A fault detection method and device based on a high-temperature resistant camera

Abstract

This invention relates to the field of fault detection technology, and discloses a fault detection method and apparatus based on a high-temperature camera. The method includes acquiring infrared and visible light images of the surface of high-temperature equipment, performing registration processing to obtain fused image data; performing temperature field gradient analysis based on the fused image data to obtain temperature gradient distribution data; identifying and locating temperature anomaly regions based on the temperature gradient distribution data to obtain temperature anomaly region data; acquiring texture features of the visible light image and performing similarity analysis to obtain deformation region data of the equipment surface; calculating the correlation between the temperature anomaly region data and the deformation region data of the equipment surface to determine whether a causal relationship exists; if a causal relationship exists and the deformation region data of the equipment surface is greater than a preset area threshold, then the equipment is determined to have a fault risk, and a fault detection result is output. This method can achieve accurate fault early warning and spatial location of high-temperature equipment.

Description

Technical Field

[0001] This invention relates to the field of fault detection technology, and in particular to a fault detection method and apparatus based on a high-temperature resistant camera. Background Technology

[0002] In modern industrial production, real-time status monitoring of critical equipment operating under extreme environments such as high temperature and high pressure is crucial, directly impacting production safety and efficiency. Abnormal temperature or physical deformation on equipment surfaces are often early signs of potential failures. Therefore, real-time and accurate online monitoring of this equipment is a key link in ensuring production safety, extending equipment lifespan, and improving economic efficiency. Currently, visual inspection technology, as a non-contact online monitoring method, is widely used in the field of equipment fault early warning due to its high efficiency and intuitive characteristics.

[0003] In existing technologies, high-temperature equipment is typically monitored using a single detection method, such as infrared thermal imagers, which measure temperature and detect anomalies by capturing high-temperature points on the surface. However, such methods have significant limitations when dealing with complex industrial environments. Due to the lack of simultaneous observation of the physical morphology of the equipment surface, when a temperature anomaly is detected, the system struggles to determine whether the anomaly is caused by actual physical deformations such as cracks or material thinning, or simply a false alarm due to complex surface textures, stains, or background thermal radiation interference. This lack of multi-source information fusion capability prevents the detection system from accurately establishing a causal relationship between temperature changes and physical deformation, thereby reducing the accuracy and reliability of fault diagnosis.

[0004] Existing technologies suffer from insufficient accuracy in identifying complex fault characteristics. Summary of the Invention

[0005] This invention provides a fault detection method and device based on a high-temperature resistant camera to solve the problem of insufficient accuracy in identifying complex fault features.

[0006] Firstly, in order to solve the above-mentioned technical problems, the present invention provides a fault detection method based on a high-temperature resistant camera, comprising:

[0007] Infrared and visible light images of the surface of high-temperature equipment are acquired and registered to obtain fused image data.

[0008] Based on the fused image data, temperature field gradient analysis is performed to obtain temperature gradient distribution data.

[0009] Spatial continuity compensation and noise suppression are performed on the temperature gradient distribution data. If the gradient amplitude exceeds a preset noise threshold, it is determined to be an effective temperature jump signal, and the first temperature jump signal is obtained.

[0010] The first temperature jump signal is scanned in a region, the segmentation threshold is calculated and spatial continuity features are identified to obtain the first candidate abnormal region.

[0011] Based on the first candidate anomaly region, the spatial clustering degree of the anomaly region is calculated and an expansion operation is performed to obtain the second candidate anomaly temperature region.

[0012] Obtain the corresponding texture features from the second candidate abnormal temperature region, calculate the similarity between the texture features and the preset normal equipment surface texture, and if the similarity is less than the preset similarity threshold, it is determined to be a deformation region, and the equipment surface deformation region is obtained.

[0013] Calculate the correlation coefficient between the deformation area on the surface of the equipment and the second candidate abnormal temperature area. If the correlation coefficient is greater than a preset correlation threshold and the abnormal temperature area is greater than a preset lower limit of area, then it is determined that the equipment has a fault risk, and the fault detection result is output.

[0014] Preferably, infrared and visible light images of the surface of the high-temperature equipment are acquired and registered to obtain fused image data, including:

[0015] Obtain infrared radiation information from the surface of high-temperature equipment;

[0016] Based on the infrared radiation information, image acquisition is performed to obtain infrared images and visible light images;

[0017] Based on the infrared and visible light images, correction and registration are performed to obtain fused image data.

[0018] Preferably, the step of performing temperature field gradient analysis based on the fused image data to obtain temperature gradient distribution data includes:

[0019] Temperature information is extracted from the fused image data, and a convolution operation is performed to obtain the first-order and second-order gradient distributions of the temperature field.

[0020] Based on the first-order and second-order gradient distributions, normalization and convolution operations are performed to obtain a multi-level temperature gradient representation.

[0021] Based on the multi-level temperature gradient, weighting factors are calculated and assigned, and weighted fusion is performed to obtain temperature gradient distribution data.

[0022] Preferably, the step of performing spatial continuity compensation and noise suppression on the temperature gradient distribution data, and determining it as a valid temperature jump signal if the gradient amplitude exceeds a preset noise threshold, to obtain a first temperature jump signal, includes:

[0023] Based on the temperature gradient distribution data, spatial continuity compensation is performed to obtain a gradient vector field distribution map;

[0024] Based on the gradient vector field distribution map, noise suppression processing is performed to obtain a set of denoised gradient data.

[0025] Based on the denoised gradient data set, threshold detection and clustering are performed to obtain the first temperature jump signal.

