Fillet weld identification method for distinguishing pseudo welding spots based on line laser

Abstract

The invention provides a fillet weld identification method for distinguishing pseudo welding spots based on line laser. The fillet weld identification method comprises the following steps: S1, image acquisition and processing: acquiring an image through an industrial gray CCD (Charge Coupled Device) camera, and sequentially performing correction, enhancement, expansion corrosion and Gaussian filtering; s2, light strip central point coarse extraction: extracting a light strip candidate region, screening a central point contour of a non-reflective region, extracting a central point of a reflective region, and combining to form a pixel-level central point; s3, subpixel-level center point fine extraction; and S4, welding seam feature point screening and pseudo welding spot distinguishing: screening wave crest points by calculating the distance from the center point to the datum line, calculating the angles of three adjacent points in combination with the cosine law, distinguishing pseudo welding spots and positioning real welding seam feature points. The problems that in the prior art, due to reflection noise interference of laser light strips, light strip centers are difficult to extract, section brightness distribution is uneven, and false welding spots are prone to being misjudged are solved, and the sub-pixel-level precision of center point extraction and the reliability of welding spot recognition are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of light bar center extraction and fillet weld extraction, and in particular to a fillet weld identification method based on line laser to distinguish pseudo welds. Background Art

[0002] Line structured light sensor technology is a common computer vision technology that captures precise geometric information in three-dimensional scenes. It also offers the advantages of being contactless, fast, and highly accurate. During robot welding, a vision system captures weld seam images, processes them on a computer, and then feeds the data back to the robot, enabling it to identify, track, and weld the seams. The ability to quickly and accurately process images containing the centerline of the light streaks and extract the true weld points directly impacts the quality of the robot's welding.

[0003] During the center point extraction process, the reflection effect caused by the laser light bar hitting the metal surface of the workpiece to be welded and the non-uniform brightness of the light bar cross section increase the difficulty of center point extraction. Its accuracy and precision directly affect the recognition of weld points and are the basis for line laser weld point recognition. The following related methods have shortcomings:

[0004] Center extraction based on the Steger algorithm: Although the method proposed by Li Dongliang et al. in the paper entitled "Based on the Hessian matrix of the linear structured light center line extraction method research" can achieve high sub-pixel accuracy, it requires multiple Gaussian kernel convolution operations, has high computational complexity and slow processing speed, and is difficult to meet the real-time requirements of welding.

[0005] Light stripe center extraction method based on principal component analysis (PCA): This method proposed by Wang Taiyong et al. in the patent with patent number "202010999750.X" relies on multiple iterative refinements to improve the stability of the results. It has a large amount of calculation and does not fully consider the local grayscale distribution non-uniformity. The fitting center position is easily affected by the uneven grayscale distribution, resulting in positioning errors.

[0006] In terms of weld point recognition, Hong Lei et al. published an article titled "Analysis of Weld Stripe Straight Line Feature Extraction Based on Slope Analysis Method" and proposed a laser center point slope analysis method. The center point set is divided and then the weld feature points are obtained by straight line fitting. However, this method ignores the situation where the width of the workpiece to be welded is relatively short. Pseudo-weld points in the image are easily mistaken for real weld points, resulting in weld path positioning deviation. In severe cases, it causes welding operation errors, affects product quality, and even causes welding failure.

[0007] Therefore, the existing technology has problems such as the difficulty in extracting the center of the laser light strip due to the interference of the reflection noise of the laser light strip, the uneven distribution of cross-sectional brightness, and the easy misjudgment of false solder joints. Summary of the Invention

[0008] The present invention provides a fillet weld identification method based on line laser to distinguish pseudo welds to solve the problems in the prior art such as difficulty in extracting the center of the laser light strip due to reflection noise interference, uneven cross-sectional brightness distribution, and easy misjudgment of pseudo welds, thereby improving the sub-pixel accuracy of center point extraction and the reliability of weld point identification.

[0009] To achieve the above object, the present invention adopts the following technical solutions:

[0010] A fillet weld identification method for distinguishing pseudo welds based on line laser includes the following steps:

[0011] S1. Image Acquisition and Processing: An industrial grayscale CCD camera is used to capture weld images with light stripes. The images are then corrected, enhanced, expanded, eroded, Gaussian filtered, and de-noised to obtain a de-noised image.

