A method and system for detecting microcracks in fire-resistant electrical wires

Abstract

This invention belongs to the field of texture feature analysis technology, specifically relating to a method and system for detecting microcracks in fire-resistant electrical wires. The method includes: constructing filter banks with different directions and scales; performing two-dimensional convolution between each filter and a pixel to obtain the response modulus of each pixel at the corresponding direction and scale; using the maximum value of all response moduli of a pixel as its directional comprehensive response value, and calculating the mean and standard deviation of the pixel's directional response accordingly; combining the maximum directional comprehensive response value of the pixel to obtain the crack significance index of each pixel; then obtaining a binarized image using the maximum inter-class variance method, performing morphological operations and connected component extraction to obtain the total defect area; calculating the comprehensive defect degree based on the defect area ratio and the average defect area, and comparing it with an anomaly threshold to obtain the detection result of the microcracks. This invention can improve the accuracy of detecting microcracks in fire-resistant electrical wires.

Description

Technical Field

[0001] This invention relates to the field of texture feature analysis technology, and in particular to a method and system for detecting microcracks in fire-resistant electrical wires. Background Technology

[0002] Fire-resistant cables are cables that can maintain normal operation for a certain period of time even in a burning environment. During the production of fire-resistant cables, such as the extrusion of the insulation and fire-resistant layers, improper control of process parameters can easily lead to stress concentration inside the cable, resulting in microcracks. In the event of a fire, flames can spread rapidly through these microcracks, preventing the fire-resistant cable from maintaining power within the specified time and creating a safety hazard. Existing technologies commonly use Gabor filter texture analysis algorithms to detect microcracks in fire-resistant cables. Specifically, this involves constructing a multi-directional, multi-scale filter bank to filter the image of the fire-resistant cable and generating a response map to enhance the linear features of the microcracks. Then, a fixed threshold is used for binarization segmentation to extract the microcrack region from the image background. However, existing technologies have serious technical flaws in testing the bending and other performance of finished fire-resistant wires before they leave the factory. The core problem is that after obtaining the response map through Gabor filtering, existing technologies heavily rely on a fixed global threshold for segmentation to extract microcracks. However, in reality, uneven illumination inevitably occurs when industrial light sources shine on the curved surface of the wire. This causes significant differences in the filtered response values ​​of the same continuous microcrack at different locations: the response is extremely strong when the microcrack is located in a high-illuminance area, but the response is significantly weakened when it is located in a low-illuminance area. Therefore, using a fixed threshold after segmentation can easily cause the originally continuous microcrack to break into multiple discontinuous segments, resulting in inaccurate detection results for microcracks. Summary of the Invention

[0003] To address the problem that existing technologies rely heavily on a fixed global threshold for segmentation to extract microcracks after obtaining the response map through Gabor filtering, which leads to the fragmentation of originally continuous microcracks into multiple discontinuous segments and thus inaccurate microcrack detection results, this invention provides a method and system for detecting microcracks in fire-resistant electrical wires.

[0004] In a first aspect, the present invention provides a method for detecting microcracks in fire-resistant electrical wires, comprising: obtaining a surface image of the fire-resistant electrical wire through image acquisition and preprocessing; constructing Gabor filter banks of different directions and scales; performing two-dimensional convolution operations between each Gabor filter in the Gabor filter bank and each pixel in the surface image to obtain the response modulus value of each pixel in the corresponding direction and scale; obtaining the directional comprehensive response value of the pixel in any direction based on the maximum value of the response modulus value of the pixel in all scales; and calculating the square root of the pixel based on the directional comprehensive response value of the pixel in all directions. The crack significance index of a pixel is obtained by multiplying the mean and standard deviation of the directional response by the power of the ratio of the standard deviation to the mean of the directional response, where the power is a constant greater than 1. Based on the crack significance index of each pixel, a segmentation threshold is adaptively determined using the maximum inter-class variance method to obtain a binary image containing only defective pixels. Morphological operations and connected component extraction are performed on the binary image to obtain several total defect regions. The comprehensive defect degree is obtained based on the proportion of the total defect area and the average defect area. The detection result of microcracks in fire-resistant wires is obtained based on the comprehensive defect degree and the anomaly threshold.

[0005] This invention significantly improves the accuracy and practicality of microcrack detection in fire-resistant electrical wires through a scientific image processing and quantitative evaluation process. By utilizing Gabor filter banks of different directions and scales and two-dimensional convolution operations, and combining directional comprehensive response values, directional response mean, and directional response standard deviation to construct a crack significance index, it accurately captures the direction and size characteristics of microcracks, effectively distinguishing them from background noise and reducing the probability of missed detections and misjudgments. The maximum inter-class variance method is used to adaptively determine the segmentation threshold, combined with morphological operations and connected region extraction, ensuring the integrity and authenticity of the total defect area in the binarized image. The comprehensive defect degree is quantified based on the defect area ratio and average defect area, taking into account both the overall coverage of microcracks and the intensity of local defects, effectively distinguishing cracks of different risk types, and the evaluation results are consistent with industrial realities. The detection process is highly automated, requiring minimal manual intervention, and key parameters can be flexibly adjusted to adapt to batch detection scenarios, providing efficient, stable, and accurate technical support for the quality control of fire-resistant electrical wires.