[0026] Preferably, the step of performing a region scan on the first temperature jump signal, calculating a segmentation threshold, and identifying spatially continuous features to obtain a first candidate anomaly region includes:

[0027] Perform a window sliding scan on the first temperature jump signal and calculate the segmentation threshold to obtain the segmentation threshold parameter;

[0028] Based on the segmentation threshold parameter, the spatial continuity features of the temperature gradient are identified, and a spatial continuity feature marker map is obtained;

[0029] Based on the spatially continuous feature map, region growing is performed to obtain the first candidate anomaly region.

[0030] Preferably, the step of calculating the spatial clustering degree of the anomalous region and performing an expansion operation based on the first candidate anomalous region to obtain the second candidate anomalous temperature region includes:

[0031] Temperature anomaly points are extracted from the first candidate anomaly region, and spatial location analysis is performed on the temperature anomaly points to obtain a regional cluster density distribution map.

[0032] Based on the regional aggregation density distribution map, the location of the density peak is found and marked as a seed candidate point, thus obtaining the set of seed candidate point coordinates;

[0033] Based on the set of seed candidate point coordinates, an expansion operation is performed to obtain the growth results of the abnormal region;

[0034] Based on the growth results of the abnormal region, regional tracking is performed to obtain the second candidate abnormal temperature region.

[0035] Preferably, the step of obtaining the corresponding texture features from the second candidate abnormal temperature region, calculating the similarity between the texture features and the preset normal equipment surface texture, and determining the deformation region as a deformation region if the similarity is less than a preset similarity threshold, includes:

[0036] Obtain the corresponding visible light image region from the second candidate abnormal temperature region;

[0037] Texture features are extracted from the visible light image region to obtain the region texture features;

[0038] Calculate the similarity between the texture features of the region and the preset normal device surface texture. If the similarity is less than the preset similarity threshold, it is determined to be a deformed region, and the deformed region of the device surface is obtained.

[0039] Preferably, the step of calculating the correlation coefficient between the deformation area on the device surface and the second candidate abnormal temperature area, and if the correlation coefficient is greater than a preset correlation threshold and the abnormal temperature area is greater than a preset lower limit of area, then it is determined that the device has a fault risk, and the fault detection result is output, including:

[0040] Calculate the spatial overlap between the deformation area on the equipment surface and the second candidate abnormal temperature area to obtain the correlation coefficient;

[0041] The correlation coefficient is compared with a preset association threshold. If the correlation coefficient is greater than the preset association threshold, it is determined that there is a causal relationship between the temperature anomaly and the deformation, and the boundary data of the anomaly area is obtained.

[0042] Extract the total pixel area from the boundary data of the abnormal region. If the total pixel area is greater than a preset lower limit value, it is determined that the device has a fault risk, and the fault detection result is output.

[0043] Secondly, the present invention provides a fault detection device based on a high-temperature resistant camera, comprising:

[0044] The data acquisition and registration module is used to acquire infrared and visible light images of the surface of high-temperature equipment, and perform registration processing to obtain fused image data;

[0045] The gradient analysis module is used to perform temperature field gradient analysis based on the fused image data to obtain temperature gradient distribution data.

[0046] The signal recognition module is used to perform spatial continuity compensation and noise suppression on the temperature gradient distribution data. If the gradient amplitude exceeds a preset noise threshold, it is determined to be an effective temperature jump signal, and the first temperature jump signal is obtained.

[0047] An abnormal region identification module is used to perform region scanning on the first temperature jump signal, calculate the segmentation threshold and identify spatial continuity features to obtain the first candidate abnormal region.

[0048] An abnormal region location module is used to calculate the spatial clustering degree of the abnormal region based on the first candidate abnormal region and perform an expansion operation to obtain a second candidate abnormal temperature region.

[0049] The deformation region identification module is used to obtain the corresponding texture features from the second candidate abnormal temperature region, calculate the similarity between the texture features and the preset normal equipment surface texture, and if the similarity is less than the preset similarity threshold, it is determined to be a deformation region, and the deformation region of the equipment surface is obtained.

[0050] The fault risk assessment module is used to calculate the correlation coefficient between the deformation area on the surface of the equipment and the second candidate abnormal temperature area. If the correlation coefficient is greater than a preset correlation threshold and the abnormal temperature area is greater than a preset area lower limit, then the equipment is determined to have a fault risk, and the fault detection result is output.

[0051] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement a fault detection method based on a high-temperature resistant camera as described in any one of the above.

[0052] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute any one of the above-described fault detection methods based on a high-temperature resistant camera.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] (1) This invention acquires infrared thermal images and visible light images of high-temperature equipment simultaneously, and uses image registration and distortion correction techniques to achieve pixel-level spatial alignment, thus constructing a dual-modal data foundation that integrates temperature and texture information. Furthermore, it performs refined analysis of the temperature field through multi-scale gradient operators and pyramid structures, thereby achieving accurate capture of subtle temperature changes at different scales. This solves the problems of low signal-to-noise ratio and susceptibility to interference from complex textures on the equipment surface and background thermal radiation caused by the reliance on a single data source in existing technologies, and provides comprehensive and reliable data support for subsequent fault identification.

[0055] (2) By adopting adaptive threshold segmentation, spatial continuity judgment and a secondary region growth algorithm based on cluster density statistics, this invention establishes a coarse-to-fine abnormal region localization system, which can accurately separate and lock the real and continuous temperature abnormal region from complex background noise. It solves the problem that the existing technology is difficult to effectively distinguish between real abnormalities and random noise due to the use of fixed thresholds, which leads to inaccurate fault location or missed detection. It significantly improves the accuracy and reliability of abnormal region localization.

