Composite board welding quality detection method and system based on image processing

Abstract

The present application belongs to the technical field of image processing, and particularly relates to a composite board welding quality detection method and system based on image processing, which comprises the following steps: performing morphological gradient and threshold segmentation preprocessing on the collected welding image to obtain a welding area image; calculating a defect spread index for any pixel point in the welding area image, and determining an optimal frequency in combination with a frequency boundary and the maximum and minimum values of the defect spread index; correcting a basic bandwidth by using the fluctuation of the optimal frequency of the neighborhood pixel points to determine a dynamic bandwidth; constructing a pixel-by-pixel adaptive Log-Gabor filter bank by using the optimal frequency and the dynamic bandwidth, extracting a defect energy distribution map, and identifying defects. The present application realizes adaptive adjustment of filter parameters through physical width perception, solves the problem that fixed parameters cannot take into account different types of defects, and improves the accuracy of quality detection.

Description

Technical Field

[0001] This invention relates to the field of image processing technology. More specifically, this invention relates to a method and system for inspecting the welding quality of composite plates based on image processing. Background Technology

[0002] Composite panels are bimetallic materials made of a base layer and a cladding layer, and are widely used due to their combination of corrosion resistance and high strength. However, during the welding process of composite panels, due to the differences in the thermophysical properties of the two metal materials and the influence of factors such as uneven release of thermal stress, it is very easy for cladding peeling or incomplete fusion to occur at the joint interface, which seriously affects the mechanical properties and sealing performance of the composite panel. Therefore, high-precision non-destructive testing of the welding quality of composite panels is particularly important.

[0003] Currently, image processing-based nondestructive testing technology is widely used. Among them, the phase consistency algorithm, which utilizes frequency domain phase information to extract features, can effectively resist the interference of non-uniform reflection and processing texture on metal surfaces, and has become a core technology for quality inspection.

[0004] However, existing phase consistency algorithms typically use a fixed-parameter Log-Gabor filter bank for frequency domain analysis. In actual composite plate welding quality inspection scenarios, the scale range of delamination and non-fusion defects is extremely large, which may include micron-level stress microcracks or wide interlayer non-fusion or voids. Using a filter with a fixed center frequency and bandwidth cannot simultaneously take into account defect features of multiple scales in the same image. When the actual physical scale of the defect exceeds the frequency response range of the preset filter, it will cause the amplitude response of the filter output to be weak, resulting in a sharp drop in the calculated phase consistency response value, causing serious missed detection problems and making the composite plate welding quality assessment results unreliable. Summary of the Invention

[0005] To address the aforementioned technical problem of scale mismatch caused by fixed filter parameters, the present invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a composite plate welding quality inspection method based on image processing, comprising: acquiring a welding image and performing morphological gradient and threshold segmentation preprocessing on the welding image to obtain a welding area image; for any pixel in the welding area image: weighting the distance from the pixel to each neighboring pixel based on the gradient magnitude of each neighboring pixel to determine a defect widening index; statistically obtaining the maximum and minimum values ​​of defect width in the project and calculating the frequency boundary; determining the optimal frequency of the pixel based on the defect widening index, the maximum and minimum values ​​of the defect widening index, and the frequency boundary; correcting the basic bandwidth using the fluctuation of the optimal frequency of the neighboring pixels to determine the dynamic bandwidth of the pixel; constructing a pixel-wise adaptive Log-Gabor filter bank using the optimal frequency and dynamic bandwidth of each pixel, using the filter bank to perform frequency domain filtering on the welding area image, extracting the amplitude and phase of each group, and substituting them into a phase consistency model to obtain a defect energy distribution map; extracting the bright area pixels in the defect energy distribution map using the Otsu method threshold segmentation algorithm and marking them as defect area pixels; if the number of defect area pixels exceeds a preset threshold, the welding quality is determined to be unqualified.

