A high-noise-resistance wave damper thermal fault positioning method based on thermal region segmentation

Abstract

The application discloses a high-noise-resistance wave choke thermal fault positioning method based on thermal region segmentation, which comprises the following steps: 1, collecting an infrared image; 2, removing the noise of the infrared image by using column weighting and layered filtering algorithm and SVD decomposition and information entropy linear fitting algorithm in sequence; 3, segmenting the infrared image by using a gradient weight region growing method; 4, obtaining an effective extreme point set according to the segmented infrared image; 5, performing clustering processing according to the effective extreme point set, then finding out a thermal abnormal region of the wave choke according to a threshold condition, and drawing a mask at a corresponding position in the infrared image, so as to obtain a heat abnormality positioning image. The application can realize high-quality noise removal and accurate segmentation in the image processing process, thereby improving the accuracy and speed of wave choke abnormal heat identification.

Description

Technical Field

[0001] The present invention belongs to the technical field of power equipment fault detection, and in particular is a method for locating thermal faults of high-noise-resistant wave traps based on thermal region segmentation. Background Art

[0002] The proper operation of power equipment is crucial to industrial production and public life. Compared to defects or malfunctions in instrument design, abnormal heating in power equipment is not only difficult to detect directly but can also impact power supply in a short period of time. Therefore, real-time monitoring of heating failures in power equipment is of practical significance. Transformers, circuit breakers, power capacitors, and surge arresters are common power equipment in substations that are prone to abnormal heating. Compared to ground-based equipment such as transformers and circuit breakers, surge arresters placed in the air are difficult to detect through direct contact and manual inspection. Therefore, designing a non-contact, automatic method for determining heating conditions can meet the needs of real-time monitoring of power equipment.

[0003] Many scholars have conducted research on abnormal heating detection of power equipment:

[0004] In his paper "Research on Denoising of Insulator Infrared Images Based on Wavelet Domain Hidden Markov Tree Model and Importance Correction", Ge Xinyuan constructed a hidden Markov tree model for wavelet coefficients, obtained the noise variance through the Donoho noise estimation method, and used the empirical Bayes formula to remove noise; Quan Pan proposed estimating the noise variance using the spatial domain correlation estimation method in his paper "Two Denoising Methods by Wavelet Transform".

[0005] In their paper “Edge Detection of Infrared Images of High-Voltage Transmission Lines,” Xiao Xiaohui et al. used the Roberts operator to detect edge contours in infrared images of power equipment and achieve equipment pattern segmentation. Patent CN112085037A combines the particle swarm algorithm (PSO) and the Otsu global threshold algorithm to separate the background and power equipment areas, thereby analyzing and processing the equipment area patterns.

[0006] In their paper "Thermal Infrared Image Recognition Method for Substation High-Voltage Equipment Faults Based on Temperature Probability Density", Xiao Yi et al. performed probability density function fitting on the temperature histogram obtained from the infrared image, used the extreme points of the probability density function that met the screening criteria as cluster temperature points, and used the clustering algorithm to segment the thermal area. However, whether the area is abnormally hot still requires manual judgment; in their paper "Research on Substation Equipment Identification and Thermal State Monitoring System Based on Infrared Images", Yin Yang used the SURF algorithm to match feature points to locate thermal anomaly areas.

[0007] Surface temperature, similar comparison, and relative temperature difference are commonly used methods for determining abnormal heating in power equipment. In their paper, "Intelligent Diagnosis of Thermal Defects in Transmission Lines Based on UAV Infrared Video," Zhang Wenfeng et al. proposed a method combining surface temperature and relative temperature difference for detecting heating faults in power equipment.

[0008] Although the typical power equipment heating fault diagnosis methods based on machine vision have been well studied in recent years in terms of equipment identification, infrared image processing, and thermal fault judgment methods, the following deficiencies still exist:

[0009] 1. Most methods fail to account for stripe noise caused by camera hardware. Common wave traps have stripe or grid-shaped housings. Stripe noise can obscure the contours of the wave trap image, severely interfering with wave trap region segmentation and the localization of abnormal hotspots. Furthermore, current general-purpose filtering operators can lose image detail when removing image noise, resulting in blurred or even distorted images.

[0010] 2. In wavelet-domain infrared image noise energy estimation, the mainstream Donoho estimation method assumes that the diagonal wavelet coefficients in the final layer of wavelet decomposition contain very little original image information. This assumes that the original image contains high noise. However, the noise estimate obtained by the Donoho method can be significantly biased in low-noise images. Under current operating conditions, infrared noise caused by optical factors is already weak, while images of wave traps with striped shells contain rich detail. Under low-noise conditions, the Donoho method tends to misinterpret edge details as noise components, resulting in image distortion.

[0011] 3. In the segmentation of equipment image regions, compared to other large-scale power equipment stored independently, there are transmission lines, insulators, hollow brackets and other equipment around the wave trap placed in the air. The threshold segmentation method in the existing scheme is easily affected by the interference of environmental objects, thus incorrectly selecting connected domains. At the same time, the accuracy of the region growing algorithm relies on the selection of seed points, but the current improved region growing algorithm still uses color similarity as the criterion and lacks the geometric configuration information of the target object as a further screening criterion.

[0012] 4. When locating thermal anomalies, the small proportion of the abnormally heated area in the infrared image makes it difficult for a small number of pixels to have sufficient feature points, limiting the accuracy of feature point matching algorithms such as SURF. Furthermore, the neighborhood of the abnormally heated area can also experience temperature increases, blurring the boundaries of the region, leading to significant deviations in similarity comparisons using regional template matching methods. Furthermore, while clustering methods based on feature region matching and temperature probability statistics offer simple steps, they often struggle to distinguish between background points and device area points in the infrared image, leading to a large number of invalid samples and a high number of iterations, resulting in low efficiency.