[0056] (3) This invention identifies physical deformation by extracting visible light image texture features corresponding to the located temperature anomaly area, and further calculates the spatial overlap and correlation coefficient between the temperature anomaly area and the deformation area, establishing a quantitative judgment mechanism for the causal relationship between temperature anomaly and physical deformation. This cross-validation of multimodal information solves the core pain point of existing technologies that cannot confirm whether temperature anomalies are caused by real physical damage, and realizes the leap from simple "phenomenon alarm" to high-confidence "fault confirmation", greatly reducing the false alarm rate. Attached Figure Description

[0057] Figure 1 This is a schematic flowchart of a fault detection method based on a high-temperature resistant camera provided in the first embodiment of the present invention;

[0058] Figure 2 This is a schematic diagram of a fault detection device based on a high-temperature resistant camera provided in the second embodiment of the present invention. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] Reference Figure 1 The first embodiment of the present invention provides a fault detection method based on a high-temperature resistant camera, comprising the following steps:

[0061] S11: Acquire infrared and visible light images of the surface of the high-temperature equipment, and perform registration processing to obtain fused image data;

[0062] S12, Based on the fused image data, perform temperature field gradient analysis to obtain temperature gradient distribution data;

[0063] S13, perform spatial continuity compensation and noise suppression on the temperature gradient distribution data. If the gradient amplitude exceeds the preset noise threshold, it is determined to be an effective temperature jump signal, and the first temperature jump signal is obtained.

[0064] S14, perform a region scan on the first temperature jump signal, calculate the segmentation threshold and identify spatial continuity features to obtain the first candidate abnormal region;

[0065] S15, Based on the first candidate abnormal region, calculate the spatial clustering degree of the abnormal region and perform an expansion operation to obtain the second candidate abnormal temperature region.

[0066] S16, Obtain the corresponding texture features from the second candidate abnormal temperature region, calculate the similarity between the texture features and the preset normal equipment surface texture, and if the similarity is less than the preset similarity threshold, determine it as a deformation region and obtain the equipment surface deformation region.

[0067] S17, calculate the correlation coefficient between the deformation area on the surface of the equipment and the second candidate abnormal temperature area. If the correlation coefficient is greater than the preset correlation threshold and the abnormal temperature area is greater than the preset area lower limit, then it is determined that the equipment has a fault risk, and the fault detection result is output.

[0068] In step S11, infrared and visible light images of the surface of the high-temperature equipment are acquired and registered to obtain fused image data, including:

[0069] Obtain infrared radiation information from the surface of high-temperature equipment;

[0070] Based on the infrared radiation information, image acquisition is performed to obtain infrared images and visible light images;

[0071] Based on the infrared and visible light images, correction and registration are performed to obtain fused image data.

[0072] In one implementation, this step is achieved through a dual-light camera system consisting of a high-resolution long-wave infrared thermal imager and a visible light industrial camera. First, the infrared thermal imager acquires the raw digital signal of infrared radiation from the equipment in real time and converts it into an actual temperature value using a built-in temperature calibration algorithm. When the temperature of a continuous N×N pixel area on the equipment surface exceeds a preset temperature trigger threshold (this threshold is set according to the equipment's safety operating specifications, such as 80 degrees Celsius), the system immediately sends a synchronous acquisition command, triggering the visible light industrial camera to take a picture at the same millisecond level.

[0073] After acquiring time-synchronized dual-modal image information, spatial alignment is required. This process is achieved through image registration techniques, specifically feature-based matching methods. For example, the Harris corner detection algorithm extracts corner features of hotspot regions in infrared images. This algorithm calculates the grayscale changes of pixels after movement in various directions, and its response function is... ,in The gradient covariance matrix is ​​2x2, and the sensitivity factor is... The value is typically set between 0.04 and 0.06. Simultaneously, the Canny edge detection algorithm is used to extract edge features of the device contour in the visible light image. By calculating descriptors (such as SIFT or SURF) for these feature points and matching them, a series of matching point pairs can be obtained. If the number of matching point pairs is less than a preset minimum feature matching threshold (e.g., 50), it indicates insufficient feature points. Based on these precise correspondences, the affine transformation matrix H required to map the infrared image to the visible light image coordinate system can be calculated.

[0074] Before registration, geometric distortion correction is necessary to eliminate image distortion caused by the lens itself. This process can be performed using intrinsic parameters obtained during camera calibration, through a radial distortion correction formula. Image processing, where For distorted coordinates, The distance from this point to the center of the image. , , This represents the radial distortion coefficient. In one specific setting, if the detected distortion level in the image edge region exceeds the maximum distortion threshold (e.g., 2 pixels), a tangential distortion correction term will be added for compensation.

[0075] Finally, the distortion-corrected infrared image is mapped to the coordinate space of the visible light image using the calculated affine transformation matrix H, establishing a pixel-level correspondence between the two images. This pixel-level correspondence means that for any pixel coordinate in the visible light image, the corresponding pixel coordinate in the infrared image, which is the same physical point observed by the same device, can be found using the transformation matrix H. Through weighted fusion processing of infrared radiation intensity information and visible light brightness information, a first fused image data containing accurate temperature information and rich texture details is finally generated. During the weighted fusion processing stage, the system dynamically adjusts the fusion weights based on the characteristics of the image content: for areas with higher temperatures, the system assigns higher weights to infrared information to highlight temperature anomalies; while for areas with rich texture details, the system increases the contribution of visible light information to preserve structural details.

