A composite board welding stress detection method and system

Abstract

The present application belongs to the technical field of image data processing, and particularly relates to a composite plate welding stress detection method and system, which comprises the following steps: acquiring a same-direction polarization image and a normal polarization image by using an active polarization imaging system; obtaining a mirror pollution index of a pixel point by evaluating the ratio of the gray difference and the sum of the two images; obtaining a reliability weight based on the mirror pollution index; performing multi-resolution pyramid decomposition and up-sampling reconstruction on the normal polarization image to obtain a reflectivity field; obtaining a normalized feature field by dividing the normal polarization image by the reflectivity field; obtaining a corrected image based on the normalized feature field and the reliability weight; and completing welding stress detection of the composite plate based on the corrected image. The present application separates the mirror highlight by dual polarization imaging, eliminates the material brightness difference by combining multi-resolution analysis, solves the image distortion problem caused by optical heterogeneity, and provides reliable image data for subsequent stress analysis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image data processing. More particularly, the present application relates to a composite plate welding stress detection method and system. BACKGROUND

[0002] In high-end equipment manufacturing such as aerospace, nuclear power, chemical industry and ocean engineering, titanium steel, nickel steel and other metal explosion composite plates are widely used due to their excellent comprehensive performance. However, due to the significant difference in thermal expansion coefficient and thermal conductivity between the cladding layer and the base material, the intense and uneven heat cycle in the welding process is prone to produce complex residual stress in the weld and heat affected zone. This stress not only causes manufacturing defects such as warping and angular deformation, but also significantly reduces the structural fatigue life and stress corrosion resistance, posing a major safety hazard.

[0003] Traditional residual stress detection methods such as X-ray diffraction method, neutron diffraction method and blind hole method have certain accuracy, but they are generally expensive, sensitive to surface state or destructive, and difficult to meet the industrial demand for online and full-field detection of large-size composite plates. In recent years, optical measurement techniques based on image processing such as digital image correlation method and structured light three-dimensional reconstruction have become a development trend due to their non-contact and full-field measurement advantages. However, when directly applied to the welding area of metal composite plates, a key bottleneck is the optical property interference caused by heterogeneous materials.

[0004] Specifically, the cladding layer, the base material and the weld fusion zone have completely different reflectivity, and the weld surface often has molten beads, oxidation skin or oil stains, which easily produce strong mirror highlights. Under the same lighting and exposure conditions, the images captured by the camera often have the phenomenon of gray saturation in the highlight area and gray hunger in the low reflectivity area, and the feature information is submerged, and the signal-to-noise ratio is insufficient. This coexistence of overexposure and underexposure image degradation seriously pollutes the speckle features required by the digital image correlation method or the stripe information required by the structured light, resulting in failure of displacement calculation or three-dimensional topography reconstruction.

[0005] Therefore, how to overcome the image quality degradation caused by optical heterogeneity in the welding area of composite materials and obtain standardized images with uniform feature distribution and complete and reliable information is a core prerequisite for realizing high-precision, non-contact residual stress detection. SUMMARY

[0006] To solve the technical problem that the image feature distortion caused by the optical heterogeneity of the composite plate welding area leads to the inability of the subsequent optical stress detection method to accurately calculate due to the lack of reliable image data, the present application provides solutions in the following aspects.

[0007] In a first aspect, the present application provides a composite plate welding stress detection method, comprising: using an active polarization imaging system to collect co-polarization images and cross-polarization images; for each pixel point in the images, obtaining a specular pollution index of the pixel point by evaluating the ratio of the gray difference and the gray sum of the co-polarization image and the cross-polarization image at the pixel point; and obtaining a reliability weight of the pixel point based on the specular pollution index; performing multi-resolution pyramid decomposition and interpolation-only up-sampling reconstruction on the cross-polarization image to estimate a reflectivity field; dividing the cross-polarization image by the reflectivity field to obtain a normalized feature field; obtaining a modified gray value of each pixel point based on the value of each pixel point in the normalized feature field and the reliability weight corresponding to the pixel point; the modified gray values of all pixel points form a modified image; and completing welding stress detection of the composite plate based on the modified image.