[0013] 5. When diagnosing thermal anomalies, a common approach for current-induced heating is to combine surface temperature with relative temperature difference to determine the abnormal heating threshold. However, in reality, both current-induced and voltage-induced heating often occur simultaneously. Since voltage-induced heating devices generate less heat and internal defects are often sealed, surface temperature cannot reflect the actual temperature at the defect site. Therefore, other criteria are needed to determine the abnormal heating threshold. Summary of the Invention

[0014] In order to address the shortcomings of the above-mentioned existing technologies, the present invention proposes a high-noise-resistant wave trap thermal fault location method based on thermal area segmentation, so as to achieve high-quality noise removal and accurate segmentation in the image processing process, thereby improving the accuracy and speed of abnormal heating identification of the wave trap.

[0015] In order to achieve the above-mentioned object, the present invention adopts the following technical solutions:

[0016] The method for locating thermal faults of a high-noise-resistant wave trap based on thermal region segmentation of the present invention is characterized by following the steps below:

[0017] Step 1: Acquire an infrared image A1 of the wave trap with a dimension of W×H, where W and H represent the width and height of the infrared image A1 respectively;

[0018] Step 2: De-noise A1 to obtain the de-noised infrared image A4.

[0019] Step 2.1: After converting A1 into a grayscale image A2 of the same scale, A2 is processed using a column weighting and layered filtering algorithm to obtain an infrared image A3 without streak noise.

[0020] Step 2.2: After logarithmic transformation of A3, perform N-layer wavelet decomposition of A3 using Coiflet wavelet basis to obtain the wavelet image set {G i |1≤i≤N}, where G i represents the wavelet image obtained by decomposition of the i-th layer;

[0021] Step 2.3, use the linear fitting algorithm based on SVD decomposition and information entropy to fit {G i |1≤i≤N} is processed to obtain an infrared image A4 without stripe noise and speckle noise;

[0022] Step 3: Use the gradient weighted region growing method to segment the infrared image A4 to obtain a segmented infrared image A5;

[0023] Step 4: According to the grayscale histogram of the segmented infrared image A5, the grayscale probability density function is obtained, and the average grayscale value T of the wave trap area in A1 is obtained by using the grayscale probability density function. device, and then obtain the effective extreme point set of the wave trap area in the three-dimensional grayscale distribution map corresponding to A5;

[0024] Step 5: Take each coordinate in the set of valid extreme points as the cluster center, and use the KNN algorithm to cluster the pixels in A5 to obtain the cluster information matrix. Then, according to the threshold condition, find the thermal anomaly area of ​​the wave trap from the cluster information matrix, and draw a mask at the corresponding position in A1 to obtain the thermal anomaly positioning image A6 of the wave trap.

[0025] The method for locating thermal faults of a high-noise-resistant wave trap based on thermal region segmentation according to the present invention is also characterized in that the column weighting and layered filtering algorithm in step 2.1 includes the following steps:

[0026] Step 2.1.1, use the gradient operator to process the grayscale image A2 to obtain the horizontal gradient map Grad;

[0027] Step 2.1.2, define the column weighted filter window:

[0028] Define a filter window of dimension (2k+1)×(2k+1), where 2k+1 is the dimension of the filter window and k is a positive integer;

[0029] The pixel value S(i, j) at the center position (i, j) of the filter window in A2 is obtained using formula (1), and the boundary constraint condition of the filter window is set to: , , thereby sliding from left to right and from top to bottom on the grayscale image A2 according to the boundary constraint conditions to obtain a sliding window image B1 containing low-frequency information;

[0030] (1)

[0031] In formula (1), represents the mean pixel value in the xth column of A2, and , represents the pixel value at coordinate (x, y) in A2; Indicates about A x The correction function of , represents an artificially determined regulatory factor, represents the mean pixel value of the x+1th column in A2, sgn(·) represents the sign function, represents the mean pixel value of the xth column in Grad, and , Represents the pixel value at coordinate (x, y) in Grad; represents the column weight of the x-th column in A2, and , σ is the variance of the set column weights;

[0032] Step 2.1.3: Subtract B1 from A2 to obtain the difference image B2. High-pass filter B2 is then processed in the column direction to obtain the high-frequency information image B3.

[0033] Step 2.1.4: Use equation (2) to process B3 and obtain the corrected image B4:

[0034] (2)

[0035] In formula (2), B4(x,y) and B3(x,y) represent the pixel values ​​at coordinates (x,y) in B4 and B3 respectively; represents the average value of pixels in B3; λ is an artificially determined noise reduction factor;

[0036] Step 2.1.5: Add B1 and B4 to obtain the true grayscale image B5, and convert B5 to the same color space as A1 to obtain the infrared image A3 without stripe noise.

[0037] Furthermore, the linear fitting algorithm based on SVD decomposition and information entropy in step 2.3 includes the following steps:

[0038] Step 2.3.1, respectively, for the wavelet image set {G i Each wavelet image in | 1 ≤ i ≤ N} is decomposed by SVD in the horizontal sub-band and vertical sub-band directions to obtain a set of non-zero singular value sequences {(Q i , P i )| 1 ≤ i ≤ N}, where Q i is the i-th wavelet image G i The non-zero singular value sequence of the sub-band in the horizontal direction, and Q i ={Q i,s | 1 ≤ s ≤S i}, Q i,s Indicates Q i The s-th horizontal non-zero singular value in S i Indicates Q i The total number of horizontal non-zero singular values ​​in P i is the i-th wavelet image G i The non-zero singular value sequence of the sub-band in the vertical direction, and P i ={P i,t | 1 ≤t ≤T i}, P i,t Indicates Q i The t-th vertical non-zero singular value in T i Indicates Q i The total number of vertical non-zero singular values ​​in ;

[0039] Step 2.3.2, traverse {(Q i ,P i ) | 1 ≤ i≤ N}, remove Q in turn i 、P i The non-zero singular value that is smaller than the median is obtained, and the set of non-zero singular value sequences after removal is obtained {(Q i ',P i ') | 1≤i≤N}; where Q i ' is the i-th wavelet image G i The non-zero singular value sequence after sub-band removal in the horizontal direction; P i ' is the i-th wavelet image G i The sequence of non-zero singular values ​​after sub-band removal in the vertical direction;