[0007] This invention addresses the issue of preprocessing the welding area to avoid background texture interference. Its core lies in calculating the defect widening index to measure the physical width of the defect, determining the optimal frequency based on engineering limits, and adjusting the bandwidth according to neighborhood frequency fluctuations to obtain dynamic bandwidth. Furthermore, it constructs a pixel-by-pixel adaptive Log-Gabor filter bank to extract defect features. This method solves the problem that fixed-parameter filters cannot simultaneously handle microcracks and large unfused defects, ensuring that the filter frequency always matches the current defect scale. This prevents microcrack signals from being smoothed and missed, and also avoids the hollowing out of large defects, significantly improving the accuracy and reliability of composite plate welding quality inspection.

[0008] Preferably, the preprocessing of the welding image by morphological gradient and threshold segmentation to obtain the welding region image includes: calculating the morphological gradient of the welding image to obtain a gradient image; performing threshold segmentation on the gradient image using the Otsu method to obtain a binarized mask; wherein, white pixels are welding region pixels, the area of ​​all white connected regions in the binarized mask is extracted, and the region containing the bounding rectangle of the connected region with the largest area is taken as the welding region image.

[0009] Preferably, the method for obtaining the neighboring pixels is as follows: taking any pixel as the center, extract all pixels within the eight neighboring range of that pixel as the neighboring pixels of that pixel.

[0010] Preferably, the defect broadening index satisfies the expression: In the formula, For the first Defect broadening index per pixel; For the first The first pixel The gradient magnitude of each neighboring pixel; For the first From the pixel point to its _th pixel The square of the Euclidean distance between the neighboring pixels; This represents the index value and total number of neighboring pixels.

[0011] This invention uses the gradient magnitude of neighboring pixels to weight the distance and evaluates the spatial dispersion of gradient energy relative to the center pixel to reflect the physical width characteristics of the defect. It effectively distinguishes between microcracks with steep and concentrated gradients and unfused voids with gentle and dispersed gradients, providing an accurate physical scale prior basis for the frequency selection of subsequent frequency domain filters and improving the accuracy of describing the characteristics of defects of different morphologies.

[0012] Preferably, the statistical acquisition of the maximum and minimum defect widths in the project and the calculation of frequency boundaries include: statistically analyzing the maximum and minimum defect widths appearing in the composite plate welding area images in the project, and taking the reciprocal of the maximum defect width as the lowest frequency and the reciprocal of the minimum defect width as the highest frequency.

[0013] This invention transforms the extreme values ​​of defect width obtained from engineering statistics into frequency boundaries, establishing the physical effective range of the filter response. This ensures that the frequency mapping can fully cover all actual defect scales, from microcracks to large-area hollows, preventing detection failures caused by frequencies exceeding the response range and guaranteeing the applicability of the algorithm in practical engineering applications.

[0014] Preferably, the optimal frequency satisfies the expression: In the formula, For the first The optimal frequency for each pixel; The highest frequency; The lowest frequency; These represent the maximum and minimum values ​​of the defect widening index; For the first Defect broadening index per pixel.

[0015] This invention achieves a precise inverse mapping from the physical broadening index to the filter center frequency. This mechanism automatically matches high frequencies to narrow defects to preserve details and low frequencies to fill the contours of wide defects. It realizes the synchronous enhancement and extraction of multi-scale defect features in the same image, effectively solving the problems of missed detection of small defects or hollowing of macroscopic defects.

[0016] Preferably, the dynamic bandwidth satisfies the expression: In the formula, For the first Dynamic bandwidth per pixel; Basic bandwidth; For the first The standard deviation of the best frequency of the neighboring pixels of a pixel; It represents the maximum value among the standard deviations of the best frequencies of the neighboring pixels of each pixel in the welding image.

[0017] This invention achieves dynamic adjustment of filter bandwidth by introducing frequency standard deviation correction bandwidth, automatically expanding the frequency coverage range in defect scale abrupt change regions, effectively preventing signal breakage caused by frequency jumps, ensuring the continuity and integrity of defect energy distribution map, and improving the reliability of defect detection.

[0018] Preferably, the preset threshold is equal to 10% of the total number of pixels in the welding area image.

[0019] Preferably, the method further includes: performing grayscale processing on the welding area image, and using the Sobel operator to calculate the gradient magnitude of each pixel in the welding area image.