[0006] Preferably, constructing Gabor filter banks with different directions and scales includes: setting the number of direction types to 13, i.e., constructing 13 Gabor filters with different directions; and recording the direction numbers as follows: The range of values ​​is Integers between 0° and 0°, 13 directions starting from 0°, with each pair of adjacent directions spaced apart. The number of scale types is set to 7, meaning 7 Gabor filters with different scales are constructed. The scale numbers are denoted as follows: The range of values ​​is Integers between [a certain range].

[0007] This method, by designing 13 different directions, can densely cover various directional textures that may exist in the image. This avoids the problem of feature omission caused by excessively large directional sampling intervals, and ensures that it can effectively capture fine textures or macroscopic structures in different orientations such as horizontal, tilted, and vertical. It is especially suitable for image scenes with complex textures and diverse orientations. The design of 7 different scales can extract information at different scales in the image, from small textures to large structures, to meet the feature acquisition requirements under different accuracy requirements. The clear direction and scale numbers not only provide clear identification for the parameter management and call control of the filter bank, but also facilitate subsequent feature tracing and parameter optimization.

[0008] Preferably, the Gabor filter banks of different scales further include: setting a first... The scale factor corresponding to each scale is: Its value is calculated according to the following formula: , It is the scale number; setting the number The wavelength at each scale is 2.5 times the scale factor; the bandwidth is set to 0.4 and the phase to 0, thus obtaining wavelengths in different directions. and different scales The Gabor filter below, denoted as .

[0009] This method first sets 13 different directions, then clarifies the core parameters for 7 different scales, and finally... Scale factor of each scale According to the relation Calculations were performed, setting the wavelength to 2.5 times the corresponding scale factor, fixing the bandwidth to 0.4 and the phase to 0, to construct a system covering different directions. and different scales The system utilizes a Gabor filter bank and performs multi-dimensional and refined filtering on the target image based on this filter bank. The system captures texture details and structural features of different orientations and sizes in the image, providing accurate and comprehensive feature support for subsequent tasks such as image classification, target recognition, and texture analysis.

[0010] Preferably, the step of performing a two-dimensional convolution operation between each Gabor filter in the Gabor filter bank and each pixel in the surface image to obtain the response modulus of each pixel in the corresponding direction and at the corresponding scale includes: converting the pixel... With in different directions and different scales Gabor filter below Perform 2D convolution, and use mirror padding at the edges to avoid boundary effects to obtain pixel points. In the Direction, First The response modulus of the scale is denoted as .

[0011] Preferably, obtaining the directional comprehensive response value of a pixel in any direction based on the maximum value of the response modulus corresponding to all scales in any direction includes: [The text abruptly ends here, so the translation stops as well.] In the The corresponding directional integrated response value under the direction is denoted as ,in, , ,and It is an integer.

[0012] Preferably, obtaining the crack saliency index of a pixel includes: ;in, It is a pixel. The crack significance index; It is a pixel. In the The overall directional response value in each direction. ; This indicates that the maximum value of the pixel's overall response value across all directions is taken; It is a pixel. Maximum directional composite response value in all directions; It is a pixel. The standard deviation of the directional response; It is a pixel. The mean directional response; It is a constant to prevent the denominator from being zero; The preset exponent is a constant with a value greater than 1.

[0013] This method uses the maximum directional composite response value of a pixel across all directions as a representation of crack texture intensity. It then multiplies this value by a preset power of the ratio of the standard deviation to the mean of the directional response. This power greater than 1 nonlinearly amplifies the ratio of the standard deviation to the mean of the directional response, significantly enhancing the high ratio characteristic of microcrack regions due to their strong directional specificity. Simultaneously, it effectively suppresses the low ratio characteristic of background regions due to their uniform directional response. Thus, while maintaining the sensitivity of the maximum directional composite response value to the crack's main strength, it further amplifies the inherent differences in the directional dimension between crack pixels and background pixels. This construction method makes the image of a microcrack... When the response of a pixel is significantly different in each direction, it obtains a higher significance index. However, even when the comprehensive response value of a pixel belonging to the background texture or illumination spot fluctuates, the ratio of its standard deviation of directional response to its mean directional response remains small. After exponential operation, it approaches zero and cannot obtain effective gain. Therefore, the crack significance index can more robustly distinguish between real microcracks and various pseudo-defects, greatly reducing false detections and false negatives caused by uneven illumination, messy surface textures, or random noise. It provides a high-contrast, high-signal-to-noise ratio feature map for subsequent adaptive threshold segmentation and complete extraction of crack regions, improving the accuracy of microcrack detection in fire-resistant wires and its stability in industrial scenarios.