[0040] Step 2.3.3, traverse {(Q i ',P i ') | 1≤i≤N}, calculate Q respectively i The average value of Q i,ave ' and P i The average value of 'P i,ave ', and select Q respectively i 'Middle is greater than Q i,ave ' non-zero singular values ​​and P i ' is greater than P i,ave After the non-zero singular values ​​of ' are reconstructed, the reconstructed wavelet image set {G i ' | 1 ≤ i ≤ N}; where G i ' represents the wavelet image after reconstruction of the i-th layer;

[0041] Step 2.3.4. Calculate G i The information entropy H of the diagonal subband wavelet coefficients i,1 , and then use formula (3) to get the transformed information entropy H i,2 :

[0042] (3)

[0043] Step 2.3.5: Based on prior information, i,2 Perform polynomial fitting and calculate G i The noise variance σ of the diagonal subband of ' i :

[0044] Step 2.3.6: Using the Empirical Bayes formula and σ i Estimation {G i ' | 1 ≤ i ≤ N}, thus obtaining the real wavelet image set {Gi ''|1≤i≤N}, where G i '' represents the true wavelet image of layer i;

[0045] Step 2.3.7, for {G i Each layer of the real wavelet image within ''|1≤i≤N} is subjected to exponential transformation and inverse wavelet transformation in sequence, and then image reconstruction is performed to obtain an infrared image A4 without stripe noise and speckle noise.

[0046] Furthermore, the gradient weighted region growing method in step 3 includes the following steps:

[0047] Step 3.1: After converting A4 into a grayscale image A4' of the same scale, perform horizontal and vertical gradient detection on the grayscale image A4', and obtain the horizontal gradient image Grad x , longitudinal gradient map Grad y ;

[0048] Step 3.2: After converting A4 to HSV space, use the watershed algorithm to pre-segment the image after HSV space conversion to obtain the first sub-region set ,in, represents the u-th first sub-region, and The set of adjacent sub-regions of , express The vth adjacent subregion of express The total number of adjacent sub-regions;

[0049] Step 3.3, initialize u=1;

[0050] Step 3.4, calculation and The root square difference of the saturation components simS u , hue component root square difference simH u , and use formula (4) to calculate and The root cosquare difference of the gradient between u ;

[0051] (4)

[0052] In formula (6), and Respectively The mean of the horizontal gradient and the mean of the vertical gradient, and Represents sub-regions and the mean transverse gradient and the mean longitudinal gradient of all its adjacent sub-regions;

[0053] Step 3.5: Calculate using formula (5) The similarity function f u If f u Less than threshold , then Mark it as the seed region and go to step 3.6; otherwise, go directly to step 3.6;

[0054] (5)

[0055] In formula (5), 、 、 are 3 weight coefficients respectively;

[0056] Step 3.6: If u is less than sub1, assign u+1 to u and return to step 3.4 to execute sequentially; otherwise, the seed region set is obtained;

[0057] Step 3.7: According to the seed growth rule, all seed regions are selected from the seed region set for growth to obtain a second sub-region set;

[0058] Step 3.8: Perform rectangle fitting on all sub-regions in the second sub-region set, and remove the rectangles whose aspect ratio after fitting is greater than the threshold. , and arrange the remaining subregions in descending order according to the distance from the geometric center to the center of A4 to obtain a third subregion set. At the same time, arrange the remaining subregions in descending order according to their own area to obtain a fourth subregion set. The subregion with the smallest sum of indexes in the third subregion set and the fourth subregion set is selected as the segmented infrared image A5.

[0059] Furthermore, the step 4 includes the following steps:

[0060] Step 4.1: Grayscale A5 and draw a grayscale histogram based on the grayscale values ​​of the pixels. Then, the kernel estimation method is used to obtain the grayscale probability density function F(t) of A5, where t represents the grayscale value and t∈(T min ,T max ); where T min 、T max are the minimum and maximum values ​​in the grayscale histogram respectively;

[0061] Step 4.2: Use the Otsu algorithm to obtain the segmentation threshold T of F(t), and take the pixel points in A5 with pixel values ​​less than T as background points, and the pixel points with pixel values ​​greater than T as the wave trap area points; thus, use equations (6) and (7) to obtain the average grayscale value T of the wave trap area respectively. device and the average gray value T of the background areaenv ;

[0062] (6)

[0063] (7)

[0064] Step 4.3, draw a three-dimensional grayscale distribution map of A2 on a rectangular coordinate system, where the x-axis and y-axis of the rectangular coordinate system represent the horizontal coordinate and vertical coordinate of the image pixel point, and the z-axis represents the grayscale value corresponding to the pixel point;

[0065] Define plane z0 and initialize it to z0=T device ;

[0066] Step 4.4, Assign a value to z0. If there is a point tangent to z=z0 in the three-dimensional grayscale distribution graph, store the x and y axis coordinates of the tangent point in the extreme point set and go to step 4.5. Otherwise, go directly to step 4.5. is the set parameter;

[0067] Step 4.5: Determine whether z0 is less than T max If yes, go to step 4.4; otherwise, output the coordinate set of the effective extreme point {(x m ,y m )|1≤m≤E max}, where (x m ,y m ) represents the coordinates of the mth effective extreme point, E max Indicates the total number of extreme points.

[0068] Furthermore, the step 5 comprises the following steps:

[0069] Step 5.1: Take each coordinate in the effective extreme point set as the cluster center, use the KNN algorithm to cluster the points in the wave trap area, and obtain the cluster information matrix , where A max represents the total number of cluster regions, S e represents the area of ​​the e-th cluster region, T e represents the average gray value of the e-th cluster area; represents the relative temperature difference of the e-th cluster area, and ;W e represents the weight of the e-th cluster area, and ;

[0070] Step 5.2: Calculate the threshold T using formula (8) th :

[0071] (8)

[0072] In formula (8), is the correction factor;

[0073] Step 5.3, initialize e=1;

[0074] Step 5.4: If Greater than the threshold T th , then according to the position of the e-th cluster area in the infrared image A1, draw a mask at the corresponding position in the infrared image A1 and go to step 5.5; otherwise, go directly to step 5.5;

[0075] Step 5.5: If e is less than A max , then e+1 is assigned to e, and the process goes to step 5.4 for sequential execution; otherwise, it means that the infrared image with the mask is obtained and output as the thermal anomaly positioning image A6.