[0020] Secondly, the present invention provides a composite plate welding quality inspection system based on image processing, 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 composite plate welding quality inspection method based on image processing is implemented.

[0021] By adopting the above technical solution, the above image processing-based composite plate welding quality inspection method is generated into a computer program and stored in a memory so that it can be loaded and executed by a processor. This allows for the creation of a terminal device based on the memory and processor, making it convenient to use.

[0022] The beneficial effects of this invention are as follows:

[0023] (1) This invention constructs an inverse mapping mechanism between the defect broadening index and the center frequency of the filter, realizing the precise matching between the filter parameters and the physical scale of the defect, effectively solving the problem of missed detection of small defects or hollowing of macroscopic defects caused by single-frequency filtering, and significantly improving the detection accuracy of multi-scale defects.

[0024] (2) The present invention corrects the basic bandwidth by adjusting the fluctuation of the neighborhood optimal frequency, which effectively enhances the adaptability of the algorithm in defect edges and scale change regions, avoids the discontinuous attenuation of the signal when switching frequencies, and improves the robustness of complex defect detection. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating the image processing-based composite plate welding quality inspection method of the present invention. Detailed Implementation

[0026] 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.

[0027] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0028] This invention discloses an image processing-based composite plate welding quality inspection method, referring to... Figure 1 This includes steps S1-S5:

[0029] S1. Acquire welding images and perform morphological gradient and threshold segmentation preprocessing on the welding images to obtain welding area images.

[0030] It should be noted that the acquired composite plate welding surface image contains two material regions: the base layer and the superimposed layer. Both of these material regions have strong texture noise. If the entire image is processed directly, the strong texture noise in the non-welding areas will seriously interfere with the accuracy of subsequent defect identification. Therefore, it is necessary to locate the core welding area first.

[0031] Specifically, a high-resolution industrial camera and a coaxial ring light source are deployed on the inspection line to obtain welding images under uniform illumination. Considering that the surface texture of the welding area is usually rougher than that of the smooth plate area, the traditional morphological gradient calculates the morphological gradient of the image through a rectangular kernel, smoothing out the gradient value of the plate area and enhancing the gradient value of the rough welding area.

[0032] Therefore, a gradient image is obtained by calculating the morphological gradient of the welding image; a binary mask is obtained by thresholding the gradient image using Otsu's method; white pixels represent welding area pixels and black pixels represent plate area pixels; all white connected regions are found and the area of ​​each connected region is counted, and the region containing the bounding rectangle of the connected region with the largest area is taken as the welding area image.

[0033] The welding area image is converted to grayscale, and the Sobel operator is used to calculate the gradient magnitude of each pixel in the welding area image. Simultaneously, the position of each pixel in the welding area image needs to be obtained. Specifically, the pixel at the bottom left corner of the welding area image is taken as the origin, and the horizontal direction to the right from the origin is used as... The positive direction of the axis is defined by taking the vertically upward direction from the origin as... The positive direction of the axis is used to construct a Cartesian coordinate system; the position of each pixel in the welding area image is obtained in the Cartesian coordinate system, including... Coordinates in the axial direction and Coordinates along the axis.

[0034] At this point, the gradient magnitude and position of each pixel in the welding area image have been obtained.

[0035] S2. For any pixel in the welding area image: Based on the gradient magnitude of each neighboring pixel, the distance from the pixel to each neighboring pixel is weighted to determine the defect widening index.

[0036] It should be noted that in the actual welding process of composite panels, considering the significant differences in the thermophysical properties of the base layer and the cladding layer, microcracks or incomplete fusion voids are easily generated at the welding interface. The edge gradient of microcracks is often steep and concentrated, with energy confined in a very small space; while the edge gradient of incomplete fusion voids is gentle and dispersed, with energy distributed in a wider space. Therefore, the physical width of potential defects can be assessed by analyzing the degree of dispersion of local gradient energy relative to the center point.

[0037] Specifically, for any pixel in the welding area image, all pixels within the eight-neighbor range of that pixel are extracted as the neighboring pixels of that pixel.