[0014] Preferably, the step of adaptively determining the segmentation threshold based on the crack significance index of each pixel using the maximum inter-class variance method to obtain a binarized image containing only defective pixels includes: if the crack significance index of a pixel is greater than or equal to the segmentation threshold, then the pixel is determined to be a defective pixel; if the crack significance index of a pixel is less than the segmentation threshold, then the pixel is determined to be a background pixel, thereby obtaining a binarized image containing only defective pixels, wherein the area of ​​defective pixels is white and the area of ​​background pixels is black.

[0015] Preferably, obtaining the comprehensive defect level based on the proportion of the total defect area and the average defect area includes: ;in, It represents the overall defect level of the fire-resistant wire corresponding to the surface image; It is the calibration coefficient; It is the percentage of the total defective area; It is the average area of ​​the defect region; It is the preset critical single defect area.

[0016] This method first identifies and segments defect regions in the surface image of fire-resistant wires based on the crack significance index of pixels, then calculates the total defect area ratio and the average defect area. Next, it introduces a calibration coefficient and a preset critical single defect area, and then applies the formula... A comprehensive calculation is performed to obtain a complete defect level that fully characterizes the surface condition of fire-resistant wires. This process reflects the overall coverage of defects by both the area ratio of the total defective region and the overall defect coverage by... This method quantifies the deviation of the average size of a single defect from the critical standard, and, combined with the calibration coefficient f, ensures the scenario adaptability of the results. It achieves dual-dimensional quantification of the surface defect coverage and the size of a single defect in fire-resistant wires, providing accurate and comprehensive quantitative basis for subsequent tasks such as quality grading and qualification determination.

[0017] Preferably, obtaining the detection result of microcracks in fire-resistant wires based on the comprehensive defect degree and anomaly threshold includes: setting a first anomaly threshold. Second abnormal threshold It needs to meet the following requirements. When the overall defect level Conditions met: At this point, the fire-resistant wire is judged to be in a qualified state, with a small coverage area of ​​microcracks and no large-sized cracks; when the overall defect degree is... Conditions met: If the fire-resistant wire is determined to have a minor defect, it needs to be returned to the factory for repair. If the overall defect severity is... Conditions met: If the fire-resistant wire is determined to be in a severely defective state, then the batch of wire must be prohibited from leaving the factory or being used.

[0018] Secondly, the present invention provides a microcrack detection system for fire-resistant electrical wires, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned method for detecting microcracks in fire-resistant electrical wires is implemented.

[0019] By adopting the above technical solution, a computer program for detecting microcracks in fire-resistant wires is generated and stored in a memory, so that it can be loaded and executed by a processor. A terminal device can then be made based on the memory and processor for convenient use.

[0020] The beneficial effects of this invention are as follows: Through a systematic image processing and feature extraction process, the accuracy of microcrack detection in fire-resistant wires is significantly improved, while simultaneously ensuring detection stability and industrial adaptability. A Gabor filter bank with different directions and scales is constructed to precisely adapt to the directionality and size differences of microcracks. Each Gabor filter is convolved with each pixel of the surface image in two dimensions, enhancing the response modulus of microcrack pixels at the corresponding direction and scale, while weakening background noise interference such as wire surface texture and light reflection. The directional comprehensive response value of the pixel is then determined by the maximum value of the response modulus across all scales in all directions. The crack significance index is calculated by combining the mean of the directional response, the standard deviation of the directional response, and the maximum directional comprehensive response value across all directions, highlighting the strong response characteristics of crack pixels and achieving pixel-level precise differentiation between cracks and background pixels, thus enabling the detection of even extremely fine cracks. The response signal is more prominent, significantly reducing the false negative rate and the probability of misjudgment due to background noise. The maximum inter-class variance method is used to adaptively determine the segmentation threshold based on the crack significance index, avoiding the problem of insufficient adaptation of fixed thresholds, ensuring that the binarized image accurately retains the real defect pixels, and further improving the accuracy of crack area recognition. Morphological operations and connected component extraction are used to post-process the binarized image to connect the crack fracture, filter isolated noise points and false response areas, and ensure the continuity, integrity and authenticity of the total defect area, laying the foundation for quantitative evaluation. The comprehensive defect degree is calculated based on the defect area ratio and the average defect area, which not only reflects the overall coverage of microcracks, but also indirectly quantifies the size and intensity of individual cracks and the density of defects, avoiding the complexity and error accumulation of multi-dimensional parameter measurement, further ensuring the accuracy of detection, and providing efficient and reliable technical support for the quality control of fire-resistant wires. Attached Figure Description

[0021] Figure 1 This is a schematic flowchart illustrating a method for detecting microcracks in fire-resistant electrical wires according to the present invention. Figure 2 This is a schematic diagram illustrating the crack significance index of a microcrack detection method for fire-resistant electrical wires according to the present invention. Figure 3 This is a schematic diagram illustrating the morphological processing of a method for detecting microcracks in fire-resistant electrical wires according to the present invention. Figure 4 This is a schematic diagram illustrating the final defect detection result of a microcrack detection method for fire-resistant electrical wires according to the present invention. Detailed Implementation

[0022] 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, not all, of the embodiments of the present invention. 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.

