Where Tn is the temperature corresponding to the nth pixel; C is the radiation constant; K is the equipment factor; E(L1,Tn) and E(L2,Tn) are the emissivity of the center wavelengths of L1 and L2, respectively, and are both functions related to Tn.

2. The method for detecting water leakage in the secondary cooling chamber of continuous casting of copper rods according to claim 1, characterized in that, When acquiring polarization images Ip(L1), Is(L1), Ip(L2), and Is(L2), timestamps are added to each polarization image. When calculating the radiation image, it is ensured that the timestamps of each polarization image are the same.

3. The method for detecting water leakage in the secondary cooling chamber of continuous casting of copper rods according to claim 1, characterized in that, When calculating the temperature of each pixel, the guessed value of Tn is continuously adjusted until the difference between the left and right sides of the temperature model formula is less than a set threshold, thereby solving for Tn.

4. The method for detecting water leakage in the secondary cooling chamber of continuous casting of copper rods according to claim 1, characterized in that, The specific steps for establishing a feature model for each pixel within the acquisition area are as follows: Let this be the Tth image acquisition. Calculate the horizontal temperature change rate of the nth pixel as Grax(n,T); calculate the vertical temperature change rate of the nth pixel as Gray(n,T); and let the temperature corresponding to the nth pixel be Tn. Set the sliding window length K; Calculate the three dimensions of the feature model corresponding to the nth pixel: Historical rate of change μGrax(n) = [Grax(n,T-1) + Grax(n,T-2) + ... + Grax(n,TK)] / K; Historical vertical change rate μGray(n) = [Gray(n,T-1) + Gray(n,T-2) + ... + Gray(n,TK)] / K; Historical temperature μTn=[T(n-1)+T(n-2)+……+T(nK)] / K.

5. The method for detecting leakage in the secondary cooling chamber of continuous casting copper rods according to claim 4, characterized in that, The specific process for determining normal or abnormal points is as follows: Calculate the overall deviation of the nth pixel D(n,T)=[0.2·(Grax(n,T)-μGrax(n))] 2 +0.2·(Gray(n,T)-μGray(n)) 2 +0.6·(Tn-μTn) 2 ] 1 / 2 ; Then, the dynamic threshold Dmax(n,T) is calculated as μD(n,T) + 2.5·σD(n,T); Where μD(n,T) is the average value of D(n,T-1), D(n,T-2), ..., D(n,TK); σD(n,T) is the standard deviation of D(n,T-1), D(n,T-2), ..., D(n,TK); If the overall deviation D(n,T) of the nth pixel is greater than the dynamic threshold Dmax, then the pixel is recorded as an abnormal pixel; otherwise, it is recorded as a normal pixel.

6. The method for detecting water leakage in the secondary cooling chamber of continuous casting of copper rods according to claim 1, characterized in that, The calculation of the comprehensive anomaly index is as follows: The outliers are divided into multiple groups, and each outlier in each group is adjacent to at least one other outlier; the number of outlier groups Ngrp is counted; the minimum bounding rectangle of the region formed by each group is calculated, and the average aspect ratio of all minimum bounding rectangles is denoted as Ara. Calculate the degree of dispersion: Sspa = Ngrp·(1+Ara); Calculate the average temperature of all outliers as μTa; calculate the average temperature of all pixels as μTglo; calculate the temperature standard deviation of all pixels as σTglo; Calculate the degree of temperature deviation Sthe = |μTa - μTglo| / σTglo; Compare the anomalies identified in the current image with those identified in the previous image, and denote the number of overlapping anomalies as Nsig; denote the total number of anomalies identified in the current image as Nglo; calculate the anomaly persistence Stem = Nsig / Nglo; The comprehensive anomaly index I = 0.2·Sspa + 0.45·Sthe + 0.35·Stem.

7. The method for detecting leakage in the secondary cooling chamber of continuous casting of copper rods according to claim 6, characterized in that, When grouping abnormal points, the intersection of all the smallest bounding rectangles is calculated, and the area with the largest area is taken as the suspected leakage area. If the comprehensive abnormal indicators exceed the alarm threshold, the suspected leakage area is taken as the leakage area in this case.

8. A leakage detection system for the secondary cooling chamber of a copper rod continuous casting machine, characterized in that, The method includes an infrared camera, a polarizing filter assembly, a storage device, and a processor; the infrared camera is used to acquire images; the polarizing filter assembly is disposed in front of the lens of the infrared camera; the storage device is used to store one or more program instructions; the processor is used to run one or more program instructions to perform the steps of the method for detecting leakage in the secondary cooling chamber of continuous casting of copper rods as described in any one of claims 1 to 7.

Images