[0008] The present application realizes a double-field photometric normalization method by separating specular highlights and diffuse reflection features using dual-polarization imaging, and combining multi-resolution reflectivity field estimation and reliability weighting. The present application can generate a modified image with uniform feature contrast and reduced interference of material optical properties and specular highlights, thereby providing high-quality and reliable input data for subsequent stress analysis algorithms.

[0009] Preferably, the specular pollution index satisfies the expression: ; wherein, is the specular pollution index of the pixel point ; and and are coordinate indices of the pixel point. is the gray value of the pixel point in the co-polarization image; is the gray value of the pixel point in the cross-polarization image; and is a normal number.

[0010] The present application calculates the specular pollution index using the ratio of the gray difference and the gray sum of the co-polarization image and the cross-polarization image, and utilizes the physical difference between specular reflection and diffuse reflection in polarization state. Through normalization processing, it only reflects the proportion of specular reflection in total reflection, thereby reducing the interference of the brightness of the material itself and enabling more accurate assessment of highlight pollution.

[0011] Preferably, the reliability weight satisfies the expression: ; wherein, is the reliability weight of the pixel point ; and and are coordinate indices of the pixel point. is the specular pollution index of the pixel point ; and is a weight sharpening coefficient.

[0012] The present application introduces a weight sharpening coefficient greater than one to calculate the reliability weight, so that the weight attenuation is nonlinear, which can tolerate slight polarization imperfection that does not affect the calculation, while applying more resolute inhibition to moderate to severe highlight pollution areas, so that the weight quickly approaches zero, thereby achieving a balance between preserving effective information and removing polluted data.

[0013] Preferably, the multi-resolution pyramid decomposition of the orthogonally polarized images and the interpolation-only up-sampling reconstruction to estimate the reflectance field comprise: taking the orthogonally polarized images as input, constructing an image pyramid by iteratively performing Gaussian blur and down-sampling operations; starting from the top image of the image pyramid, iteratively performing up-sampling operations containing only interpolation to restore the image to the original size to obtain the reflectance field.

[0014] The present application uses Gaussian pyramid decomposition and interpolation-only up-sampling reconstruction to reflectance field, which can effectively smooth out high-frequency features while preserving steep brightness boundaries between different materials in low-frequency fields, reducing the color bleeding and blur produced by traditional large-radius Gaussian blur at material boundaries, and obtaining accurate background brightness estimation.

[0015] Preferably, the normalized feature field satisfies the expression: ; wherein, is the normalized feature field at pixel point ; and is the coordinate index of the pixel point; is the gray value of pixel point in the orthogonally polarized image; is the reflectance field at pixel point ; is a constant.

[0016] Preferably, the modified image satisfies the expression: ; wherein, is the modified image gray at pixel point ; and is the coordinate index of the pixel point; is the reliability weight at pixel point ; is the normalized feature field at pixel point .

[0017] The application can retain reliable and material brightness normalized features, replace unreliable pixels polluted by specular highlights with neutral background values, and finally output a corrected image with uniform feature contrast and no highlight artifacts by point-by-point weighting of normalized feature fields with reliability weights.

[0018] Preferably, the collecting of the co-polarization image and the cross-polarization image comprises:

[0019] The co-polarization image is collected when the polarization axes of the light source and the camera are parallel, and the cross-polarization image is collected when the polarization axes of the light source and the camera are orthogonal.

[0020] Preferably, the constructing of the image pyramid comprises: taking the cross-polarization image as the zeroth layer of the pyramid; iteratively performing Gaussian blur and two-fold downsampling on a current level image to generate a next level image; and repeating the iteration until a preset total number of layers is reached to obtain the image pyramid.

[0021] Preferably, the detecting of the welding stress of the composite plate based on the corrected image comprises: reconstructing the three-dimensional topography of the weld and the heat-affected zone based on the corrected image using a structured light algorithm; and inversely calculating the residual stress distribution of the welding area of the composite plate based on the three-dimensional topography and in combination with a material mechanics model to realize the detection of the welding stress of the composite plate.