[0076] It should be noted that this method of acquiring and fusing dual-modal images can provide a unified and information-rich perspective for subsequent fault analysis. Time-synchronized acquisition ensures the immediacy of the data, while precise geometric correction and registration criteria ensure the accuracy of spatial location. The registration accuracy can reach the sub-pixel level, laying the foundation for the subsequent precise correlation between temperature anomalies and physical deformation.

[0077] In step S12, temperature field gradient analysis is performed based on the fused image data to obtain temperature gradient distribution data, including:

[0078] Temperature information is extracted from the fused image data, and a convolution operation is performed to obtain the first-order and second-order gradient distributions of the temperature field.

[0079] Based on the first-order and second-order gradient distributions, normalization and convolution operations are performed to obtain a multi-level temperature gradient representation.

[0080] Based on the multi-level temperature gradient, weighting factors are calculated and assigned, and weighted fusion is performed to obtain temperature gradient distribution data.

[0081] In one implementation, this step first extracts temperature information from the fused image data to form a two-dimensional temperature matrix. Specifically, since the fused image data contains temperature information collected by an infrared thermal imager, the system can directly parse the temperature value corresponding to each pixel from the fused data, thus obtaining a two-dimensional temperature matrix. Each element in this matrix represents the temperature value at a specific physical location on the device surface, in degrees Celsius. Next, the temperature field is convolved using the Sobel operator to obtain the first-order partial derivative matrices of the temperature field in the horizontal and vertical directions. and The gradient magnitude is determined by the formula... The calculation is performed. Simultaneously, the Laplace operator is used to perform a second-order differential operation on the temperature field to obtain a second-order gradient distribution, which is particularly sensitive to inflection points and extreme points of temperature changes.

[0082] After obtaining the original gradient distribution, a multi-level temperature gradient representation is created by constructing a Gaussian pyramid to perform analysis at different scales. The pyramid construction here is an iterative dimensionality reduction process from high resolution to low resolution, implemented as follows:

[0083] First, the gradient magnitude data after maximum-minimum normalization is used as the 0th layer of the pyramid, denoted as . Secondly, construct the first level of the pyramid ( Gaussian smoothing is applied to the L0 image, i.e., using a Gaussian kernel with a specific standard deviation. (For example, Gaussian filter with =0.8) and A convolution operation is performed to obtain a blurred image. Then, this blurred image is downsampled, specifically by removing all even-numbered rows and columns, reducing its size to half its original width and height. The resulting image, after smoothing and downsampling, has a size of... A quarter of the image, which is the first level of the pyramid. Then, construct the second layer of the pyramid ( ). The first layer image As input, repeat the above process. That is, using a standard deviation with a Gaussian kernel. (For example, =1.6) Gaussian filter and Perform convolution, then downsample the result (removing even-numbered rows and columns). The resulting smaller image is the second layer of the pyramid. Finally, the third layer of the pyramid is constructed. (With the second layer image) For the input, repeat the process again, using a Gaussian kernel standard deviation. (For example, A Gaussian filter (=3.2) is used for convolution smoothing and downsampling to finally obtain the third layer L3 of the pyramid.

[0084] Finally, the three temperature gradient representations are weighted and fused. The weighting factor is assigned based on the rate of change of each gradient representation: if the rate of change of a certain gradient exceeds a preset gradient strength threshold (this threshold is usually set after statistical analysis of historical data from a large number of normally operating devices, for example, 0.6), then the feature of that layer is considered more significant, and it is given a higher weighting factor (e.g., ...). =0.5); otherwise, assign a lower weight (e.g., =0.3 or =0.2). Using the weighted summation formula... Ultimately, a second gradient intensity matrix, which integrates multi-scale information and highlights key temperature change regions, is obtained, namely, temperature gradient distribution data.

[0085] In step S13, spatial continuity compensation and noise suppression are performed on the temperature gradient distribution data. If the gradient amplitude exceeds a preset noise threshold, it is determined to be a valid temperature jump signal, and a first temperature jump signal is obtained, including:

[0086] Based on the temperature gradient distribution data, spatial continuity compensation is performed to obtain a gradient vector field distribution map;

[0087] Based on the gradient vector field distribution map, noise suppression processing is performed to obtain a set of denoised gradient data.

[0088] Based on the denoised gradient data set, threshold detection and clustering are performed to obtain the first temperature jump signal.

[0089] In one implementation, spatial continuity compensation is first performed on the second gradient intensity matrix. A bilinear interpolation algorithm is then used to fill in the blank areas between discrete data points in the matrix. Specifically, the algorithm estimates the value of the target point using the values ​​of its four nearest-neighbor known data points. The core idea is to first perform two independent linear interpolations in the first direction (e.g., horizontal) to obtain estimated values ​​for two intermediate points; then, based on the values ​​of these two intermediate points, perform a single linear interpolation in the second direction (e.g., vertical) to calculate the final value of the target point.

[0090] Next, to suppress random noise, noise detection is performed on the gradient vector field distribution map. This process is based on an adaptive Wiener filter, which dynamically adjusts the filter parameters by analyzing the local gradient change rate. For example, when the gradient fluctuation frequency in a certain region is higher than the background noise frequency, the system automatically reduces the filtering intensity in that region. The filter estimates the true signal by minimizing the mean square error between the signal and the noise, ultimately obtaining the denoised gradient data set.