[0012] S2. Coarse extraction of light streak centers: Dynamic threshold segmentation is performed on the de-noised image to extract candidate light streak regions. Lower and upper limits for pixel width are set and pixel segments are merged. The center point outlines of the non-reflective areas are then filtered out, and the center points of the reflective areas are extracted. Finally, the center point outlines of the non-reflective areas are merged with the center points extracted from the reflective areas to form pixel-level center points.

[0013] S3. Sub-pixel center point extraction: Based on the pixel-level center point obtained in S2, the normal direction is first determined, the light stripe width is estimated, and finally the sub-pixel center point is extracted;

[0014] S4. Weld feature point screening and pseudo-weld point differentiation: Calculate the distance from each sub-pixel center point to the line connecting the first and last points, and smooth the distances to obtain a distance set. Use the cosine theorem to calculate the angles of three adjacent points in the distance set. Filter the first three points with the largest angles. Combine the distance elements to select the minimum and maximum values ​​to differentiate pseudo-weld points. Use the fitted straight line intersections as true weld feature points.

[0015] In this specification, in S1, the dilation erosion selects a convolution kernel of 3×3 to eliminate small holes in the stripe area; the Gaussian filter selects a convolution kernel of 5×5 to remove noise.

[0016] In this specification, in S1, the formula for calculating the grayscale value after image enhancement is: s = cr γ , where r is the input value of the grayscale image, ranging from [0,1], s is the grayscale output value after gamma transformation, c is the grayscale scaling coefficient, and γ is the gamma factor.

[0017] In this specification, in S2, the method of extracting the light stripe candidate area is: setting an initial light stripe segmentation threshold T0=160, segmenting each row of pixels in the image, and for each pixel in each row, if the pixel point is greater than the threshold, it is determined to be a foreground pixel, otherwise it is determined to be a background pixel.

[0018] In this specification, in S2, dynamic threshold segmentation is: if the foreground area with continuity characteristics cannot be segmented under the initial grayscale threshold condition, the threshold is gradually reduced by 30 steps to the minimum threshold T min =100, and image segmentation is re-executed after each round of threshold update.

[0019] In this specification, in S2, the lower limit W of the pixel width is set. min =5, upper limit W max =10; then set a distance threshold T D , used to determine whether adjacent pixel segments should be merged. If the distance between adjacent pixel segments is less than T D , they are considered to belong to the same laser light stripe area and are merged into an overall pixel segment.

[0020] In this specification, in S2, the center point of the reflective area is extracted using the brightness-distance trade-off function and the greedy optimization algorithm; the brightness-distance trade-off function is defined as: W = 10 -(s / β)+d ; Where d represents the Euclidean distance between two points, s represents the grayscale value of the second point, and β represents the angle between the vector between the two points and the reference vector.

[0021] In this manual, the execution process of the greedy optimization algorithm is as follows: use the end point of the previous contour of the reflective area as the starting point of the path, use the starting point of the next center point contour of the reflective area as the end point of the path, and optimize the path selection point by point based on the principle of minimizing the trade-off between the distance traveled and the brightness distance consumed. Repeat this process to gradually construct the entire path.

[0022] In this specification, in S3, the purpose of light strip width estimation is to extract effective width light strips to avoid redundant calculations; by extending along both sides of the normal direction and searching for points with a grayscale value of 25% as the starting point and end point of the normal cross section of the light strip at that location, the effective width of the light strip is calculated.

[0023] In this specification, in S3, based on the Gaussian function model, the LM algorithm is used to extract the sub-pixel center point. In summary, the present invention has at least the following beneficial effects:

[0024] The present invention enhances the contrast between the light strip and the background and reduces some noise in the image by correcting, gamma enhancing, dilating, corroding and filtering the light strip image. Combined with the center point contour screening of the non-reflective area and the reflective area, the center point with pixel-level accuracy is obtained using a brightness-distance trade-off function. The coordinates of the light strip center point are then extracted using an optimized LM fine extraction method based on the normal direction and cross-sectional width of the center point, effectively improving the sub-pixel accuracy and stability of the weld image center point extraction under reflective conditions. The angles of three adjacent center points and the distances from the center point to the straight line are calculated according to the cosine theorem to screen out actual weld points and pseudo weld points, thereby improving the accuracy and stability of weld point extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0026] Figure 1 Schematic diagram of the fillet weld identification method for distinguishing false welds based on line laser in the present invention.