[0022] In a second aspect, the application provides a composite plate welding stress detection system, comprising a processor and a memory, and the memory stores computer program instructions which, when executed by the processor, implement the above-mentioned composite plate welding stress detection method.

[0023] By using the above technical solution, the above-mentioned composite plate welding stress detection method is generated into a computer program and stored in the memory to be loaded and executed by the processor, so that a terminal device is manufactured according to the memory and the processor, and convenient use is achieved.

[0024] The application has the following beneficial effects:

[0025] The application obtains reliability weights by dual-polarization imaging to separate specular highlights, and combines multi-resolution analysis reflectance fields to eliminate inherent brightness differences of different materials, thereby solving the problem of coexistence of overexposure and underexposure in the composite plate welding area, and providing high-robustness image data basis for detection in complex optical environments.

[0026] The output corrected image has uniform contrast and average gray scale on different material backgrounds, and the high light pollution area is replaced by reliable background values, thereby providing high-quality and high signal-to-noise ratio input data for subsequent analysis algorithms such as digital image correlation method or structured light method, and ensuring the accuracy of the final stress and strain analysis results.

[0027] The present application combines multi-resolution reflectivity field estimation and reliability weighting, the former preserves the steep boundary between different materials while reducing the background brightness difference, and the latter accurately locates and suppresses the specular high light pollution pixels, thereby ensuring that the optical heterogeneity interference is reduced while the surface detail features such as speckle or stripe are preserved, the integrity of these features is the key to the success of subsequent stress and strain calculation, thereby providing direct data guarantee for high-precision composite plate welding stress detection. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 is a flow chart schematically showing a composite plate welding stress detection method in the present application. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0030] The specific embodiments of the present application will be described in detail below with reference to the drawings.

[0031] The embodiments of the present application disclose a composite plate welding stress detection method, referring to Figure 1 , comprising steps S1-S4:

[0032] S1: An active polarization imaging system is used to acquire co-polarization images and cross-polarization images.

[0033] It should be noted that, in order to achieve the overall goal of eliminating optical heterogeneity in the composite plate welding area of the present application, the present application uses the physical principle that specular reflection, i.e. high light, maintains the polarization state of the light source, while diffuse reflection, i.e. features, randomizes the polarization state, uses an existing active polarization imaging system to obtain two images containing different optical components from the same detection position, and provides the necessary input data for subsequent separation of specular high light and diffuse reflection features.

[0034] Specifically, the present application adopts the prior art active polarization imaging system, which acquires two real-time images of the welding area of the composite panel twice by respectively placing a linear polarizer in front of a light source and a camera lens and controlling the relative angle of the polarization axes of the two; when the polarization axes of the light source and the camera are parallel, a same-direction polarization image is acquired , which retains the specular highlight component to the greatest extent; when the polarization axes of the light source and the camera are orthogonal, an orthogonal polarization image is acquired , which suppresses the specular highlight component to the greatest extent and can be regarded as an approximation of the pure diffuse reflection feature.

[0035] At this point, two original gray-scale images containing different optical components are obtained and , which are the data basis for all subsequent calculations.

[0036] S2: For each pixel point in the image, the specular pollution index of the pixel point is obtained by evaluating the ratio of the gray-scale difference and the gray-scale sum of the same-direction polarization image and the orthogonal polarization image at the pixel point; and the reliability weight of the pixel point is obtained based on the specular pollution index.

[0037] It should be noted that after the same-direction polarization image and the orthogonal polarization image are obtained, since the same-direction polarization image approximately contains the diffuse reflection component and the specular reflection component, and the orthogonal polarization image can be regarded as the best approximation of the diffuse reflection component, the difference between the two is the approximate intensity of the specular reflection component; if this difference is directly used, it will be disturbed by the brightness of the object itself, i.e., the high reflectivity of titanium and the low reflectivity of steel, for example, a weak highlight on a bright diffuse reflector may have the same difference as a strong highlight on a dark diffuse reflector. Therefore, the present application eliminates this influence by obtaining the specular pollution index, so that it only reflects the proportion of the specular reflection in the total reflection, and then calculates an index for accurately evaluating the degree of pollution of each pixel point in the image by the specular highlight, and generates a reliability weight field accordingly. The weight field will be used to dynamically suppress those pixel points that are polluted by highlights and are unreliable in subsequent steps, while retaining those pixel points that are intact and reliable.