[0091] Finally, the denoised gradient data set is subjected to threshold detection and clustering to obtain the first temperature jump signal. This process employs a dual judgment criterion. First, when the gradient magnitude of a data point exceeds a preset noise threshold (e.g., 5.2, obtained by statistically analyzing the gradient data of normally operating equipment in a fault-free state, taking the mean and adding three times the standard deviation), the system further checks the continuity of the signal. If the number of consecutive pixels exceeds the minimum signal width requirement (e.g., 6 pixel units), the data point is marked as a candidate temperature jump location. Then, a connected component labeling algorithm is used to cluster adjacent candidate jump locations. If the distance between adjacent candidate locations is less than a preset aggregation radius (e.g., 10 pixel units), they are merged into the same temperature jump event. The precise location coordinates of each jump event are determined using a centroid calculation method to obtain the first temperature jump signal.

[0092] In step S14, texture features of a visible light image are obtained from the temperature anomaly region, and similarity analysis is performed to obtain data on the deformation region of the device surface, including:

[0093] The first temperature jump signal is scanned to obtain a first candidate anomaly region by calculating a segmentation threshold and identifying spatial continuity features, including:

[0094] Perform a window sliding scan on the first temperature jump signal and calculate the segmentation threshold to obtain the segmentation threshold parameter;

[0095] Based on the segmentation threshold parameter, the spatial continuity features of the temperature gradient are identified, and a spatial continuity feature marker map is obtained;

[0096] Based on the spatially continuous feature map, region growing is performed to obtain the first candidate anomaly region.

[0097] In one implementation, the spatial coordinates of the first temperature jump signal are first acquired, and a fixed-size rectangular window is used to perform a row-by-row, column-by-column sliding scan of the data region. If the current window position exceeds the data boundary range (e.g., the data boundary is 512×384 pixels), the system automatically adjusts the window size to adapt to the boundary constraints, obtaining a complete region scanning path sequence. Based on the numerical distribution of temperature data within each window, the fluctuation amplitude of the temperature data within the window is analyzed using a statistical variance calculator. If the fluctuation amplitude is higher than a preset benchmark value (e.g., 6.0), the threshold adjustment factor value is increased using a preset linear adjustment formula, for example, new threshold adjustment factor = original threshold adjustment factor * (fluctuation amplitude / preset benchmark value), thereby obtaining the dynamically adjusted segmentation threshold parameter.

[0098] Next, based on the dynamically adjusted segmentation threshold parameters, a pixel adjacency detector is used to verify the continuity of temperature data points within each scanning window. Specifically, if the temperature difference between three adjacent pixels is less than the segmentation threshold and the spatial distance is within a preset neighborhood radius (e.g., 2 pixel units), the pixel combination is marked as a continuous temperature gradient feature, resulting in a spatially continuous feature label map.

[0099] Finally, based on the marker distribution pattern in the spatially continuous feature marker map, a region growing algorithm is used to aggregate pixels with the same marker attributes. The seed point selection rule for this algorithm is to randomly select an unaggregated pixel from the marker map as the initial seed point. Then, the algorithm searches for pixels with the same marker attributes in the eight adjacent directions of the seed point. If the number of aggregated pixels exceeds a preset minimum region area threshold (e.g., 300 pixels), the aggregated region is identified as a candidate abnormal temperature region, thus obtaining the first set of candidate abnormal regions.

[0100] In step S15, based on the first candidate anomaly region, the spatial clustering degree of the anomaly region is calculated and expanded to obtain the second candidate anomaly temperature region, including:

[0101] Temperature anomaly points are extracted from the first candidate anomaly region, and spatial location analysis is performed on the temperature anomaly points to obtain a regional cluster density distribution map.

[0102] Based on the regional aggregation density distribution map, the location of the density peak is found and marked as a seed candidate point, thus obtaining the set of seed candidate point coordinates;

[0103] Based on the set of seed candidate point coordinates, an expansion operation is performed to obtain the growth results of the abnormal region;

[0104] Based on the growth results of the abnormal region, regional tracking is performed to obtain the second candidate abnormal temperature region.

[0105] In one implementation, the pixel coordinate data of each region in the first candidate anomaly region set is first obtained. Then, spatial location analysis is performed on the temperature anomalies within the region. This process is achieved through the following steps: If the Euclidean distance between anomalies is less than a preset clustering radius (e.g., 5.0 pixels) and the number of anomalies exceeds a minimum clustering threshold (e.g., 12), then the kernel density estimation method is used to calculate the clustering density parameter value of the region, resulting in a regional clustering density distribution map. Specifically, kernel density estimation is a non-parametric estimation method used to estimate the probability density function of a random variable. Its calculation process involves calculating the density contribution value of each pixel within the region based on the number and distance of its surrounding temperature anomalies, ultimately generating a regional clustering density distribution map reflecting the spatial clustering degree of anomalies. Next, based on the density value distribution in the regional clustering density distribution map, by finding the density peak position in each candidate region, if the density value of the current pixel is greater than the density values ​​of all pixels within its eight neighboring regions, then the pixel is marked as a seed candidate point, resulting in a set of seed candidate point coordinates.