[0038] Based on the gradient magnitude of each neighboring pixel, the distance from this pixel to each neighboring pixel is weighted to determine the defect widening index, which satisfies the expression:

[0039]

[0040] In the formula, For the first Defect broadening index per pixel; For the first The first pixel The gradient magnitude of each neighboring pixel; For the first From the pixel point to its _th pixel The square of the Euclidean distance between the neighboring pixels; This represents the index value and total number of neighboring pixels.

[0041] in, Reflecting the The spatial dispersion of the gradient magnitude of a pixel relative to the gradient magnitudes of its neighboring pixels, when the... When a pixel is located at a macroscopic defect on a relatively wide physical scale, the neighboring pixels with high gradients are distributed at a distance of [missing value]. The position of the farth pixel makes the first pixel... The larger the defect broadening index of a pixel, the greater the defect broadening index; when the... When a pixel might be located at a microscopic defect on an extremely narrow physical scale, high-gradient neighboring pixels will cluster tightly together at the first... Around the nth pixel, making the nth pixel The smaller the defect broadening index of each pixel, the better; it should be added that when the... When the sum of the gradient magnitudes of the neighboring pixels of a pixel is 0, it indicates that there is no gradient change in the neighborhood of that pixel, and the defect broadening index of that pixel is set to 0.

[0042] At this point, the defect widening index of all pixels in the welding area image has been obtained.

[0043] S3. Statistically obtain the maximum and minimum values ​​of defect width in the project and calculate the frequency boundary; determine the optimal frequency of the pixel based on the defect widening index, the maximum and minimum values ​​of the defect widening index, and the frequency boundary.

[0044] It should be noted that in frequency domain image processing, there is an inverse relationship between the center frequency of a filter and the physical scale of the features it can respond to. High-frequency components correspond to detailed features with drastic gray-level changes and extremely narrow physical scales in the image, while low-frequency components correspond to regional features with gradual gray-level changes and wider physical scales. Therefore, using a single fixed-frequency filter cannot simultaneously handle defects of different scales. For example, microcracks in multilayer peeling are extremely narrow and deep, with a very small defect widening index. If a low-frequency filter is used, the tiny signals from the crack will be treated as noise due to the excessively long wavelength. Smoothing can lead to missed detections; and the physical morphology of unfused hollow bulges is wide and flat with a large defect broadening index. If a high-frequency filter is used for processing, the wavelength is too short, and the filter can only respond to the boundary of the hollow bulge, but cannot cover the gradual gray-scale changes inside the hollow bulge, resulting in the detected defect appearing hollow and the defect area cannot be accurately assessed. Therefore, a frequency domain inverse mapping mechanism based on physical width perception is proposed. That is, the broadening index is used as prior knowledge to dynamically resolve the optimal frequency required for the current pixel, thereby ensuring that the filter always works at the optimal frequency point that best matches the defect scale.

[0045] Specifically, frequency boundaries are obtained, including the lowest and highest frequencies. The method of obtaining these frequencies is to statistically analyze the maximum and minimum defect widths appearing in the composite plate welding area images during the project, and to take the reciprocal of the maximum defect width as the lowest frequency and the reciprocal of the minimum defect width as the highest frequency.

[0046] According to actual statistics, microcracks caused by multilayer peeling are usually represented as extremely fine lines in the image. The width of such defects is usually between 2 and 5 pixels. However, for interlayer non-fusion or voids, the physical scale is larger, and the width of such defects is usually between 30 and 60 pixels. Therefore, in this embodiment, the minimum defect width is taken as 2 pixels and the maximum defect width is taken as 60 pixels.

[0047] Statistically calculate the defect widening index of all pixels in the weld area image, and obtain the maximum and minimum values ​​of the defect widening index.

[0048] The optimal frequency for any pixel is determined based on its defect broadening index, maximum and minimum defect broadening index values, and frequency boundaries; the optimal frequency satisfies the expression:

[0049]

[0050] In the formula, For the first The optimal frequency for each pixel; The highest frequency; The lowest frequency; These represent the maximum and minimum values ​​of the defect widening index; For the first Defect broadening index per pixel.