[0076] An electronic device of the present invention includes a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the high-noise-resistance wave trap thermal fault locating method, and the processor is configured to execute the program stored in the memory.

[0077] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program executes the steps of the method for locating thermal faults of high-noise-resistant wave traps when the computer program is run by a processor.

[0078] Compared with the prior art, the present invention has the following beneficial effects:

[0079] 1. Since the stripe noise in the image is vertical and the pixel values ​​of the noise pixels in the same column are similar, the present invention designs a new filtering operator. The mean pixel value of each column of the grayscale image is used as the processing unit to better meet the characteristics of stripe noise. Since the stripe noise in different columns of the image has different intensities and the differences with adjacent real patterns are also different, the new filtering operator combines horizontal gradient information and performs numerical compensation on the noise pixels according to the rate of change of pixel values ​​in the horizontal direction and the noise intensity. This can achieve the effect of adaptive smoothing and alleviate the blurring effect caused by over-smoothing in weak noise areas. Based on the continuity of pixel values ​​in the real pattern, the new filtering operator sets weights for the horizontal distances from different columns to the image center according to the Gaussian distribution, which strengthens the influence of close-range pixel information on the operator center and suppresses image distortion caused by long-range information.

[0080] 2. The column-weighted and layered filtering algorithm proposed in this invention smoothes the original image to obtain a smoothed image. The original image and the smoothed image are then subtracted and subjected to horizontal high-pass filtering to produce a high-frequency image containing significant noise. Simultaneously, the high-frequency image is filtered again based on horizontal gradient information to obtain a smoothed high-frequency image. This smoothed high-frequency image is then added to the smoothed image to produce a denoised image. Compared to single filtering, smoothing a noisy high-frequency image mitigates information loss caused by oversmoothing. The denoised image obtained using this method further mitigates the blurring effect caused by smoothing.

[0081] 3. The linear fitting algorithm based on SVD decomposition and information entropy proposed in the present invention uses an adaptive screening method to remove smaller singular values ​​in the horizontal subbands and vertical subbands obtained by SVD decomposition of the wavelet image, and can retain larger singular values ​​to obtain a true pattern; at the same time, compared with the traditional Donoho estimation method, when the noise intensity is reduced, the linear relationship between the entropy value of the noisy image in the wavelet subband and the noise variance is used to estimate the noise energy variance of the diagonal subband of the wavelet image, which is closer to the true value and less susceptible to interference from image details.

[0082] 4. This invention uses an improved seed region growing method to extract device image regions, and pre-segments the image using a watershed algorithm to generate a large number of subregions. Taking into account the hollowed-out appearance of wave trap devices, this method incorporates gradient information as a new similarity criterion, building on the regional similarity and Euclidean distance criteria. This allows for the selection of more appropriate seed regions, avoiding the generation of erroneous regions due to improper seed selection. Furthermore, the distance from the region to the image center and the region's area are used as screening criteria to further filter out authentic device images, thereby mitigating interference from external background devices in the image.

[0083] 5. This invention uses the Otsu algorithm to derive the pixel value ranges of background and device region points from the grayscale histogram, thereby further distinguishing the background and device regions in the image and improving the segmentation accuracy of device hot regions. Furthermore, based on the known condition that pixel grayscale information can reflect temperature information, a three-dimensional grayscale distribution map is constructed. Starting from the average grayscale value of the background region, extreme points are searched for in high-pixel value regions and used as cluster centers to identify high-temperature points in the two-dimensional image, thereby reducing invalid sampling and improving clustering speed.

[0084] 6. In the case of voltage-induced heating problems, the present invention, based on the characteristic that the area of ​​abnormal heating areas is small, introduces the area of ​​clustered areas as an influencing factor to determine the abnormal heating threshold, thereby reducing the false positive rate of thermal anomaly recognition when the heating phenomenon is not obvious. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 Flow chart of the steps of the present invention;

[0086] Figure 2 A schematic diagram of wavelet decomposition in the present application;

[0087] Figure 3 A comparison chart of noise standard deviation estimation results in the present application. DETAILED DESCRIPTION

[0088] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. In order to make the purpose, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0089] In the present embodiment, a high-noise-resistance wave trap thermal fault positioning method based on thermal region segmentation is shown in FIG. 1, which is performed according to the following steps: Figure 1

[0090] Step 1, an infrared image A1 with a dimension of WxH is collected, wherein W and H respectively represent the width and height of the infrared image A1; in the present embodiment, W is 1024 and H is 1024.

[0091] Step 2, the A1 is denoised to obtain a denoised infrared image A4;

[0092] Step 2.1, after the A1 is converted into a gray-scale image A2 with the same scale, the column weighting and layered filtering algorithm is used to process the A2 to obtain an infrared image A3 without stripe noise.

[0093] Step 2.1.1, the gradient operator is used to process the gray-scale image A2 to obtain a horizontal gradient image Grad;

[0094] Step 2.1.2, a column weighting filtering window is defined as:

[0095] A filtering window with a dimension of (2k+1)x(2k+1) is defined, wherein 2k+1 is the dimension of the filtering window, and in the present embodiment, k takes a positive integer 3.

[0096] A loop program is constructed to realize the traversal of the pixel points, the pixel value S(i,j) of the center position (i,j) of the filtering window in the A2 is obtained by using formula (1), and the boundary constraint condition of the filtering window is set as: Accordingly, the sliding window image B1 is obtained by sliding on the gray-scale image A2 in the order from left to right and from top to bottom according to the boundary constraint condition.

[0097] (1)​​

[0098] In formula (1), represents the mean pixel value in the xth column of A2, and , represents the pixel value at coordinate (x, y) in A2; Indicates about A x The correction function of , represents an artificially determined regulatory factor. In this embodiment, The value is 0.2, represents the mean pixel value of the x+1th column in A2, sgn(·) represents the sign function, represents the mean pixel value of the xth column in Grad, and , represents the pixel value at coordinate (x, y) in Grad; f(x) represents the column weight of the x-th column in A2, and , σ is the variance of the column weight determined artificially. In this embodiment, σ is set to 1.2.