[0106] The expansion operation is implemented through the following steps: starting from the seed candidate point, an expansion operation is performed towards surrounding pixels. If the temperature difference between adjacent pixels in the 8-connected neighborhood is less than the growth threshold (e.g., 2.5℃), then it is assigned to the growth region of the current seed point. It should be noted that this growth region is a dynamically aggregated set of pixels, including the starting seed point and all adjacent pixels that meet the growth conditions. This process continues until no new pixels meet the growth conditions, ultimately yielding the abnormal region growth result. Finally, region tracking is achieved by performing single-pixel precision contour tracking on the region edges in the abnormal region growth result, thereby obtaining the second candidate abnormal temperature region data. Specifically, this process first identifies the starting point of the growth region edge, then traverses and records all pixels on the region boundary pixel by pixel in a preset direction (e.g., clockwise), obtaining a sequence containing the coordinates of all boundary pixels, thus achieving single-pixel precision contour extraction.

[0107] In step S16, the corresponding texture features are obtained from the second candidate abnormal temperature region, and the similarity between the texture features and the preset normal equipment surface texture is calculated. If the similarity is less than the preset similarity threshold, it is determined to be a deformation region, thus obtaining the equipment surface deformation region, including:

[0108] Obtain the corresponding visible light image region from the second candidate abnormal temperature region;

[0109] Texture features are extracted from the visible light image region to obtain the region texture features;

[0110] Calculate the similarity between the texture features of the region and the preset normal device surface texture. If the similarity is less than the preset similarity threshold, it is determined to be a deformed region, and the deformed region of the device surface is obtained.

[0111] First, based on the image registration mapping relationship established in step S11, the system maps the pixel coordinate range of the second candidate abnormal temperature region to the corresponding visible light image, thereby accurately obtaining the visible light image region to be analyzed.

[0112] Texture features refer to the set of statistical texture feature parameters calculated using Local Binary Pattern (LBP) descriptors and Gray-Level Co-occurrence Matrix (GLCM). These features are used to quantify the local microscopic patterns and global statistical properties of the surface texture of a device. In practice, the region to be analyzed in the visible light image is first encoded using Local Binary Pattern (LBP). This process compares the gray value of each pixel with the gray values ​​of its eight neighboring pixels. If the gray value of a neighboring pixel is greater than or equal to that of the center pixel, it is marked as 1; otherwise, it is marked as 0. This generates a binary encoded sequence as the local texture descriptor for that pixel. Next, the Gray-Level Co-occurrence Matrix (GLCM) is constructed. This matrix describes the global characteristics of the texture by statistically analyzing the frequency of pixel pairs with a specific spatial relationship (e.g., a distance of 1 pixel unit and an orientation of 0 degrees). Specifically, the value of each element (i, j) in the matrix represents the number of times pixel pairs with gray levels i and j in the image appear with a specific spatial relationship. Based on this matrix, multiple texture statistical parameters such as texture contrast, entropy, second moment of angle, and correlation are calculated, which together form a set of regional texture statistical feature parameters.

[0113] To determine whether a region is a deformed area, the system pre-establishes a normal device surface texture feature library. This library is a set of standard texture feature vectors extracted and stored using the same method when the device is in a known normal state. Then, a similarity metric is used to compare the set of texture statistical feature parameters of the current region with all the standard feature vectors in the feature library, calculating the Pearson correlation coefficient between the two as a quantitative indicator of similarity. If all calculated correlation coefficients are less than a preset similarity threshold (e.g., 0.65), it indicates that the texture features of the current region deviate significantly from the normal state. Based on this, the system determines that the region is an abnormal surface deformation and outputs the location data of the deformed area on the device surface.

[0114] In step S17, the correlation coefficient between the deformation area on the device surface and the second candidate abnormal temperature area is calculated. If the correlation coefficient is greater than a preset correlation threshold and the abnormal temperature area is greater than a preset lower limit of area, then the device is determined to have a fault risk, and the fault detection result is output, including:

[0115] Calculate the spatial overlap between the deformation area on the equipment surface and the second candidate abnormal temperature area to obtain the correlation coefficient;

[0116] The correlation coefficient is compared with a preset association threshold. If the correlation coefficient is greater than the preset association threshold, it is determined that there is a causal relationship between the temperature anomaly and the deformation, and the boundary data of the anomaly area is obtained.

[0117] Extract the total pixel area from the boundary data of the abnormal region. If the total pixel area is greater than a preset lower limit value, it is determined that the device has a fault risk, and the fault detection result is output.

[0118] First, the system acquires the location data of the deformation area on the equipment surface and the data of the second candidate abnormal temperature area, and calculates the spatial overlap between the two. In practice, this process is based on the geometric analysis of the boundary coordinates of the regions, and quantifies their spatial correlation by calculating the ratio of the intersection and union areas of the two regions.

[0119] The correlation coefficient is calculated using the Pearson correlation algorithm, which is based on statistical linear correlation analysis. In practice, the system extracts spatial distribution feature vectors of the deformed and temperature anomaly regions, including parameters such as region center coordinates, aspect ratio, and boundary regularity. A quantitative correlation coefficient is then calculated using the Pearson correlation formula.

[0120] Next, the correlation coefficient is compared with a preset correlation threshold (for example, the threshold is 0.75). If the correlation coefficient is greater than the threshold, the system determines that there is a strong causal relationship between the temperature anomaly and the surface deformation, confirms that the two are composite anomalies caused by the same fault source, and merges the boundaries of the two causally related anomaly regions to obtain the final anomaly region boundary data.