[0027] Figure 2 It is the center point outline extracted from the non-reflective area.

[0028] Figure 3 It is the center point outline extracted using the brightness distance trade-off function in the reflective area.

[0029] Figure 4 is the precisely extracted center point contour.

[0030] Figure 5 Schematic diagram of the processed center point set.

[0031] Figure 6 This is a simplified schematic diagram of the light bar outline.

[0032] Figure 7 Schematic diagram of the angular relationship between three adjacent center points.

[0033] Figure 8 This is a schematic diagram of the filtered weld feature points in the image.

[0034] Figure 9 Schematic diagram of the weld point in the image. DETAILED DESCRIPTION

[0035] Hereinafter, only certain exemplary embodiments are briefly described. As will be appreciated by those skilled in the art, the described embodiments may be modified in various ways without departing from the spirit or scope of the embodiments of the present invention. Therefore, the drawings and description are to be regarded as illustrative in nature and not restrictive.

[0036] The disclosure below provides many different embodiments or examples for implementing different structures of the embodiments of the present invention. In order to simplify the disclosure of the embodiments of the present invention, the components and configurations of specific examples are described below. Of course, these are merely examples and are not intended to limit the embodiments of the present invention. In addition, the embodiments of the present invention may repeat reference numerals and / or reference letters in different examples. Such repetition is for the purpose of simplicity and clarity and does not in itself indicate the relationship between the various embodiments and / or configurations discussed.

[0037] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0038] like Figure 1 As shown, this embodiment provides a fillet weld identification method based on line laser to distinguish pseudo welds, including the following steps:

[0039] S1. Image Acquisition and Processing: An industrial grayscale CCD camera is used to capture weld images with light stripes. The images are then corrected, enhanced, expanded, eroded, Gaussian filtered, and de-noised to obtain a de-noised image.

[0040] S2. Coarse extraction of light streak centers: Dynamic threshold segmentation is performed on the de-noised image to extract candidate light streak regions. Lower and upper limits for pixel width are set and pixel segments are merged. The center point outlines of the non-reflective areas are then filtered out, and the center points of the reflective areas are extracted. Finally, the center point outlines of the non-reflective areas are merged with the center points extracted from the reflective areas to form pixel-level center points.

[0041] S3. Sub-pixel center point extraction: Based on the pixel-level center point obtained in S2, the normal direction is first determined, the light stripe width is estimated, and finally the sub-pixel center point is extracted;

[0042] S4. Weld feature point screening and pseudo-weld point differentiation: Calculate the distance from each sub-pixel center point to the line connecting the first and last points, and smooth the distances to obtain a distance set. Use the cosine theorem to calculate the angles of three adjacent points in the distance set. Filter the first three points with the largest angles. Combine the distance elements to select the minimum and maximum values ​​to differentiate pseudo-weld points. Use the fitted straight line intersections as true weld feature points.

[0043] In some embodiments, step S1: using an industrial grayscale CCD camera, aiming the camera at the laser light stripe hitting the weld area to take a picture and acquire an image of the weld with the light stripe. The image is first corrected and processed by the image processing system in the computer; then the light stripe brightness is enhanced to obtain a clearer light stripe image, which facilitates the subsequent extraction of the light stripe center. The formula for calculating the grayscale value of the enhanced image is as follows:

[0044] s=cr γ ;

[0045] Among them, r is the input value of the grayscale image (the original grayscale value), and its value range is [0,1]; s is the grayscale output value after gamma transformation; c is the grayscale scaling coefficient, which is usually 1; γ is the gamma factor, and its size controls the scaling degree of the entire transformation.

[0046] Then, the enhanced image is first dilated and then eroded, and the convolution kernel is selected as 3×3, which helps to smooth the stripe contour and eliminate the small holes in the stripe area; then Gaussian filtering is performed first, and the convolution kernel is selected as 5×5. The enhanced image is convolved to remove the noise that affects the subsequent center extraction image.