[0099] Step 2.1.3: Subtract B1 from A2 to obtain the difference image B2. High-pass filter B2 is then processed in the column direction to obtain the high-frequency information image B3.

[0100] Step 2.1.4: Use equation (2) to process B3 and obtain the corrected image B4:

[0101] (2)

[0102] In formula (2), B4(x,y) and B3(x,y) represent the pixel values ​​at coordinates (x,y) in B4 and B3 respectively; represents the average value of the pixels in B3; λ is an artificially determined noise reduction factor; in this embodiment, λ is set to 0.15.

[0103] Step 2.1.5: Add B1 and B4 to obtain the true grayscale image B5, and convert B5 to the same color space as A1 to obtain the infrared image A3 without stripe noise.

[0104] Step 2.2: After taking the logarithmic transformation of A3, perform N-layer wavelet decomposition of A3 using the Coiflet wavelet basis, as follows: Figure 2 As shown, in this embodiment, three-layer wavelet decomposition is performed, and N is 3; and Figure 2 Where LH1, LH2, LH3 are horizontal sub-bands, HL1, HL2, HL3 are vertical sub-bands, HH1, HH2, HH3 are diagonal sub-bands; the wavelet image set {G i | 1 ≤ i ≤ N}, where Gi Represents the wavelet image obtained by decomposition at the i-th layer.

[0105] Step 2.3, use the linear fitting algorithm based on SVD decomposition and information entropy to fit {G i |1≤i≤N} is processed to obtain an infrared image A4 without stripe noise and speckle noise;

[0106] Step 2.3.1, respectively, for the wavelet image set {G i Each wavelet image in | 1 ≤ i ≤ N} is decomposed by SVD in the horizontal sub-band and vertical sub-band directions to obtain a set of non-zero singular value sequences {(Q i , P i )| 1 ≤ i ≤ N}, where Q i is the i-th wavelet image G i The non-zero singular value sequence of the sub-band in the horizontal direction, and Q i ={Q i,s | 1 ≤ s ≤S i}, Q i,s Represents Q i The s-th horizontal non-zero singular value in S i Represents Q i The total number of horizontal non-zero singular values ​​in P i is the i-th wavelet image G i The non-zero singular value sequence of the sub-band in the vertical direction, and P i ={P i,t | 1 ≤t ≤T i}, P i,t Represents Q i The t-th vertical non-zero singular value in T i Indicates Q i The total number of vertical nonzero singular values ​​in .

[0107] Step 2.3.2, traverse {(Q i ,P i ) | 1 ≤ i≤ N}, remove Q in turn i 、P i The non-zero singular value that is smaller than the median is obtained, and the set of non-zero singular value sequences after removal is obtained {(Q i ',P i ') | 1≤i≤N}; where Q i ' is the i-th wavelet image G i The non-zero singular value sequence after sub-band removal in the horizontal direction; P i ' is the i-th wavelet image G i The sequence of non-zero singular values ​​after sub-band removal in the vertical direction;

[0108] Step 2.3.3, traverse {(Q i ’, P i ’ | 1≤i≤N} respectively to calculate the average value Q i ’ of Q i,ave ’ and the average value P i ’ of P i,ave ’, and select the non-zero singular values greater than Q i ’ in Q i,ave ’ and the non-zero singular values greater than P i ’ in P i,ave ’ respectively to reconstruct to obtain the reconstructed wavelet image set {G i ’ | 1≤i≤N}; wherein G i ’ represents the reconstructed wavelet image of the i-th layer.

[0109] Step 2.3.4, calculate the information entropy H i ’ of the diagonal sub-band wavelet coefficients of G i,1 ’, so as to obtain the transformed information entropy H i,2 ’ by formula (3):

[0110] (3)

[0111] Step 2.3.5, polynomial fitting is performed on H i,2 based on prior information, and the noise variance σ i of the diagonal sub-band of G i ’ is calculated: in this embodiment, through the data statistics of 240 images, the calculation formula of σ i is obtained as follows:

[0112]

[0113] Step 2.3.6: the real wavelet coefficients of {G i ’ | 1≤i≤N} are estimated by using the empirical Bayes formula and σ i , so as to obtain the real wavelet image set {G i ’’ | 1≤i≤N}, wherein G i ’’ represents the i-th layer real wavelet image; in this embodiment, the Bayes empirical formula used is as formula (3a):

[0114] (3a)

[0115] In formula (3a), represents the j-th real wavelet coefficient in the diagonal sub-band of G i ’’, represents the j-th wavelet coefficient in the diagonal sub-band of G i ’, Represents G i 'The observed variance of the jth wavelet coefficient in the diagonal subband.

[0116] like Figure 3 As shown, is the noise standard deviation of the preset value during simulation and verification of this embodiment, which are 15, 20, 25, 30, and 35 respectively. is the standard deviation estimated by this method, is the standard deviation obtained by the Donoho estimation method. After testing 500 images, the method used in this embodiment more accurately estimates the true wavelet coefficients compared to the traditional Donoho estimation method. This method significantly outperforms the Donoho estimation method, especially when the noise variance is small.

[0117] Step 2.3.7, for {G i Each layer of the real wavelet image within '' | 1 ≤ i ≤ N} is subjected to exponential transformation and inverse wavelet transformation in sequence, and then image reconstruction is performed to obtain an infrared image A4 without stripe noise and speckle noise.

[0118] Step 3: Use the gradient weighted region growing method to segment the infrared image A4 to obtain a segmented infrared image A5;

[0119] Step 3.1: After converting A4 into a grayscale image A4' of the same scale, perform horizontal and vertical gradient detection on the grayscale image A4', and obtain the horizontal gradient image Grad x , longitudinal gradient map Grad y ;

[0120] Step 3.2: After converting A4 to HSV space, use the watershed algorithm to pre-segment the image after HSV space conversion to obtain the first sub-region set ,in, represents the u-th first sub-region, and The set of adjacent sub-regions of , express The vth adjacent subregion of express The total number of adjacent sub-regions.