[0121] Finally, the system uses a pixel counting function to calculate the total pixel area of ​​the abnormal region. When processing an abnormal region with a boundary range of [195, 145] to [270, 205], the counter scans all pixels within this rectangular range line by line, counting the number of valid pixels belonging to the abnormal region. After a complete traversal, if the total pixel area (e.g., 4125 pixels) is greater than a preset lower limit (e.g., 3000 pixels), the system determines that the device has a fault risk and outputs the fault detection result and spatial location coordinate calibration information.

[0122] It should be noted that causality is defined as a multi-stage judgment rule. The first stage involves calculating the spatial overlap between temperature anomaly region A and deformation region B, which can be achieved using the Jaccard coefficient, i.e. .like If the overlap exceeds a preset spatial overlap threshold T_overlap (e.g., T_overlap=0.5), the process proceeds to the second stage. In the second stage, spatial distribution feature vectors (e.g., vectors containing centroid coordinates, aspect ratio, and boundary regularity) are extracted from the two regions, and the Pearson correlation coefficient r between the two vectors is calculated. It should be noted that boundary regularity refers to the smoothness or regularity of the region boundary, calculated as the ratio of the region's perimeter to the perimeter of a circle of the same area; a larger ratio indicates a more irregular boundary. If r exceeds a preset feature correlation threshold T_corr (e.g., T_corr=0.75), a strong causal relationship is determined between the temperature anomaly and surface deformation.

[0123] In summary, this invention achieves accurate fault warning and spatial positioning of high-temperature equipment by fusing analysis of temperature jumps in infrared images and surface deformation in visible light images, and establishing a causal relationship judgment mechanism between the two.

[0124] Reference Figure 2 The second embodiment of the present invention provides a fault detection device based on a high-temperature resistant camera, comprising:

[0125] The data acquisition and registration module is used to acquire infrared and visible light images of the surface of high-temperature equipment, and perform registration processing to obtain fused image data;

[0126] The gradient analysis module is used to perform temperature field gradient analysis based on the fused image data to obtain temperature gradient distribution data.

[0127] The signal recognition module is used to perform spatial continuity compensation and noise suppression on the temperature gradient distribution data. If the gradient amplitude exceeds a preset noise threshold, it is determined to be an effective temperature jump signal, and the first temperature jump signal is obtained.

[0128] An abnormal region identification module is used to perform region scanning on the first temperature jump signal, calculate the segmentation threshold and identify spatial continuity features to obtain the first candidate abnormal region.

[0129] An abnormal region location module is used to calculate the spatial clustering degree of the abnormal region based on the first candidate abnormal region and perform an expansion operation to obtain a second candidate abnormal temperature region.

[0130] The deformation region identification module is used to obtain the corresponding texture features from the second candidate abnormal temperature region, calculate the similarity between the texture features and the preset normal equipment surface texture, and if the similarity is less than the preset similarity threshold, it is determined to be a deformation region, and the deformation region of the equipment surface is obtained.

[0131] The fault risk assessment module is used to calculate the correlation coefficient between the deformation area on the surface of the equipment and the second candidate abnormal temperature area. If the correlation coefficient is greater than a preset correlation threshold and the abnormal temperature area is greater than a preset area lower limit, then the equipment is determined to have a fault risk, and the fault detection result is output.

[0132] It should be noted that the fault detection device based on a high-temperature camera provided in this embodiment of the invention is used to execute all the process steps of the fault detection method based on a high-temperature camera in the above embodiment. The working principle and beneficial effects of the two are one-to-one, so they will not be described again.

[0133] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a fault detection program for a high-temperature camera. When the processor executes the computer program, it implements the steps in the various fault detection method embodiments based on high-temperature cameras described above, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the signal recognition module.

[0134] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0135] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0136] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting various parts of the electronic device through various interfaces and lines.

[0137] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD card), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0138] If the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0139] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0140] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A fault detection method based on a high-temperature resistant camera, characterized in that, include: Infrared and visible light images of the surface of high-temperature equipment are acquired and registered to obtain fused image data. Based on the fused image data, temperature field gradient analysis is performed to obtain temperature gradient distribution data. Spatial continuity compensation and noise suppression are performed on the temperature gradient distribution data. If the gradient amplitude exceeds a preset noise threshold, it is determined to be an effective temperature jump signal, and the first temperature jump signal is obtained. The first temperature jump signal is scanned in a region, the segmentation threshold is calculated and spatial continuity features are identified to obtain the first candidate abnormal region. Based on the first candidate anomaly region, the spatial clustering degree of the anomaly region is calculated and an expansion operation is performed to obtain the second candidate anomaly temperature region. Obtain the corresponding texture features from the second candidate abnormal temperature region, calculate the similarity between the texture features and the preset normal equipment surface texture, and if the similarity is less than the preset similarity threshold, it is determined to be a deformation region, and the equipment surface deformation region is obtained. Calculate the correlation coefficient between the deformation area on the surface of the equipment and the second candidate abnormal temperature area. If the correlation coefficient is greater than a preset correlation threshold and the abnormal temperature area is greater than a preset area lower limit, then it is determined that the equipment has a fault risk, and the fault detection result is output. The step of calculating the spatial clustering degree of the anomalous region and performing an expansion operation based on the first candidate anomalous region to obtain the second candidate anomalous temperature region includes: Temperature anomaly points are extracted from the first candidate anomaly region, and spatial location analysis is performed on the temperature anomaly points to obtain a regional cluster density distribution map. Based on the regional aggregation density distribution map, the location of the density peak is found and marked as a seed candidate point, thus obtaining the set of seed candidate point coordinates; Based on the set of seed candidate point coordinates, an expansion operation is performed to obtain the growth results of the abnormal region; Based on the growth results of the abnormal region, regional tracking is performed to obtain the second candidate abnormal temperature region.