[0023] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0024] This invention discloses a method for detecting microcracks in fire-resistant electrical wires, referring to... Figure 1 This includes steps S1 to S4: S1. Obtain the surface image of the preprocessed fire-resistant wire through image acquisition and preprocessing.

[0025] It should be noted that the bending test platform for fire-resistant wires was chosen as the data collection point. The reason is that the bending test will cause stress concentration inside the wire, and potential microcracks will appear or expand due to stress release. For example, the crack edge opening will increase, making it easier to capture compared to the static state. At the same time, the curvature change of the wire surface is more significant under bending conditions, which can better expose the impact of uneven lighting on subsequent testing and provide a basis for subsequent targeted treatment.

[0026] Specifically, a high-precision industrial camera is used to photograph the surface of the fire-resistant wire to ensure that the details of the microcracks on the surface of the fire-resistant wire can be clearly shown; the optical axis of the camera is perpendicular to the plane where the axis of the fire-resistant wire is located, that is, the outward convex surface of the wire after bending is photographed perpendicularly; a flicker-free high-brightness light source is used to obtain an RGB color image of the fire-resistant wire.

[0027] Furthermore, the surface images of the acquired fire-resistant wires are preprocessed: the RGB color images are converted to grayscale images using a weighted average formula; Gaussian filtering is then applied to the grayscale images to suppress random noise from the sensor, wherein the Gaussian filtering employs... A Gaussian convolution kernel of a specific size; processing of the grayscale image using a contrast-limited adaptive histogram equalization algorithm, including: segmenting the grayscale image into... The non-overlapping sub-blocks are used to calculate the gray-level histogram for each sub-block. By limiting the contrast gain, usually set to 2.0, the gray-level distribution of the image is made more uniform, and the visibility of micro-cracks in dark areas is enhanced.

[0028] Specifically, a surface image with reasonable grayscale distribution, low noise, and prominent edges and shadows of microcracks is obtained, laying a reliable foundation for subsequent Gabor filtering to extract crack texture. Among these, grayscale processing, Gaussian filtering, and contrast-limited adaptive histogram equalization are all existing mature image processing techniques, which will not be elaborated here.

[0029] S2. Construct Gabor filter banks with different directions and scales. Through different filters, obtain the response magnitude of each pixel in the corresponding direction and scale. Based on the response magnitude of the pixel, obtain the comprehensive response value in that direction. Based on the comprehensive response values ​​of all directions of the pixel, obtain its mean and standard deviation of directional response. Combine the maximum response magnitude of the pixel to obtain its crack significance index.

[0030] It should be noted that microcracks on the surface of fire-resistant wires exhibit both random orientation and diverse width characteristics under bending stress. Filters relying solely on a single scale or direction cannot fully capture these characteristics. Therefore, by simulating the human visual system's perception of fine textures, a multi-scale, multi-directional Gabor filter bank is constructed. Convolution is used to obtain the directional response matrix of each pixel at multiple scales. Utilizing the difference between the strong response of microcracks in one direction and the weak response in other directions, and the uniform response of the background in all directions, a crack saliency index for each pixel is constructed. Power terms are used to compensate for response attenuation caused by uneven illumination, ensuring that pixels belonging to microcracks in both high- and low-light areas maintain high saliency, thus supporting subsequent segmentation.

[0031] Specifically, a Gabor filter bank containing different directions and scales is constructed: for example, the number of direction types is set to 13, that is, 13 Gabor filters with different directions are constructed; the direction numbers are denoted as... The range of values ​​is Integers between 0° and 0°, 13 directions starting from 0°, with each pair of adjacent directions spaced apart. Corresponding to 0°, 15°, 30°, ..., 180° respectively, providing complete coverage. The range should be set to ensure that there are no blind spots in the response between adjacent directions, and to avoid missing some microcracks. It should be noted that the microcrack direction detection should not exceed 180° to avoid directional redundancy.

[0032] Furthermore, for example, the number of scale types is set to 7, that is, 7 Gabor filters of different scales are constructed, and the scale numbers are denoted as follows: The range of values ​​is Integers between 1 and 2; the scaling factor is a core parameter of the Gabor filter, essentially the standard deviation of the Gaussian envelope, used to define the effective detection range of the filter. Each scale corresponds to a scaling factor. For example, the 1st... The scale factor corresponding to each scale is denoted as . Its value is calculated according to the following formula: This design allows the scale factor corresponding to each scale to increase uniformly with the sequence number, ensuring that the detection ranges of the seven Gabor filters at different scales are closely connected; among them, 0.8 is an example value and can be modified according to actual needs.