[0121] Step 3.3, initialize u = 1;

[0122] Step 3.4, first calculate and The root square difference of the saturation components simS u In this embodiment, simS u It is obtained from the following formula (3b):

[0123] (3b)

[0124] In formula (3b), express The mean of the saturation component, express All regions and The mean of the saturation component.

[0125] Secondly, calculate the root square difference of the hue component simH u In this embodiment, simH u It is obtained from the following formula (3c):

[0126] (3c)

[0127] In formula (3c), express The mean value of the hue component, express All regions and The mean hue component of .

[0128] Finally, use formula (4) to calculate and The root cosquare difference of the gradient between u ;

[0129] (4)

[0130] In formula (4), and Respectively The mean of the horizontal gradient and the mean of the vertical gradient, and Represents sub-regions and the mean transverse and longitudinal gradients of all its adjacent sub-regions.

[0131] Step 3.5: Calculate using formula (5) The similarity function f u If f u Less than threshold , then Mark it as the seed region and go to step 3.6; otherwise, go directly to step 3.6;

[0132] (5)

[0133] In formula (5), 、 、 are three weight coefficients respectively. In this embodiment, Take 0.6, Take 0.2, Take 0.2.

[0134] Step 3.6: If u is less than sub1, assign u+1 to u and return to step 3.4 to execute sequentially; otherwise, the seed region set is obtained;

[0135] Step 3.7: According to the seed growth rule, all seed regions are selected from the seed region set for growth to obtain a second sub-region set.

[0136] Step 3.8: Perform rectangle fitting on all sub-regions in the second sub-region set, and remove the rectangles whose aspect ratio after fitting is greater than the threshold. , and arrange the remaining subregions in descending order according to the distance from the geometric center to the center of A4 to obtain a third subregion set. At the same time, arrange the remaining subregions in descending order according to their own area to obtain a fourth subregion set. The subregion with the smallest sum of indexes in the third subregion set and the fourth subregion set is selected as the segmented infrared image A5.

[0137] Step 4: According to the grayscale histogram of the segmented infrared image A5, the grayscale probability density function is obtained, and the average grayscale value T of the wave trap area in A1 is obtained by using the grayscale probability density function. device , and then obtain the effective extreme point set of the wave trap area in the three-dimensional grayscale distribution map corresponding to A5;

[0138] Step 4.1: Grayscale A5 and draw a grayscale histogram based on the grayscale values ​​of the pixels. Then, the kernel estimation method is used to obtain the grayscale probability density function F(t) of A5, where t represents the grayscale value and t∈(T min ,T max ); where T min 、T max are the minimum and maximum values ​​in the grayscale histogram respectively; in this embodiment, the kernel function used is the Gaussian distribution kernel function.

[0139] Step 4.2: Use the Otsu algorithm to obtain the segmentation threshold T of F(t), and take the pixel points in A5 with pixel values ​​less than T as background points, and the pixel points with pixel values ​​greater than T as the wave trap area points; thus, use equations (6) and (7) to obtain the average grayscale value T of the wave trap area respectively. device and the average gray value T of the background area env ;

[0140] (6)

[0141] (7)

[0142] The principle of this step is: there are fewer abnormally hot pixels in the device area, and thus fewer pixels with high grayscale values. Therefore, the grayscale histogram can be approximately composed of two peaks, which correspond to the peak of the background area and the peak of the device area respectively. The hot area can be ignored, so the Otsu algorithm for handling binary classification problems can be introduced.

[0143] Step 4.3, draw a three-dimensional grayscale distribution map of A2 on a rectangular coordinate system, where the x-axis and y-axis of the rectangular coordinate system represent the horizontal coordinate and vertical coordinate of the image pixel point, and the z-axis represents the grayscale value corresponding to the pixel point;

[0144] Define plane z0 and initialize it to z0=T device ;

[0145] Step 4.4, Assign a value to z0. If there is a point tangent to z=z0 in the three-dimensional grayscale distribution graph, store the x and y axis coordinates of the tangent point in the extreme point set and go to step 4.5. Otherwise, go directly to step 4.5. is the set parameter;

[0146] Step 4.5: Determine whether z0 is less than T max If yes, go to step 4.4; otherwise, output the coordinate set of the effective extreme point {(x m ,y m )|1 ≤ m ≤E max}, where (x m ,y m ) represents the coordinates of the mth effective extreme point, E max Indicates the total number of extreme points.

[0147] Step 5: Use each coordinate in the set of valid extreme points as the cluster center, and use the KNN algorithm to cluster the pixels in A5 to obtain the cluster information matrix. Then, according to the threshold condition, find the thermal anomaly area of ​​the wave trap from the cluster information matrix, and draw a mask at the corresponding position in A1 to obtain the thermal anomaly positioning image A6 of the wave trap.

[0148] Step 5.1: Take each coordinate in the set of effective extreme points as the cluster center, use the KNN algorithm to cluster the points in the wave trap area, and obtain the cluster information matrix:

[0149]

[0150] Among them, A max represents the total number of cluster regions, S e represents the area of ​​the e-th cluster region, T e represents the average gray value of the e-th cluster area; represents the relative gray difference of the e-th cluster region, and ; W e represents the weight of the e-th cluster region, and .

[0151] Step 5.2, calculating the threshold T by using formula (8) th , in formula (8), is a correction factor;

[0152] (8)

[0153] In the embodiment, is 7; in the places where the embodiment is implemented, generally, T device is about 150, T env is about 65; according to the actual inspection, the W e of the abnormal heat point is about 0.02; generally, when the gray value exceeds 210, it can be considered that the heat is abnormal.

[0154] Step 5.3, initializing e = 1, and starting to traverse the cluster information matrix from the first row;

[0155] Step 5.4, if is greater than the threshold T th , then according to the position of the e-th cluster region in the infrared image A1, a mask is drawn on the corresponding position in the infrared image A1, and step 5.5 is entered; otherwise, step 5.5 is directly entered;

[0156] Step 5.5, if e is less than A max , then e + 1 is assigned to e, and step 5.4 is sequentially executed; otherwise, the infrared image with the mask is obtained and output as a heat anomaly positioning image A6.