2. The fault detection method based on a high-temperature resistant camera according to claim 1, characterized in that, The process of acquiring infrared and visible light images of the surface of the high-temperature equipment and performing registration processing to obtain fused image data includes: Obtain infrared radiation information from the surface of high-temperature equipment; Based on the infrared radiation information, image acquisition is performed to obtain infrared images and visible light images; Based on the infrared and visible light images, correction and registration are performed to obtain fused image data.

3. The fault detection method based on a high-temperature resistant camera according to claim 1, characterized in that, The step of performing temperature field gradient analysis based on the fused image data to obtain temperature gradient distribution data includes: Temperature information is extracted from the fused image data, and a convolution operation is performed to obtain the first-order and second-order gradient distributions of the temperature field. Based on the first-order and second-order gradient distributions, normalization and convolution operations are performed to obtain a multi-level temperature gradient representation. Based on the multi-level temperature gradient, weighted fusion is performed to obtain temperature gradient distribution data.

4. The fault detection method based on a high-temperature resistant camera according to claim 1, characterized in that, The process of performing spatial continuity compensation and noise suppression on the temperature gradient distribution data, wherein if the gradient amplitude exceeds a preset noise threshold, it is determined to be a valid temperature jump signal, and a first temperature jump signal is obtained, includes: Based on the temperature gradient distribution data, spatial continuity compensation is performed to obtain a gradient vector field distribution map; Based on the gradient vector field distribution map, noise suppression processing is performed to obtain a set of denoised gradient data. Based on the denoised gradient data set, threshold detection and clustering are performed to obtain the first temperature jump signal.

5. The fault detection method based on a high-temperature resistant camera according to claim 1, characterized in that, The step of performing a region scan on the first temperature jump signal, calculating a segmentation threshold, and identifying spatially continuous features to obtain a first candidate anomaly region includes: Perform a window sliding scan on the first temperature jump signal and calculate the segmentation threshold to obtain the segmentation threshold parameter; Based on the segmentation threshold parameter, the spatial continuity features of the temperature gradient are identified, and a spatial continuity feature marker map is obtained; Based on the spatially continuous feature map, region growing is performed to obtain the first candidate anomaly region.

6. The fault detection method based on a high-temperature resistant camera according to claim 1, characterized in that, The process involves obtaining corresponding texture features from the second candidate abnormal temperature region, calculating the similarity between the texture features and a preset normal equipment surface texture, and determining the deformation region as a deformation region of the equipment surface if the similarity is less than a preset similarity threshold. Obtain the corresponding visible light image region from the second candidate abnormal temperature region; Texture features are extracted from the visible light image region to obtain the region texture features; Calculate the similarity between the texture features of the region and the preset normal device surface texture. If the similarity is less than the preset similarity threshold, it is determined to be a deformed region, and the deformed region of the device surface is obtained.

7. The fault detection method based on a high-temperature resistant camera according to claim 1, characterized in that, The correlation coefficient between the deformation area on the equipment surface and the second candidate abnormal temperature area is calculated. If the correlation coefficient is greater than a preset correlation threshold and the abnormal temperature area is greater than a preset lower limit of area, then the equipment is determined to have a fault risk, and the fault detection result is output, including: Calculate the spatial overlap between the deformation area on the equipment surface and the second candidate abnormal temperature area to obtain the correlation coefficient; The correlation coefficient is compared with a preset association threshold. If the correlation coefficient is greater than the preset association threshold, it is determined that there is a causal relationship between the temperature anomaly and the deformation, and the boundary data of the anomaly area is obtained. Extract the total pixel area from the boundary data of the abnormal region. If the total pixel area is greater than a preset lower limit value, it is determined that the device has a fault risk, and the fault detection result is output.

8. A fault detection device based on a high-temperature resistant camera, characterized in that, The method for implementing the fault detection method based on a high-temperature resistant camera as described in any one of claims 1 to 7 includes: The data acquisition and registration module is used to acquire infrared and visible light images of the surface of high-temperature equipment, and perform registration processing to obtain fused image data; The gradient analysis module is used to perform temperature field gradient analysis based on the fused image data to obtain temperature gradient distribution data. The signal recognition module is used to perform spatial continuity compensation and noise suppression on the temperature gradient distribution data. If the gradient amplitude exceeds a preset noise threshold, it is determined to be an effective temperature jump signal, and the first temperature jump signal is obtained. An abnormal region identification module is used to perform region scanning on the first temperature jump signal, calculate the segmentation threshold and identify spatial continuity features to obtain the first candidate abnormal region. An abnormal region location module is used to calculate the spatial clustering degree of the abnormal region based on the first candidate abnormal region and perform an expansion operation to obtain a second candidate abnormal temperature region. The deformation region identification module is used to obtain the corresponding texture features from the second candidate abnormal temperature region, calculate the similarity between the texture features and the preset normal equipment surface texture, and if the similarity is less than the preset similarity threshold, it is determined to be a deformation region, and the deformation region of the equipment surface is obtained. The fault risk assessment module is used to calculate the correlation coefficient between the deformation area on the surface of the equipment and the second candidate abnormal temperature area. If the correlation coefficient is greater than a preset correlation threshold and the abnormal temperature area is greater than a preset area lower limit, then the equipment is determined to have a fault risk, and the fault detection result is output.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform a fault detection method based on a high-temperature resistant camera as described in any one of claims 1 to 7.

Images