[0038] Specifically, for any pixel point in the image, the specular pollution index of the pixel point is calculated as follows:

[0039]

[0040] In the formula, is the specular pollution index of the pixel point ; is the gray-scale value of the pixel point in the same-direction polarization image; is the gray-scale value of the pixel point the gray value in the orthogonal polarization image; is a very small positive number, empirically .

[0041] It is noted that, this term represents the specular reflection, i.e. the highlight intensity, at the pixel point ; This term is related to the total light intensity received by the pixel point , and serves as a normalization denominator to eliminate the influence of the material brightness itself; the output value range of the specular pollution index is theoretically close to the interval of 0 to 1; when the gray value of the co-polarization image is less than the gray value of the orthogonal polarization image, resulting in the index being less than 0, it indicates that there is no specular reflection enhancement in this area, which belongs to the non-highlight area; when a pixel point is a pure diffuse reflector, the gray value of the co-polarization image and the gray value of the orthogonal polarization image are close, resulting in the numerator tending to 0, and the specular pollution index tends to 0; when a pixel point is a pure specular reflector, the gray value of the orthogonal polarization image tends to 0, resulting in the specular pollution index tends to 1; therefore, the physical meaning of the specular pollution index is how much proportion of the signal energy at the pixel point comes from the specular reflection, and the higher the value, the lower the reliability of the diffuse reflection features such as digital image correlation method speckle or structured light stripe contained.

[0042] Further, based on the specular pollution index, the present application defines a reliability weight for representing the data reliability of the pixel point:

[0043]

[0044] wherein, is the reliability weight of the pixel point ; is the specular pollution index of the pixel point ; is the weight sharpening coefficient.

[0045] It is noted that when the weight sharpening coefficient is 1, the weight is linearly decreased; when the weight sharpening coefficient is greater than 1, the attenuation of the weight is nonlinear; the purpose of this nonlinear attenuation is to make the reliability weight insensitive to slight increase of the specular pollution index , i.e. slight pollution, and the weight is still close to 1, but once the specular pollution index exceeds a certain threshold, the reliability weight It will quickly drop to 0; achieve more resolute closing effect to moderate to heavy pollution area, while lenient slightly affect the calculation of polarization imperfection.

[0046] It is necessary to supplement that the weight sharpening coefficient The value range is usually between (1, 3], if the value is too small close to 1, the inhibition effect of moderate pollution is not strong, if the value is too large, it may be too sensitive to very slight noise; In the embodiment of the application, the weight sharpening coefficient 2.0 can achieve a good balance between resolute inhibition of high light and leniency to slight noise.

[0047] S3: Multi-resolution pyramid decomposition and interpolation-only up-sampling reconstruction of the orthogonal polarization image are performed to estimate the reflectivity field; the orthogonal polarization image is divided by the reflectivity field to obtain the normalized feature field.

[0048] It should be noted that for the orthogonal polarization image, different materials such as titanium and steel have different base reflectivity, resulting in that the average brightness of the titanium region in the orthogonal polarization image is much higher than that of the steel region; The purpose of the application is to accurately estimate this spatially slowly changing background brightness determined by the nature of the material.

[0049] It should be further supplemented that the orthogonal polarization image Physically, it can be modeled as a multiplicative model, in which Is the reflectivity field representing the inherent brightness of the material that the application wants to estimate, and Is the high-frequency feature field representing the surface details that the application wants to preserve; traditional image data processing methods, such as using a super large radius Gaussian blur to smooth To estimate This method is effective inside the material, but at the boundary of different materials such as the weld edge, it will produce serious color bleeding and blurring, resulting in that the estimated Is severely distorted at the boundary; Therefore, the application proposes a multi-resolution analysis method, the core of which is to separate information at different scales using a Gaussian pyramid, which can effectively smooth out high-frequency features While completely preserving the steep boundaries between different materials in the low-frequency field .