[0157] In the embodiment, an electronic device includes a memory and a processor, the memory is used to store a program supporting the processor to execute the above method, and the processor is configured to execute the program stored in the memory.

[0158] In the embodiment, a computer readable storage medium has a computer program stored thereon, and the computer program is executed by a processor to perform the steps of the above method.

Claims

1. A method for locating thermal faults of high noise resistant wave traps based on thermal region segmentation, characterized in that: Follow these steps: Step 1: Acquire an infrared image A1 of the wave trap with a dimension of W×H, where W and H represent the width and height of the infrared image A1 respectively; Step 2: De-noise A1 to obtain the de-noised infrared image A4. Step 2.1: After converting A1 into a grayscale image A2 of the same scale, A2 is processed using a column-weighted and layered filtering algorithm. Specifically, the filter window is constructed using the horizontal gradient map and a column-weighted function based on a Gaussian distribution to obtain an infrared image A3 without streak noise. Step 2.2: After logarithmic transformation of A3, perform N-layer wavelet decomposition of A3 using Coiflet wavelet basis to obtain the wavelet image set {G i |1≤i≤N}, where G i represents the wavelet image obtained by decomposition of the i-th layer; Step 2.3, use the linear fitting algorithm based on SVD decomposition and information entropy to fit {G i |1≤i≤N} is processed, specifically including: performing SVD decomposition on the wavelet image, screening the singular value sequence, estimating the noise variance based on information entropy linear fitting, and then estimating the true wavelet coefficients using the empirical Bayes formula to obtain an infrared image A4 without streak noise and speckle noise; Step 3: Use the gradient weighted region growing method to segment the infrared image A4 to obtain a segmented infrared image A5; The gradient weighted region growing method in step 3 comprises the following steps: Step 3.1: After converting A4 into a grayscale image A4' of the same scale, perform horizontal and vertical gradient detection on the grayscale image A4', and obtain the horizontal gradient image Grad x , longitudinal gradient map Grad y ; Step 3.2: After converting A4 to HSV space, use the watershed algorithm to pre-segment the image after HSV space conversion to obtain the first sub-region set ,in, represents the u-th first sub-region, and The set of adjacent sub-regions of , express The vth adjacent subregion of express The total number of adjacent sub-regions; Step 3.3, initialize u=1; Step 3.4, calculation and The root square difference of the saturation components simS u , hue component root square difference simH u , and calculate and The root cosquare difference of the gradient between u ; Step 3.5, calculation The similarity function f u If f u Less than threshold , then Mark it as the seed region and go to step 3.6; otherwise, go directly to step 3.6; Step 3.6: If u is less than sub1, assign u+1 to u and return to step 3.4 to execute sequentially; otherwise, the seed region set is obtained; Step 3.7: According to the seed growth rule, all seed regions are selected from the seed region set for growth to obtain a second sub-region set; Step 3.8: Perform rectangle fitting on all sub-regions in the second sub-region set, and remove the rectangles whose aspect ratio after fitting is greater than the threshold. The remaining sub-regions are sorted in descending order according to the distance from the geometric center to the center of A4 to obtain a third sub-region set. At the same time, the remaining sub-regions are sorted in descending order according to their own area to obtain a fourth sub-region set. The sub-region with the smallest sum of indexes in the third sub-region set and the fourth sub-region set is selected as the segmented infrared image A5; Step 4: According to the grayscale histogram of the segmented infrared image A5, the grayscale probability density function is obtained, and the average grayscale value T of the wave trap area in A1 is obtained by using the grayscale probability density function. device , and then obtain the effective extreme point set of the wave trap area in the three-dimensional grayscale distribution map corresponding to A5; Step 5: Take each coordinate in the set of valid extreme points as the cluster center, and use the KNN algorithm to cluster the pixels in A5 to obtain the cluster information matrix. Then, according to the threshold condition, find the thermal anomaly area of ​​the wave trap from the cluster information matrix, and draw a mask at the corresponding position in A1 to obtain the thermal anomaly positioning image A6 of the wave trap.

2. The method for locating thermal faults of high noise resistant wave traps based on thermal region segmentation according to claim 1, characterized in that: The column weighting and layered filtering algorithm of step 2.1 includes the following steps: Step 2.1.1, use the gradient operator to process the grayscale image A2 to obtain the horizontal gradient map Grad; Step 2.1.2, define the column weighted filter window: Define a filter window of dimension (2k+1)×(2k+1), where 2k+1 is the dimension of the filter window and k is a positive integer; The pixel value S(i, j) at the center position (i, j) of the filter window in A2 is obtained using formula (1), and the boundary constraint condition of the filter window is set to: , , thereby sliding from left to right and from top to bottom on the grayscale image A2 according to the boundary constraint conditions to obtain a sliding window image B1 containing low-frequency information; (1) In formula (1), represents the mean pixel value in the xth column of A2, and , represents the pixel value at coordinate (x, y) in A2; Indicates about A x The correction function of , represents an artificially determined regulatory factor, represents the mean pixel value of the x+1th column in A2, sgn(·) represents the sign function, represents the mean pixel value of the xth column in Grad, and , Represents the pixel value at coordinate (x, y) in Grad; represents the column weight of the x-th column in A2, and , σ is the variance of the set column weights; Step 2.1.3: Subtract B1 from A2 to obtain the difference image B2. High-pass filter B2 is then processed in the column direction to obtain the high-frequency information image B3. Step 2.1.4: Use equation (2) to process B3 and obtain the corrected image B4: (2) In formula (2), B4(x,y) and B3(x,y) represent the pixel values ​​at coordinates (x,y) in B4 and B3 respectively; represents the average value of pixels in B3; λ is an artificially determined noise reduction factor; Step 2.1.5: Add B1 and B4 to obtain the true grayscale image B5, and convert B5 to the same color space as A1 to obtain the infrared image A3 without stripe noise.