[0033] Specifically, because the Gabor filter follows Based on the coverage principle, 99.7% of the effective detection energy is concentrated in the core region. Based on this principle, the width of the microcrack is twice the scale factor, ensuring that microcracks of the corresponding width are completely embedded in the effective detection area of ​​the filter, enabling the filter to generate the strongest response to microcracks of that width and achieve accurate identification. Furthermore, a third... The wavelength at each scale is 2.5 times the scale factor, aiming to match the width of the microcrack corresponding to that scale, so that the spatial frequency of the microcrack texture resonates with the filter carrier frequency, thereby enhancing the response intensity. For example, a bandwidth of 0.4 and a phase of 0 are set to simplify calculations and enhance the contrast between light and dark areas at the microcrack edges; thus obtaining different wavelengths in different directions. and different scales The Gabor filter below, denoted as .

[0034] It should be noted that in order to eliminate the detection uncertainty caused by the arbitrariness of crack direction and to solve the problem of insufficient contrast of a single response map under uneven illumination, it is necessary to extract feature indicators with illumination robustness from the multidimensional response data, and to analyze the response modulus obtained by convolution.

[0035] Specifically, using pixels in the surface image Let's take a case study: [Analysis of pixel points] With in different directions and different scales Gabor filter below Perform 2D convolution, and use mirror padding at the edges to avoid boundary effects to obtain pixel points. In the Direction, First The response modulus of the scale is denoted as ; and thus obtain all pixels in the surface image at the th Direction, First The response modulus of the scale.

[0036] Furthermore, based on pixels Calculate the directional comprehensive response value of a pixel in any given direction by considering the response magnitudes at all scales. Given that microcracks produce the strongest response at the optimal scale matching their width, the maximum response magnitude across all scales in that direction can be used as the directional comprehensive response value. For example, the pixel... In the The corresponding directional integrated response value under the direction is denoted as ,in, , ,and The value is an integer; at this time, each direction corresponds to only one direction comprehensive response value, while one pixel corresponds to multiple directions, that is, multiple direction comprehensive response values; this processing not only preserves the strongest response characteristics of the microcrack in that direction, but also achieves dimensionality reduction of the scale dimension, laying the foundation for subsequent analysis of response differences between different directions.

[0037] It should be noted that microcrack regions typically exhibit extremely strong response values ​​in a specific direction, corresponding to the direction of the microcrack, while the response is weaker in the direction perpendicular to it, demonstrating significant anisotropy. Conversely, background regions typically exhibit isotropy. It is important to note that background regions refer to all regions remaining in the surface image except for the microcrack regions, including the illuminated areas. Based on this characteristic, a crack saliency index is constructed for each pixel to address the technical problem of crack fracture caused by uneven illumination.

[0038] Furthermore, the pixels in the surface image The average of the combined response values ​​in all directions is used as the pixel value. directional response mean Used to reflect pixels Average response intensity in all directions; pixels in the surface image The standard deviation of the overall response values ​​in all directions is used as the pixel value. Directional response standard deviation Used to reflect pixels The degree of difference in response in different directions; among them, the pixels belonging to the microcrack region have a significantly higher comprehensive response value in a certain direction, resulting in a less uniform distribution of their directional response mean, while the pixels belonging to the background region have smaller differences in comprehensive response values ​​in all directions, so their directional response mean distribution is more uniform; the pixels belonging to the microcrack region have a larger standard deviation of directional response due to the significant contrast between strong and weak response directions, while the pixels belonging to the background region have a relatively smooth response in all directions, and their directional response standard deviation is smaller.

[0039] It should be noted that because pixels belonging to microcracks are precisely matched with filters of a certain direction and scale, the response modulus of these pixels is significantly higher than that of pixels in the background area. Therefore, the maximum response modulus obtained under all directions and scales can directly reflect the texture intensity of the microcracks; the standard deviation of the directional response of a pixel can reflect the distribution difference of the response modulus of that pixel; the mean of the directional response of a pixel can reflect the overall level of the response modulus of that pixel; the coefficient of variation is a well-known technique and will not be elaborated here.

[0040] Specifically, using pixels in the surface image Taking this as an example, we can obtain the pixel points. The crack significance index can be found in the following formula: ; in, It is a pixel. The crack significance index; It is a pixel. In the The overall directional response value in each direction. ; This means taking the maximum value of the pixel's combined response value across all directions, i.e., the strongest response in the optimal direction, to capture the intensity of the crack texture; It is a pixel. Maximum directional composite response value in all directions; It is a pixel. The standard deviation of the directional response; It is a pixel. The mean directional response; It is a constant to prevent the denominator from being zero, and is usually taken as... The purpose is to prevent the denominator from being zero; The preset exponent is a constant with a value greater than 1. For example, [the exponent is...]. It is used to nonlinearly amplify the coefficient of variation.

[0041] It should be noted that in the relational expression Its function is to capture the texture intensity of microcracks. Microcracks, as slender linear structures, only produce a large response modulus when matched with the optimal direction of their orientation and the optimal scale of their width; the relationship in... Its function is to reflect the orientation specificity of microcracks; essentially, it is a measure of the coefficient of variation. The reason for performing power-law enhancement is that pixels belonging to microcracks have strong direction specificity, while pixels belonging to the background region have weak direction specificity. That is, pixels belonging to microcracks only produce strong responses in the optimal direction, and the responses in other directions are significantly weakened. Pixels belonging to the background region have no fixed orientation or scale characteristics, so their responses are uniform and small in all directions and scales.