[0047] In some embodiments, by dynamically adjusting the light stripe segmentation threshold and processing the reflection noise, the low brightness, unevenness, and reflection noise problems in the light stripe image can be effectively solved. The following steps are included:

[0048] Step 1: Grayscale threshold segmentation and dynamic adjustment mechanism;

[0049] Set an initial light streak segmentation threshold, T0 = 160, and segment each row of pixels in the image. For each row, if the pixel value is greater than the threshold, it is considered a foreground pixel (possibly a laser streak); otherwise, it is considered a background pixel. This method extracts preliminary light streak candidate regions.

[0050] If the foreground area with continuity characteristics cannot be segmented under the initial grayscale threshold condition, the segmentation threshold is dynamically adjusted using a decreasing strategy. A fixed threshold attenuation step of 30 is set, the current threshold is gradually reduced, and the image segmentation operation is re-executed after each round of update. The current threshold is reduced to the preset minimum allowable threshold lower limit T min = 100, further segmentation attempts will be terminated.

[0051] Step 2: Calculate the width of continuous foreground pixels;

[0052] After completing the foreground extraction based on the grayscale threshold, calculate the width of the continuous foreground pixels in each row and set the lower limit of the width W min =5 and upper limit W max= 10. Only pixel segments that meet the above width constraints are retained, and invalid response areas that are too narrow or too wide are filtered out to remove noise and non-laser light stripe structures. Furthermore, in order to enhance the spatial coherence of foreground extraction, a distance threshold T is set. D , used to determine whether adjacent pixel segments should be merged. If the distance between adjacent pixel segments is less than T D , they are considered to belong to the same laser light stripe area and are merged into an overall pixel segment.

[0053] Step 3: Filter out the center point outline of the non-reflective area;

[0054] Calculate the center point of each row of continuous pixel foregrounds, and then find the center point of each row row by row. The calculation formula of the center point is:

[0055]

[0056] Among them, C x 、C y are the horizontal and vertical coordinates of the center point, n is the number of continuous foreground pixels, (x i ,y i ) is the coordinate of the i-th foreground pixel.

[0057] Determine the relationship between the center point and the center point of the previous row of pixel segments to screen the candidate center point contours. Let f( x,y) As the judgment sign, when f( x,y) =1, the center point belongs to the gth center point contour. The judgment conditions are as follows:

[0058]

[0059] Where N is the number of center point contours.

[0060] Then, traverse from the first center point contour, requiring the length of each contour to be greater than the set threshold contour_thresh=20 to remove short discontinuous contours.

[0061] To find subsequent adjacent candidate contours, the following conditions must be met:

[0062]

[0063] in, is the coordinate of the last point of the current contour, is the coordinate of the first point of the candidate contour;

[0064] If there are multiple adjacent contours, calculate the angle between the current contour and the candidate contour, and select the contour closest in direction and position for connection. The angle calculation formula is as follows:

[0065]

[0066] Select the contour that meets the closest principle; thus, when no more contours that meet the conditions can be found, the screening is completed and all the true center point contours of the non-reflective area are obtained. The center point contours extracted in the non-reflective area are as follows: Figure 2 shown.

[0067] Step 4: Use the brightness-distance trade-off function to perform rough extraction on the center point of the reflective area;

[0068] The brightness-distance trade-off function is used to evaluate the quality of a path or between two points. The brightness-distance trade-off function combines the distance, grayscale value, and the angle between the two points, giving priority to points with close distance and high grayscale to achieve path optimization. The brightness-distance trade-off function is defined as:

[0069] W=10 -(s / β)+d ;

[0070] Where d represents the Euclidean distance between two points, which is calculated as: (x1, y1) and (x2, y2) are the coordinates of two points. The farther the distance, the larger the value of d. s represents the grayscale value of the second point (assuming the image is a single-channel grayscale image). The higher the grayscale value, the brighter the point in the image. β represents the angle between the vector between the two points and the reference vector. The center point contour extracted using the brightness distance trade-off function in the reflective area is as follows: Figure 3 shown.

[0071] Based on this brightness-distance trade-off function, during the actual path construction process, the endpoint of the previous reflective area's contour is used as the path's starting point, and the starting point of the next center point's contour is used as the path's end point. A greedy optimization algorithm optimizes path selection point by point, minimizing the distance traveled and the brightness-distance trade-off. At each step, the greedy optimization algorithm selects the current optimal solution—that is, the point with the smallest brightness-distance trade-off—as the next path point, gradually constructing the entire path.