[0050] Specifically, taking the orthogonal polarization image As input, a Gaussian blur and down-sampling operation is iteratively performed to construct an image pyramid, which will gradually erase high-frequency surface detail features layer by layer until the top image only retains the average brightness information of the material.

[0051] Furthermore, starting with the top-level image of the pyramid, the present invention restores the image to its original size by iteratively performing an upsampling operation that involves only interpolation, thus obtaining the final reflectance field. .

[0052] It should be noted that during the decomposition process, Gaussian blurring and downsampling are efficient low-pass filters that quickly erase high-frequency, small-sized features. When reaching the top layer, the remaining information only represents the average brightness of large areas such as titanium plates, steel plates, and welds. During reconstruction, this invention only uses interpolation for upsampling, preserving the large-scale edges that survived in the low-resolution model—that is, the brightness abrupt changes between materials—so that these boundaries appear in the reconstructed image. It remains clear and is not completely blurred like a single large-radius Gaussian blur.

[0053] S4: Obtain the corrected grayscale value of each pixel based on the value of each pixel in the normalized feature field and the reliability weight corresponding to that pixel; the corrected grayscale values ​​of all pixels form a corrected image; and the welding stress detection of the composite plate is completed based on the corrected image.

[0054] It should be noted that the purpose of this invention is to fuse reliability weights and reflectivity fields to reconstruct a corrected image that can be used for image data processing algorithm analysis. The desired outcome of this invention The image should possess two characteristics: firstly, regardless of whether it is on highly reflective titanium or low-reflective steel, its average background brightness should be normalized to the same level to eliminate... The first is the impact of specular highlights; the second is that in areas contaminated by specular highlights, the features should be completely suppressed and replaced with harmless background values.

[0055] Specifically, the present invention performs feature field normalization, transforming orthogonal polarization images... Divide by the estimated reflectivity field To eliminate differences in background brightness, a normalized feature field is extracted. :

[0056]

[0057] In the formula, For pixels Normalized feature field at the location; For pixels Gray values ​​in orthogonally polarized images; For pixels The reflectivity field at that location; It is the stability constant.

[0058] In the formula, according to The model back-calculates So that regardless Is it high or low? The average values ​​will be normalized to Left and right, while the multiplicative characteristics are preserved; stability constant This setting is to prevent the denominator from being set. The value approaches zero in extremely dark regions, causing computational instability; this value should be a small positive number, for example... Its size is small enough relative to the dynamic range of the image (0-255) to not affect the normalization result, while ensuring the numerical stability of the division.

[0059] Furthermore, the present invention performs reliability-weighted reconstruction, using reliability weights. The result obtained in the previous step Weighting is applied to flatten areas contaminated by highlights, generating a corrected image. :

[0060]

[0061] In the formula, For pixels Correcting the grayscale of the image at that location; For pixels Reliability weight at the location; For pixels The normalized characteristic field at that location.

[0062] It should be noted that, The average value is The normalized feature field is transformed into a zero-mean feature field, whose value is in The values ​​fluctuate in the vicinity, with positive values ​​representing bright speckle and negative values ​​representing dark speckle. Multiplying by the zero-mean characteristic field is to multiply the signal by its corresponding reliability weight; finally, add... It involves shifting the weighted feature field back to... The mean.

[0063] When a pixel is a reliable diffuse point, its specular contamination index Approaching zero, reliability weight As it approaches 1, the formula becomes The result is equal to This indicates that the present invention fully trusts in and retains this characteristic that has been normalized for material brightness; when a pixel is a specular highlight, its specular contamination index... Approaching 1, reliability weight As it approaches zero, the formula becomes The result is equal to That is, neutral background gray, which shows that the application replaces all the unreliable pixels contaminated by specular highlight with neutral background gray, while keeping all the reliable features which have been normalized by material brightness intact.