3. The method for locating thermal faults of a high noise resistant wave trap based on thermal region segmentation according to claim 2, characterized in that: The linear fitting algorithm based on SVD decomposition and information entropy in step 2.3 includes the following steps: Step 2.3.1, respectively, for the wavelet image set {G i Each wavelet image in | 1 ≤ i ≤ N} is decomposed by SVD in the horizontal sub-band and vertical sub-band directions to obtain a set of non-zero singular value sequences {(Q i , P i )| 1 ≤ i ≤ N}, where Q i is the i-th wavelet image G i The non-zero singular value sequence of the sub-band in the horizontal direction, and Q i ={Q i,s | 1 ≤ s ≤S i }, Q i,s Indicates Q i The s-th horizontal non-zero singular value in S i Indicates Q i The total number of horizontal non-zero singular values ​​in P i is the i-th wavelet image G i The non-zero singular value sequence of the sub-band in the vertical direction, and P i ={P i,t | 1 ≤t ≤T i }, P i,t Indicates Q i The t-th vertical non-zero singular value in T i Indicates Q i The total number of vertical non-zero singular values ​​in ; Step 2.3.2, traverse {(Q i ,P i ) | 1 ≤ i≤ N}, remove Q in turn i 、P i The non-zero singular value that is smaller than the median is obtained, and the set of non-zero singular value sequences after removal is obtained {(Q i ',P i ') | 1≤i≤N}; where Q i ' is the i-th wavelet image G i The non-zero singular value sequence after sub-band removal in the horizontal direction; P i ' is the i-th wavelet image G i The sequence of non-zero singular values ​​after sub-band removal in the vertical direction; Step 2.3.3, traverse {(Q i ',P i ') | 1≤i≤N}, calculate Q respectively i The average value of Q i,ave ' and P i The average value of 'P i,ave ', and select Q respectively i 'Middle is greater than Q i,ave ' non-zero singular values ​​and P i ' is greater than P i,ave After the non-zero singular values ​​of ' are reconstructed, the reconstructed wavelet image set {G i ' | 1 ≤ i ≤ N}; where G i ' represents the wavelet image after reconstruction of the i-th layer; Step 2.3.

4. Calculate G i The information entropy H of the diagonal subband wavelet coefficients i,1 , and then use formula (3) to get the transformed information entropy H i,2 : (3) Step 2.3.5: Based on prior information, i,2 Perform polynomial fitting and calculate G i The noise variance σ of the diagonal subband of ' i : Step 2.3.6: Using the Empirical Bayes formula and σ i Estimation {G i ' | 1 ≤ i ≤ N}, thus obtaining the real wavelet image set {G i ''|1≤i≤N}, where G i '' represents the true wavelet image of layer i; Step 2.3.7, for {G i Each layer of the real wavelet image within ''|1≤i≤N} is subjected to exponential transformation and inverse wavelet transformation in sequence, and then image reconstruction is performed to obtain an infrared image A4 without stripe noise and speckle noise.

4. The method for locating thermal faults of a high noise resistant wave trap based on thermal region segmentation according to claim 3, characterized in that: Step 4 comprises the following steps: Step 4.1: Grayscale A5 and draw a grayscale histogram based on the grayscale values ​​of the pixels. Then, the kernel estimation method is used to obtain the grayscale probability density function F(t) of A5, where t represents the grayscale value and t∈(T min ,T max ); where T min 、T max are the minimum and maximum values ​​in the grayscale histogram respectively; Step 4.2: Use the Otsu algorithm to obtain the segmentation threshold T of F(t), and take the pixel points in A5 with pixel values ​​less than T as background points, and the pixel points with pixel values ​​greater than T as the wave trap area points; thus, use equations (6) and (7) to obtain the average grayscale value T of the wave trap area respectively. device and the average gray value T of the background area env ; (6) (7) Step 4.3, draw a three-dimensional grayscale distribution map of A2 on a rectangular coordinate system, where the x-axis and y-axis of the rectangular coordinate system represent the horizontal coordinate and vertical coordinate of the image pixel point, and the z-axis represents the grayscale value corresponding to the pixel point; Define plane z0 and initialize it to z0=T device ; Step 4.4, Assign a value to z0. If there is a point tangent to z=z0 in the three-dimensional grayscale distribution graph, store the x and y axis coordinates of the tangent point in the extreme point set and go to step 4.

5. Otherwise, go directly to step 4.

5. is the set parameter; Step 4.5: Determine whether z0 is less than T max If yes, go to step 4.4; otherwise, output the coordinate set of the effective extreme point {(x m ,y m )|1≤m≤E max }, where (x m ,y m ) represents the coordinates of the mth effective extreme point, E max Indicates the total number of extreme points.

5. The method for locating thermal faults of high noise resistant wave traps based on thermal region segmentation according to claim 4, characterized in that: The step 5 comprises the following steps: Step 5.1: Take each coordinate in the effective extreme point set as the cluster center, use the KNN algorithm to cluster the points in the wave trap area, and obtain the cluster information matrix , where A max represents the total number of cluster regions, S e represents the area of ​​the e-th cluster region, T e represents the average gray value of the e-th cluster area; represents the relative temperature difference of the e-th cluster area, and ;W e represents the weight of the e-th cluster area, and ; Step 5.2: Calculate the threshold T using formula (8) th : (8) In formula (8), is the correction factor; Step 5.3, initialize e=1; Step 5.4: If Greater than the threshold T th , then according to the position of the e-th cluster area in the infrared image A1, draw a mask at the corresponding position in the infrared image A1 and go to step 5.5; otherwise, go directly to step 5.5; Step 5.5: If e is less than A max , then e+1 is assigned to e, and the process goes to step 5.4 for sequential execution; otherwise, it means that an infrared image with a mask is obtained and output as the thermal anomaly positioning image A6.

6. An electronic device comprising a memory and a processor, characterized in that: The memory is used to store a program that supports the processor to execute the high-noise-resistant wave trap thermal fault location method according to any one of claims 1 to 5, and the processor is configured to execute the program stored in the memory.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for locating thermal faults of a high-noise-resistant wave trap according to any one of claims 1 to 5 are executed.

Images