[0072] Step 5: Perform sub-pixel precision extraction on the center of the light stripe;

[0073] The center point extracted by the trade-off function between the center point outline of the non-reflective area and the brightness distance of the reflective area can only achieve pixel-level accuracy, and further refinement is required using a sub-pixel center extraction method. The following steps are included:

[0074] a: Determine the normal direction;

[0075] The center point of the normal direction to be calculated is recorded as (x i ,y i ), take this point as the center, and extract a number of coarse center point sets N groups of data U(xi ,y i ), i=1,2,...,N. Fit a straight line to the center point of the point set, and finally determine the center point (x i ,y i ). The functional relationship between x and y in the linear equation of the fitted line is:

[0076] y=a+b·x;

[0077] There are two parameters to be determined in the formula, a represents the intercept and b represents the slope. i ,y i ), i=1,2,...,N, fit a straight line, and require the observed value y i The weighted sum of squares of the deviations f(x) is minimized, and the formula is as follows:

[0078]

[0079] The above formula is used to obtain partial derivatives of a and b:

[0080]

[0081] After sorting, we get the system of equations:

[0082] aN+b∑x i =∑y i ;

[0083]

[0084] Solving the above system of equations gives the best estimates of the line parameters a and b as follows:

[0085]

[0086] At this time, the slope of the normal direction of the center point can be obtained It can be seen that the normal equation of the center point to be detected is

[0087] b: light bar width estimation;

[0088] The purpose of estimating the width of the light strip is to extract the effective width of the light strip and avoid redundant calculations. The line width w is estimated based on the relationship between the grayscale values ​​of the light strip cross section in the non-reflective area. The grayscale value of the center point C of the light strip is set to p c , taking point C as the center, extend along the normal direction on both sides and search for about p c The point with gray value of 25% is taken as the starting point A and the ending point B of the normal cross section of the light strip, and the effective light strip sequence L0=x i (i∈[y A ,yB ]), x A ,x B are the horizontal coordinates of the starting point A and the ending point B, respectively. i is the horizontal coordinate of the i-th point in the light strip sequence, then the length of the L0 sequence is the estimated line width w = x B -x A .

[0089] c: Extract the sub-pixel center of the laser stripe;

[0090] In most cases, the laser light strip obeys a Gaussian distribution on the light strip cross section. The grayscale value and the grayscale coordinates are used as observation data in the width interval of the light strip cross section. The constructor model is as follows:

[0091]

[0092] Among them, (x, y) is the coordinate information, θ=(A0,x0,y0,σ x ,σ y ) is a parameter list, A0 is the peak amplitude; x0 is the coordinate of the sub-pixel center of the light bar in the x direction, y0 is the coordinate of the sub-pixel center of the light bar in the y direction, σ x , σ y are the standard deviations in the x and y directions, respectively.

[0093] In order to accurately extract the sub-pixel center position of the laser stripe, the coordinates of each pixel point in a certain section of the laser stripe are recorded as (x i ,y i ), the corresponding gray value is I(x i ,y i ), a parameterized Gaussian function f(x, y; θ) is used to describe the ideal grayscale distribution of the laser stripes in the image.

[0094] The residual of each sampling point is defined as the difference between the model prediction value and the actual image grayscale value, that is:

[0095] r i (θ)=f(x,y;θ)-I(x i ,y i );

[0096] where r i is the residual function, which is used to measure the fitting accuracy. The goal of this method is to adjust the parameter θ so that the sum of squared residuals of all sampling points, i.e., the cost function, reaches the minimum value. The cost function is as follows:

[0097]

[0098] Among them, r(θ) is the residual vector, θ=(A0,x0,y0,σ x ,σ y ) is the parameter list to be optimized, r(θ) T is the rotation vector of r(θ).