[0064] Further, based on the corrected image, the three-dimensional topography of the weld and the heat-affected zone is reconstructed by using a structured light algorithm; based on the three-dimensional topography, the residual stress distribution of the composite plate welding area is inversely calculated by combining a material mechanics model, so that the detection of the welding stress of the composite plate is realized.

[0065] The embodiment of the application further discloses a composite plate welding stress detection system, comprising a processor and a memory, and the memory stores computer program instructions, which realize the composite plate welding stress detection method according to the application when executed by the processor.

[0066] The above system further comprises a communication bus and a communication interface and other components well known to those skilled in the art, the setting and functions of which are known in the art, and thus will not be described here.

Claims

1. A method for detecting welding stress in composite plates, characterized in that, include: An active polarization imaging system is used to acquire co-polarized and orthogonal polarized images; For each pixel in the image, the specular contamination index of that pixel is obtained by evaluating the ratio of the gray level difference between the same polarization image and the orthogonal polarization image at that pixel to the total gray level; and the reliability weight of that pixel is obtained based on the specular contamination index. Multi-resolution pyramid decomposition and interpolation-only upsampling reconstruction are performed on the orthogonal polarization image to estimate the reflectance field; the orthogonal polarization image is divided by the reflectance field to obtain the normalized feature field. The corrected grayscale value of each pixel is obtained based on the value of each pixel in the normalized feature field and the reliability weight corresponding to that pixel. The corrected image is composed of the corrected grayscale values ​​of all pixels; Welding stress detection of composite plates was performed based on the corrected images; The mirror pollution index satisfies the following expression: ; For pixels The mirror pollution index at the location; and The coordinate index of the pixel; For pixels Gray values ​​in a co-polarized image; For pixels Gray values ​​in orthogonally polarized images; It is a positive number; The reliability weights satisfy the expression: ; For pixels Reliability weight at the location; and The coordinate index of the pixel; This is the weighted sharpening coefficient; The corrected image satisfies the expression: ; For pixels Correcting the image grayscale at the location; For pixels Reliability weight at the location; For pixels The normalized characteristic field at that location.

2. The method for detecting welding stress in composite plates according to claim 1, characterized in that, The step of performing multi-resolution pyramid decomposition and interpolation-only upsampling reconstruction on orthogonal polarization images to estimate the reflectivity field includes: Using orthogonal polarization images as input, an image pyramid is constructed by iteratively performing Gaussian blur and downsampling operations. Starting from the top image of the image pyramid, the image is restored to its original size by iteratively performing upsampling operations that only involve interpolation, thus obtaining the reflectance field.

3. The method for detecting welding stress in composite plates according to claim 1, characterized in that, The normalized feature field satisfies the expression: ； in, For pixels Normalized feature field at the location; and The coordinate index of the pixel; For pixels Gray values ​​in orthogonally polarized images; For pixels The reflectivity field at that location; It is the stability constant.

4. The method for detecting welding stress in composite plates according to claim 1, characterized in that, The acquisition of co-polarized and orthogonally polarized images includes: Images with the same polarization direction are acquired when the polarization axes of the light source and the camera are parallel; images with orthogonal polarization are acquired when the polarization axes of the light source and the camera are orthogonal.

5. The method for detecting welding stress in composite plates according to claim 2, characterized in that, The construction of the image pyramid includes: The orthogonal polarization image is used as the zeroth layer of the pyramid; Gaussian blur and double downsampling are performed iteratively on the current layer image to generate the next layer image; the iterative operation is repeated until a preset total number of layers is reached to obtain the image pyramid.

6. The method for detecting welding stress in composite plates according to claim 1, characterized in that, The method of detecting welding stress in composite plates based on corrected images includes: Based on the corrected image, the three-dimensional morphology of the weld and heat-affected zone is reconstructed using a structured light algorithm. Based on the three-dimensional morphology and combined with a material mechanics model, the residual stress distribution in the welded area of ​​the composite plate is calculated, thereby enabling the detection of welding stress in the composite plate.

7. A composite plate welding stress detection system, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a composite plate welding stress detection method according to any one of claims 1-6.

Images