[0042] Specifically, when pixel The larger the maximum value of the combined response in all directions, the greater the power term. When the size is larger, the number of pixels The larger the crack significance index, the lower the value; further, obtain the crack significance index for each pixel.

[0043] It should be noted that when the pixel In low-light areas, the overall response intensity of the Gabor filter decreases. However, since the orientation of the microcracks remains unchanged, the distribution characteristic of strong response in the optimal direction and weak response in other directions still exists. Therefore, the gain of its power-term output is still relatively large, and the pixel value remains high. The crack saliency index of the microcrack is much higher than that of the background pixels. Therefore, it can ensure that pixels belonging to the microcrack region in the low-light area are prevented from having their microcrack features obscured by the light. Meanwhile, the pixels in the background region do not have additional output gain. Therefore, it can separate the pixels in the microcrack region from the pixels in the background region, providing a robust feature basis for subsequent continuous connections.

[0044] Please see Figure 2 , Figure 2 The diagram illustrates the crack significance index. As shown in the diagram, the crack significance index achieves a higher value in the crack area, forming a continuous high-response band consistent with the crack direction, while the background area on the cable surface shows a lower value. This effectively distinguishes crack features from background noise, preserving the fine morphology and branch details of the crack while suppressing interference from normal textures and uneven lighting, thus providing a reliable feature basis for subsequent defect segmentation and quantitative evaluation.

[0045] S3. Based on the crack significance index of the pixel, the segmentation threshold is determined by the maximum inter-class variance method to generate a binarized image; morphological operations and connected component extraction are performed on the binarized image to obtain several total defect regions; the comprehensive defect degree is obtained based on the defect area ratio of the total defect regions and the average defect area.

[0046] It should be noted that in step S2, the crack significance index of each pixel in the surface image is obtained. Its core feature is that the crack significance index of the pixels in the microcrack region is significantly higher than that of the pixels in the background region. Moreover, after illumination compensation, the crack significance index can maintain a clear numerical difference in both high-illuminance and low-illuminance regions. Therefore, the crack significance index can be used as the core basis to screen candidate crack regions through threshold segmentation. Combined with morphological operations and connected component analysis, the fractured microcracks can be repaired, ultimately achieving accurate extraction and detection of microcracks.

[0047] Specifically, an adaptive threshold segmentation method is first used to screen candidate crack regions: For all crack significance indices obtained in step S2, the Otsu's method is used to adaptively determine the segmentation threshold. The Otsu's method can automatically divide defect pixels with high crack significance indices and background pixels with low crack significance indices based on the grayscale distribution of the index map. To avoid the problem of missing dark cracks or falsely detecting bright backgrounds in areas with uneven illumination when using a fixed threshold, the following segmentation rules are set: If the crack significance index of a pixel is greater than or equal to the segmentation threshold, the pixel is determined to be a defect pixel; if the crack significance index of a pixel is less than the segmentation threshold, the pixel is determined to be a background pixel. This results in a binary image containing only defect pixels, where the area of ​​defect pixels is white and the area of ​​background pixels is black. The Otsu's method is an existing technology and will not be described in detail here.

[0048] Furthermore, morphological processing is performed on the binarized image to optimize the continuity and integrity of the candidate crack regions: [The following is a separate, unrelated sentence:] Using... The rectangular structuring element is used to dilate the binarized image once to connect crack breaks caused by uneven lighting or tiny gaps. This dilation operation fills the small gaps between defective pixels, restoring the continuous morphology of the microcracks. After dilation, the same structuring element is eroded once to eliminate isolated noise points amplified during the dilation process, such as pixels in the background with crack saliency indices close to the segmentation threshold, while maintaining the original width and edge morphology of the microcracks and preventing excessive expansion of their outlines. Connected region extraction is then performed on the morphologically processed binarized image, with a threshold area set for the connected region. (Example provided). The value is greater than or equal to 10 pixels, and the filtered area is less than the connected region area threshold. Isolated small connected regions are mostly residual false responses of illumination spots. Connected regions with an area greater than or equal to the connected region area threshold are taken as the final total defect region, resulting in several total defect regions. The connected region area threshold is related to the image resolution. At other resolutions, it can be adjusted according to the area of ​​the pixel corresponding to the smallest physical size of the actual microcrack. If the Euclidean distance between different connected regions is less than 3 pixels, they can be merged into one total defect region to reflect the continuity of the actual crack.

[0049] Please see Figure 3 , Figure 3 The diagram illustrates the morphological processing. As shown in the figure, after morphological processing, the crack area is extracted as continuous connected lines. At the same time, isolated noise points and small stray interferences are effectively filtered out, while the complete direction and branch structure of the crack are preserved. This provides a clear binary feature basis for subsequent area statistics, morphological analysis and the generation of the final defect detection results.