[0099] Performing a first-order Taylor expansion on the residual function, it is approximately:

[0100] r(θ+Δθ)≈r(θ)+JΔθ;

[0101] Where J is the Jacobian matrix of the residuals with respect to the parameters The calculation of the second-order derivative is avoided by approximating the Hessian matrix with the Jacobian matrix. In order to improve the stability of the iteration, the damping factor λ is introduced to solve J T The singularity problem of J and the identity matrix I reduce the iteration step size. Therefore, the LM iteration formula can be expressed as follows:

[0102] (J T J+λI)Δθ=-J T r;

[0103] By solving the above formula, we get the parameter increment Δθ, and then update the parameter to the following formula:

[0104] θ (k+1) =θ (k) +Δθ;

[0105] Where k is the current iteration number, and the damping factor λ is a positive real number. When λ≥0, the correct direction of iteration is guaranteed. When λ>0 and approaches 0, the system automatically enters the Gauss-Newton method, and the update formula at this time is approximately:

[0106] θ (k+1) =θ (k) -(J T J) -1 J T r;

[0107] When λ>0 is very large, the formula for gradient descent method is:

[0108] θ (k+1) =θ (k) -(λI) -1 J T r;

[0109] Substitute the new update parameters into the cost function and calculate whether the residual sum of squares of the sampling point is smaller. If it is smaller, accept the updated parameters; otherwise, reject the update and increase the damping factor λ:

[0110] λ=10*λ;

[0111] If the new updated parameters are better, the damping factor λ is reduced:

[0112] λ=0.1*λ;

[0113] In order to avoid the numerical instability problem caused by the infinite decrease of the damping factor, the lower limit threshold λ of the damping factor is set min =10 -8 .

[0114] After changing the parameters, the new parameters θ (k+1) The corresponding residual sum of squares becomes S(θ (k+1) ), if the cost function value change satisfies the set iteration accuracy threshold ε S ,Right now:

[0115] |S(θ (k) )-S(θ (k+1) )|<ε S ;

[0116] At this point, the cost function has basically stopped decreasing, and further iterations are meaningless. It can be considered to have converged, or the iteration process is terminated when the maximum number of iterations is reached. The optimized LM algorithm above finally finds the optimal solution for the parameter θ, and thus the coordinate information (x0, y0) of the sub-pixel center of a certain cross section of the light strip can be obtained. The precisely extracted center point contour is as follows Figure 4 shown.

[0117] In some embodiments, step S4 improves the accuracy of corner weld point positioning and effectively distinguishes actual weld points from pseudo weld point information, thereby avoiding the phenomenon of failure to identify or misidentification of weld points and improving the reliability and robustness of the weld point recognition system.

[0118] The processed center point will be obtained, such as Figure 5 , sorted by coordinates, and the set of its center points is A={(x1,y1),(x2,y2),,,(x n ,y n )}. Then calculate the equation of the line Lb between the first and last coordinate points in set A:

[0119]

[0120] Then calculate the points (x i ,y i ) to the straight line Lb, as follows:

[0121]

[0122] Then all the distances are stored in the set B = {d1, d2, ..., d n} (note that the ordinal numbers of the elements in B correspond to the ordinal numbers of the elements in A).

[0123] First find the element d in set B that satisfies the following conditions p , d p is the distance from point p to line Lb, with the following conditions:

[0124]

[0125] Then take element d p Center, z is the filter window radius, and the surrounding elements are smoothed to obtain the new distance d' p , the formula is as follows:

[0126]

[0127] Where z is the filter window radius, the default size is 3, and σ is the standard deviation.

[0128] Then, all the distances d' p There is a set M, and the coordinates of the center point in the set A corresponding to the set M are as follows Figure 6 As shown in the figure, the outline of the simplified light strip is obtained, which is beneficial to the next step of extracting the weld feature points.

[0129] Only the peaks that are significantly higher than the average level are retained; using the law of cosines, we can find Figure 7 , the angle C between every three adjacent coordinate points p , the formula is as follows:

[0130]

[0131] in,

[0132] The following formula is further obtained:

[0133]

[0134] N p Expressed as the sharpness of point p, angle C p The smaller it is, the more likely point p is to be a peak. Finally, we can get N corresponding to all elements of the set M. p value, then for all N p Sort the values ​​and select the top three N with the largest values p The coordinates P of the center point in A corresponding to the value p1 ,P p2 ,P p3 ,like Figure 8 shown.