[0051] in, Reflects the range of frequency variation; Reflecting the The frequency variation adjustment factor for the nth pixel; the larger this value, the more significant the adjustment. The closer the defect broadening index of the nth pixel is to the minimum value of the defect broadening index, that is, the more accurate the calculation of the defect broadening index. The narrower the physical width of a defect per pixel, the higher the frequency needs to be matched to resolve detailed features and prevent small defects from being missed; conversely, the lower the frequency needs to be matched to fully cover the macroscopic outline and prevent large defects from becoming hollow. Reflecting the The actual frequency increment required for each pixel relative to the lowest frequency boundary is obtained by adding the frequency increment to the lowest frequency to obtain the optimal frequency that matches the current physical scale of the defect; it should be added that when and When they are equal, it indicates that the image defect broadening index is constant. At this point, let... equal and The average value.

[0052] At this point, the optimal frequency for all pixels in the welding area image has been obtained.

[0053] S4. The base bandwidth is corrected by utilizing the fluctuation of the optimal frequency of the neighboring pixels of the pixel to determine the dynamic bandwidth of the pixel.

[0054] It should be noted that, considering that the defect morphology in the actual welding process is often continuously changing, the defect broadening index of pixels in this area may change drastically, causing the calculated frequency to jump drastically in space. If the filter bandwidth is too narrow at this time, the signal will be lost because the frequency cannot keep up. Therefore, by analyzing the instability of all the optimal frequencies, the bandwidth of the filter is adjusted to ensure the best filtering effect.

[0055] Specifically, the base bandwidth is corrected based on the fluctuation of the optimal frequency of the neighboring pixels of any given pixel to determine the dynamic bandwidth of that pixel; the dynamic bandwidth satisfies the expression:

[0056]

[0057] In the formula, For the first Dynamic bandwidth per pixel; Basic bandwidth; For the first The standard deviation of the best frequency of the neighboring pixels of a pixel; It represents the maximum value among the standard deviations of the best frequencies of the neighboring pixels of each pixel in the welding image.

[0058] in, Reflecting the The relative drasticness of the optimal frequency change within the neighborhood of the nth pixel; the larger this value, the more pronounced the change. A sharp jump in the spatial distribution of the optimal frequency around a pixel indicates that the pixel may be in a complex transition region of a defect morphology or a region of abrupt scale change. In this case, a significantly increased bandwidth is needed to expand the frequency coverage of the filter and prevent signal breakage due to frequency inaccuracy. Conversely, maintaining or slightly exceeding the basic bandwidth is sufficient. It should be noted that the basic bandwidth of a Log-Gabor filter, i.e., the minimum allowable bandwidth, is typically 0.55. When the value is 0, it indicates that the optimal frequency distribution of the image is very stable, and no bandwidth adjustment is needed. At this point, the bandwidth can be set... It equals the base bandwidth.

[0059] At this point, the dynamic bandwidth of all pixels in the welding area image has been obtained.

[0060] S5. Utilize the optimal frequency and dynamic bandwidth of each pixel to construct a dynamic Log-Gabor filter, and perform frequency domain analysis on the welding image to obtain the defect energy distribution map; use the Otsu method threshold segmentation algorithm to extract the bright pixels in the defect energy distribution map as defect region pixels. If the number of defect region pixels exceeds the preset threshold, the welding quality is determined to be unqualified.

[0061] It should be noted that traditional phase consistency models, when acquiring the defect energy distribution map of the welding area image, usually rely on a set of preset Log-Gabor filters with fixed frequency and bandwidth to perform uniform convolution processing on the entire image. Since filters with fixed parameters are prone to signal breakage or missed detection, the final quality detection is unreliable. Therefore, based on the optimal frequency and dynamic bandwidth calculated in the previous steps, filter parameters are constructed for each pixel, thereby achieving adaptive and accurate capture of multi-scale defects in the energy map.

[0062] Specifically, by utilizing the optimal frequency and dynamic bandwidth of each pixel in the welding area image, a pixel-wise adaptive Log-Gabor filter bank is constructed. This filter bank is then used to perform frequency domain filtering on the welding area image, extracting the amplitude and phase of each group, and substituting them into the phase consistency model to obtain the defect energy distribution map.