[0050] Specifically, the total number of pixels in all defective regions and the total number of pixels in the surface image are counted, and the ratio of the two is taken as the area ratio of the total defective regions, denoted as . This ratio directly reflects the coverage of microcracks on the surface of fire-resistant wires and is a core parameter for assessing the severity of defects. The larger the ratio, the wider the crack distribution and the more serious the damage to the overall structural integrity and fire resistance of the wire.

[0051] Furthermore, the ratio of the total number of pixels in all defective regions to the total number of defective regions is taken as the average defective region area, denoted as . When the total number of pixels in the total defect area is fixed, the average defect area is... The size directly reflects the distribution characteristics of the microcrack defect: if the average defect area A larger average defect area indicates a greater actual propagation range for a single microcrack. These concentrated, large-size cracks pose a more significant risk of penetration and structural tearing of the wire insulation. Smaller, but accounting for a smaller percentage of the total defect area. A higher number of microcracks means that the microcracks exist in the form of densely distributed small areas. Although the harm of a single crack is limited, the superposition of hidden dangers in multiple areas will reduce the overall fire resistance reliability and structural integrity of the wire, indirectly reflecting the core impact of the total number of microcracks.

[0052] It should be noted that, in addition, the average defect area can avoid the limitations of using only the area ratio of the total defect area for assessment: when the area ratio of the total defect area in the surface images of different fire-resistant wires is the same, the average defect area can effectively distinguish between two different risk types: a small number of large cracks and a large number of small cracks. This provides a more comprehensive area-derived dimension for the subsequent calculation of the overall defect level, ensuring that the assessment results cover both the overall coverage area and the intensity of local defects.

[0053] Furthermore, based on the minimum size of a single microcrack that needs to be considered in industrial scenarios, a preset critical single defect area is set, denoted as... An example value of 50 pixels can be used, but this can be adjusted according to the wire type and inspection standards; based on the proportion of the total defective area. and average defect area and the preset critical single defect area Obtain the overall defect level of the fire-resistant wire corresponding to the surface image. For details, please refer to the following relation: ; in, It represents the overall defect level of the fire-resistant wire corresponding to the surface image; It is a calibration coefficient based on the average defect area. The maximum area of ​​a single defect shall not exceed twice the preset critical area. The assumption, exemplarily taken ; It is the percentage of the total defective area; It is the average area of ​​the defect region; It is a preset critical single defect area; when the area of ​​the total defect area is larger and the average defect area is larger, the comprehensive defect degree of the fire-resistant wire corresponding to the surface image is greater, and vice versa.

[0054] S4. Based on the comparison results of the comprehensive defect degree and the abnormal threshold, the detection results of microcracks in fire-resistant wires are obtained.

[0055] Based on the overall degree of defect The detection results of microcracks in fire-resistant wires are obtained by comparing the results with the abnormal threshold. The specific method is as follows: Set the first abnormal threshold. Second abnormal threshold It can be configured according to actual needs, and must meet certain requirements. When the overall defect level Conditions met: At this point, the fire-resistant wire is judged to be in a qualified state, with a small coverage area of ​​microcracks and no large-sized cracks; when the overall defect degree is... Conditions met: If the fire-resistant wire is determined to have a minor defect, it needs to be returned to the factory for repair. If the overall defect severity is... Conditions met: If the fire-resistant wire is determined to be severely defective, then the entire batch of wire must be prohibited from leaving the factory or being used. A comprehensive investigation of the production process is necessary to identify the root causes of problems, such as in the raw materials, extrusion process, and bending test procedures. After rectification, a small batch of samples must be reproduced and tested throughout the entire process to confirm the overall degree of defect. Less than or equal to the first abnormal threshold Only then can mass production or use be resumed.

[0056] Please see Figure 4 , Figure 4This diagram illustrates the final defect detection results. As shown in the diagram, the final defect detection results accurately distinguish the cracked area from the normal surface of the cable. The complete shape, direction, and branch structure of the crack are clearly marked. This not only accurately restores the crack distribution in the original image but also avoids misjudging the normal texture. It intuitively presents the actual defect state of the microcracks on the surface of the fire-resistant wire, providing an intuitive and reliable basis for subsequent defect degree quantification and quality classification.

[0057] This invention also discloses a microcrack detection system for fire-resistant electrical wires, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a microcrack detection method for fire-resistant electrical wires according to the present invention.