[0135] Then select P from set B p1,P p2 ,P p3 The corresponding distance elements are d p1 ,d p2 ,d p3 , then in d p1 ,d p2 ,d p3 Select the minimum value d p2 and the maximum value d p3 , we can get the coordinate points corresponding to the coordinates (x p2 ,y p2 ),(x p3 ,y p3 ). Thus, P p1 ,P p2 ,P p3 The three coordinates are distinguished.

[0136] Finally, using the traditional straight line fitting algorithm, we can fit the straight line equation L1 of all coordinate points between the p2th point and the p3th point in the set A, and the straight line equation L2 of all coordinate points between the p3th point and the nth point. The intersection point p of the straight line L1 and the straight line L2 is O The weld point, such as Figure 9 shown.

[0137] The above embodiments are intended to illustrate the present invention, not to limit the present invention. Therefore, changes in illustrative values ​​or substitutions of equivalent components should still fall within the scope of the present invention.

[0138] From the above detailed description, it will be clear to those skilled in the art that the present invention can indeed achieve the aforementioned objectives and is in compliance with the provisions of the Patent Law.

[0139] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they become aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as covering the preferred embodiments and all changes and modifications that fall within the scope of the invention. The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

[0140] It should be noted that the above description of the relevant processes is for illustration and purpose only and does not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to the processes under the guidance of this specification. However, such modifications and changes are still within the scope of this specification.

[0141] The basic concepts have been described above. It will be apparent to those skilled in the art after reading this application that the above disclosures are merely illustrative and do not constitute limitations on this application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and amendments to this application. Such modifications, improvements, and amendments are suggested in this application and remain within the spirit and scope of the exemplary embodiments of this application.

[0142] At the same time, this application uses specific terms to describe the embodiments of this application. For example, "one embodiment," "an embodiment," and / or "some embodiments" refer to a certain feature, structure, or characteristic related to at least one embodiment of this application. Therefore, it should be emphasized and noted that "one embodiment," "an embodiment," or "an alternative embodiment" mentioned twice or more in different places in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of this application may be appropriately combined.

[0143] In addition, it will be understood by those skilled in the art that various aspects of the present application can be illustrated and described by a number of patentable categories or situations, including any new and useful combination of processes, machines, products or substances, or any new and useful improvements thereto. Therefore, various aspects of the present application can be implemented entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software can all be referred to as "units", "modules" or "systems". In addition, various aspects of the present application can take the form of a computer program product embodied in one or more computer-readable media, wherein computer-readable program code is contained therein.

[0144] The computer program code required for the operation of each part of the application can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, conventional procedural programming languages ​​such as C programming language, VisualBasic, Fortran2103, Perl, COBOL2102, PHP, ABAP, dynamic programming languages ​​such as Python, Ruby and Groovy or other programming languages. The program code can be run completely on the user's computer, or run on the user's computer as an independent software package, or run partly on the user's computer and partly on a remote computer, or run completely on a remote computer or server. In the latter case, the remote computer can be connected to the user's computer through any network form, such as a local area network (LAN) or a wide area network (WAN), or be connected to an external computer (such as by the Internet), or in a cloud computing environment, or used as a service such as software as a service (SaaS).

[0145] In addition, unless expressly stated in the claims, the order of the processing elements and sequences described in this application, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this application. Although the above disclosure discusses some embodiments of the invention that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this application. For example, although the implementation of the various components described above can be embodied in a hardware device, it can also be implemented as a pure software solution, for example, installation on an existing server or mobile device.

[0146] Similarly, it should be noted that in order to simplify the presentation of this disclosure and thereby facilitate understanding of one or more of the invention's embodiments, the foregoing descriptions of the embodiments of this disclosure sometimes combine multiple features into a single embodiment, figure, or description thereof. However, this approach should not be interpreted as reflecting an intention that the claimed subject matter requires more features than expressly recited in each claim. Rather, the subject matter of the invention may possess fewer features than the single embodiment described above.