[0063] Among them, the defect energy distribution map shows high response values ​​at microcracks and non-fusion sites, which are significantly different from normal welding areas; and the phase consistency model is a well-known technology, so it will not be elaborated here.

[0064] The Otsu threshold segmentation algorithm is further used to extract the pixels in the high-brightness area of ​​the defect energy distribution map and mark them as defect area pixels. If the number of defect area pixels exceeds the preset threshold, the welding quality is determined to be unqualified and an early warning is triggered immediately.

[0065] The preset threshold can be 10% of the total number of pixels in the welding area image, and can be adjusted according to the actual situation.

[0066] This invention also discloses an image processing-based composite plate welding quality inspection system, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the image processing-based composite plate welding quality inspection method according to this invention.

[0067] 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 composite plate welding quality inspection method based on image processing, characterized in that, include: Welding images are acquired, and morphological gradient and threshold segmentation preprocessing is performed on the welding images to obtain welding area images; For any pixel in the weld area image: based on the gradient magnitude of each neighboring pixel, the distance from the pixel to each neighboring pixel is weighted to determine the defect widening index, satisfying the expression: , For the first Defect broadening index per pixel For the first The first pixel The gradient magnitude of each neighboring pixel. For the first From the pixel point to its _th pixel The square of the Euclidean distance between each neighboring pixel. , These are the index values ​​and total number of neighboring pixels; The maximum and minimum defect widths in the project are statistically obtained, and the frequency boundaries are calculated. This includes: statistically analyzing the maximum and minimum defect widths appearing in the composite plate welding area images, and taking the reciprocal of the maximum defect width as the lowest frequency and the reciprocal of the minimum defect width as the highest frequency; based on the defect widening index of the pixel, the maximum and minimum defect widths, and the frequency boundaries, the optimal frequency of the pixel is determined, satisfying the expression: , For the first The optimal frequency for each pixel The highest frequency, The lowest frequency, , The maximum and minimum values ​​of the defect broadening exponent are given; the base bandwidth is corrected by utilizing the fluctuation of the optimal frequency of the neighboring pixels of a pixel to determine the dynamic bandwidth of the pixel, satisfying the expression: , For the first Dynamic bandwidth per pixel Based on bandwidth, For the first The standard deviation of the best frequency of the neighboring pixels of a pixel The maximum value among the standard deviations of the best frequencies of the neighboring pixels of each pixel in the welding image; By utilizing the optimal frequency and dynamic bandwidth of each pixel, a pixel-adaptive Log-Gabor filter bank is constructed. This filter bank is used to perform frequency domain filtering on the welding area image, extract the amplitude and phase of each group, and substitute them into the phase consistency model to obtain the defect energy distribution map. The Otsu threshold segmentation algorithm is used to extract the pixels in the high-brightness area of ​​the defect energy distribution map and mark them as defect area pixels; if the number of defect area pixels exceeds the preset threshold, the welding quality is judged to be unqualified.

2. The composite plate welding quality inspection method based on image processing according to claim 1, characterized in that, The preprocessing of the welding image using morphological gradient and threshold segmentation to obtain the welding area image includes: The morphological gradient of the welding image is calculated to obtain the gradient image. The gradient image is then thresholded using the Otsu method to obtain a binary mask. White pixels represent the welding area pixels. The area of ​​all white connected regions in the binary mask is extracted, and the region containing the bounding rectangle of the largest connected region is taken as the welding area image.

3. The composite plate welding quality inspection method based on image processing according to claim 1, characterized in that, The method for obtaining the neighboring pixels is as follows: taking any pixel as the center, extract all pixels within the eight neighboring range of that pixel, and use them as the neighboring pixels of that pixel.

4. The image processing-based composite plate welding quality inspection method according to claim 1, characterized in that, The preset threshold is equal to 10% of the total number of pixels in the welding area image.

5. The composite plate welding quality inspection method based on image processing according to claim 1, characterized in that, The method further includes: performing grayscale processing on the welding area image, and using the Sobel operator to calculate the gradient magnitude of each pixel in the welding area image.

6. A composite plate welding quality inspection system based on image processing, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the image processing-based composite plate welding quality inspection method according to any one of claims 1-5.

Images