[0058] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A method for detecting microcracks in fire-resistant electrical wires, characterized in that, include: Image images of the fire-resistant wire surface were obtained through image acquisition and preprocessing; Gabor filter banks of different directions and scales were constructed. Each Gabor filter in the Gabor filter bank is convolved with each pixel in the surface image in two dimensions to obtain the response magnitude of each pixel in the corresponding direction and scale. The directional comprehensive response value of the pixel in any direction is obtained based on the maximum response magnitude of the pixel in all scales. The mean and standard deviation of the directional response of the pixel are calculated based on the directional comprehensive response value of the pixel in all directions. The crack significance index of a pixel is obtained by multiplying the maximum directional comprehensive response value by the power of the ratio of the standard deviation of the directional response to the mean of the directional response, where the power is a constant greater than 1. Based on the crack significance index of each pixel, the segmentation threshold is adaptively determined by the maximum inter-class variance method to obtain a binarized image containing only defect pixels; morphological operations and connected component extraction are performed on the binarized image to obtain several total defect regions; the comprehensive defect degree is obtained based on the defect area ratio of the total defect regions and the average defect region area. Based on the comprehensive defect level and abnormal threshold, the detection results of microcracks in fire-resistant wires are obtained.

2. The method for detecting microcracks in fire-resistant electrical wires according to claim 1, characterized in that, The construction of Gabor filter banks with different orientations and scales includes: The number of direction types is set to 13, meaning 13 Gabor filters with different directions are constructed; the direction numbers are denoted as... The range of values ​​is Integers between 0° and 0°, 13 directions starting from 0°, with each pair of adjacent directions spaced apart. The number of scale types is set to 7, meaning 7 Gabor filters with different scales are constructed. The scale numbers are denoted as follows: The range of values ​​is Integers between [a certain range].

3. The method for detecting microcracks in fire-resistant electrical wires according to claim 1, characterized in that, The Gabor filter banks of different scales also include: Set the first The scale factor corresponding to each scale is: Its value is calculated according to the following formula: , It is the scale number; setting the number The wavelength at each scale is 2.5 times the scale factor; the bandwidth is set to 0.4 and the phase to 0, thus obtaining wavelengths in different directions. and different scales The Gabor filter below, denoted as .

4. The method for detecting microcracks in fire-resistant electrical wires according to claim 1, characterized in that, The step of performing a two-dimensional convolution operation between each Gabor filter in the Gabor filter bank and each pixel in the surface image to obtain the response magnitude of each pixel in the corresponding direction and at the corresponding scale includes: pixels With in different directions and different scales Gabor filter below Perform 2D convolution, and use mirror padding at the edges to avoid boundary effects to obtain pixel points. In the Direction, First The response modulus of the scale is denoted as .

5. The method for detecting microcracks in fire-resistant electrical wires according to claim 4, characterized in that, The step of obtaining the directional comprehensive response value of a pixel in any direction based on the maximum value of the response modulus corresponding to all scales in any direction includes: pixels In the The corresponding directional integrated response value under the direction is denoted as ,in, , ,and It is an integer.

6. The method for detecting microcracks in fire-resistant electrical wires according to claim 1, characterized in that, The method for obtaining the crack significance index of a pixel includes: ； in, It is a pixel. The crack significance index; It is a pixel. In the The overall directional response value in each direction. ; This indicates that the maximum value of the pixel's overall response value across all directions is taken; It is a pixel. Maximum directional composite response value in all directions; It is a pixel. The standard deviation of the directional response; It is a pixel. The mean directional response; It is a constant to prevent the denominator from being zero; The preset exponent is a constant with a value greater than 1.

7. The method for detecting microcracks in fire-resistant electrical wires according to claim 1, characterized in that, The step of adaptively determining the segmentation threshold based on the crack significance index of each pixel using the maximum inter-class variance method to obtain a binarized image containing only defect pixels includes: If the crack significance index of a pixel is greater than or equal to the segmentation threshold, the pixel is determined to be a defective pixel; if the crack significance index of a pixel is less than the segmentation threshold, the pixel is determined to be a background pixel. This results in a binary image containing only defective pixels, where the area of ​​defective pixels is white and the area of ​​background pixels is black.

8. The method for detecting microcracks in fire-resistant electrical wires according to claim 1, characterized in that, The method of obtaining the comprehensive defect level based on the proportion of the total defect area and the average defect area includes: ； in, It represents the overall defect level of the fire-resistant wire corresponding to the surface image; It is the calibration coefficient; It is the percentage of the total defective area; It is the average area of ​​the defect region; It is the preset critical single defect area.

9. The method for detecting microcracks in fire-resistant electrical wires according to claim 1, characterized in that, The method for obtaining the detection results of microcracks in fire-resistant electrical wires based on the comprehensive defect degree and abnormal threshold includes: Set the first abnormal threshold. Second abnormal threshold It needs to meet the following requirements. When the overall defect level Conditions met: At this point, the fire-resistant wire is judged to be in a qualified state, with a small coverage area of ​​microcracks and no large-sized cracks; when the overall defect degree is... Conditions met: If the fire-resistant wire is determined to have a minor defect, it needs to be returned to the factory for repair. If the overall defect severity is... Conditions met: If the fire-resistant wire is determined to be in a severely defective state, then the batch of wire must be prohibited from leaving the factory or being used.

10. A microcrack detection system for fire-resistant electrical wires, characterized in that, include: The processor and memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the method for detecting microcracks in fire-resistant electrical wires according to any one of claims 1-9.

Images