Claims

1. A fillet weld identification method based on line laser to distinguish pseudo welds, characterized in that: The following steps are involved: S1. Image Acquisition and Processing: An industrial grayscale CCD camera is used to capture weld images with light stripes. The images are then corrected, enhanced, expanded, eroded, Gaussian filtered, and de-noised to obtain a de-noised image. S2. Coarse extraction of light streak centers: Dynamic threshold segmentation is performed on the de-noised image to extract candidate light streak regions. Lower and upper limits for pixel width are set and pixel segments are merged. The center point outlines of the non-reflective areas are then filtered out, and the center points of the reflective areas are extracted. Finally, the center point outlines of the non-reflective areas are merged with the center points extracted from the reflective areas to form pixel-level center points. S3. Sub-pixel center point extraction: Based on the pixel-level center point obtained in S2, the normal direction is first determined, the light stripe width is estimated, and finally the sub-pixel center point is extracted; S4. Weld feature point screening and pseudo-weld point differentiation: Calculate the distance from each sub-pixel center point to the line connecting the first and last points, and smooth the distances to obtain a distance set. Use the cosine theorem to calculate the angles of three adjacent points in the distance set. Filter the first three points with the largest angles. Combine the distance elements to select the minimum and maximum values ​​to differentiate pseudo-weld points. Use the fitted straight line intersections as true weld feature points.

2. The fillet weld identification method based on line laser to distinguish pseudo welds according to claim 1, characterized in that: In S1, the dilation erosion selects a convolution kernel of 3×3 to eliminate small holes in the stripe area; the Gaussian filter selects a convolution kernel of 5×5 to remove noise.

3. The fillet weld identification method based on line laser to distinguish pseudo welds according to claim 1, characterized in that: In S1, the formula for calculating the grayscale value after image enhancement is: s = cr γ , where r is the input value of the grayscale image, ranging from [0,1], s is the grayscale output value after gamma transformation, c is the grayscale scaling coefficient, and γ is the gamma factor.

4. The fillet weld identification method based on line laser to distinguish pseudo welds according to claim 1, characterized in that: In S2, the method of extracting the light stripe candidate area is as follows: an initial light stripe segmentation threshold T0=160 is set, and each row of pixels in the image is segmented. For each pixel in each row, if the pixel point is greater than the threshold, it is determined to be a foreground pixel, otherwise it is determined to be a background pixel.

5. The fillet weld identification method based on line laser to distinguish pseudo welds according to claim 1, characterized in that: In S2, dynamic threshold segmentation is as follows: if the foreground area with continuity characteristics cannot be segmented under the initial gray threshold condition, the threshold is gradually reduced by 30 steps to the minimum threshold T min =100, and image segmentation is re-executed after each round of threshold update.

6. The fillet weld identification method based on line laser to distinguish pseudo welds according to claim 1, characterized in that: In S2, set the lower limit of pixel width W min =5, upper limit W max =10; Then set a distance threshold T D , used to determine whether adjacent pixel segments should be merged. If the distance between adjacent pixel segments is less than T D , they are considered to belong to the same laser light stripe area and are merged into an overall pixel segment.

7. The fillet weld identification method based on line laser to distinguish pseudo welds according to claim 1, characterized in that: In S2, the center point of the reflective area is extracted using the brightness-distance trade-off function and the greedy optimization algorithm; the brightness-distance trade-off function is defined as: W = 10 -(s / β)+d ; Where d represents the Euclidean distance between two points, s represents the grayscale value of the second point, and β represents the angle between the vector between the two points and the reference vector.

8. The fillet weld identification method based on line laser to distinguish pseudo welds according to claim 7, characterized in that: The execution process of the greedy optimization algorithm is as follows: use the end point of the previous contour of the reflective area as the starting point of the path, use the starting point of the next center point contour of the reflective area as the end point of the path, and optimize the path selection point by point based on the principle of minimizing the trade-off between the distance traveled and the brightness distance consumed. Repeat this process to gradually construct the entire path.

9. The fillet weld identification method based on line laser to distinguish pseudo welds according to claim 1, characterized in that: In S3, the purpose of light strip width estimation is to extract the effective width light strip to avoid redundant calculations; the effective width of the light strip is calculated by extending along both sides of the normal direction and searching for points with a grayscale value of 25% as the starting and ending points of the normal cross section of the light strip at that location.

10. The fillet weld identification method based on line laser to distinguish pseudo welds according to claim 1, characterized in that: In S3, based on the Gaussian function model, the LM algorithm is used to extract the sub-pixel center point.

Images