Reflection-noise-resistant fillet weld identification method for robot vision welding

Abstract

The invention provides an anti-reflective noise fillet weld recognition method for robot vision welding. The method comprises the steps that a region of interest is set; solving the sum of gray values of each row of pixels in the RO I, and solving the row with the maximum cumulative sum; searching a row with the minimum gray value accumulated sum; searching a coarse coordinate point of the welding seam; the line with the maximum gray value accumulated sum on the two sides of the weld joint is obtained; respectively solving the initial positions of the light strips; solving a central point through an internal propulsion algorithm; and center fitting and weld joint identification. The method can quickly identify the fillet weld.

Description

Technical Field

[0001] The invention relates to the technical field of welding, and in particular to a fillet weld recognition method for robot vision welding with resistance to reflection noise. Background Art

[0002] Intelligent manufacturing technology is particularly important for the development of the manufacturing industry. Welding technology is an important technology in the manufacturing field. How to achieve automation, efficiency and intelligence in welding is particularly important for the realization of intelligent manufacturing. Welding robots are gradually being used in manufacturing. Adding vision to automatically identify the position of welds to welding robots makes them more automated. Intelligence is a hot topic in the current research of intelligent manufacturing.

[0003] Fillet welds are a type of weld that is widely used in manufacturing. When the welding material that makes up the fillet weld is a highly reflective metal material, the weld image captured by the line structured light vision sensor will have strong reflective noise, especially for fillet welds composed of bright metal surfaces such as stainless steel, aluminum alloy, and titanium alloy. There will be a lot of reflective noise. However, many existing fillet weld recognition methods can only recognize fillet welds without reflective noise, with large center extraction errors, slow speed, and complex methods. Summary of the invention

[0004] The present invention provides a fillet weld recognition method for robot vision welding that is resistant to reflective noise. A line structured light vision sensor composed of a line laser (or line structured light) and a camera is used to collect images of fillet welds irradiated by the line laser. The method is used to quickly process the images and recognize the welds.

[0005] This specification discloses a fillet weld recognition method for robot vision welding with anti-reflective noise, comprising:

[0006] Collect weld images through line structured light vision sensor;

[0007] Set a rectangular region of interest in the weld image. The minimum row and column coordinates of the rectangular region of interest are (x s ,y s );

[0008] Sum each row of pixel grayscale values ​​in the rectangular area of ​​interest and save each row y i The corresponding pixel gray value accumulation sum i , i is the row number in the image, by comparing the grayscale values ​​of each row of pixels and sum i The size of , find the cumulative sum i The maximum value of sum max and the corresponding row position, which is recorded as y1. The row with the largest cumulative sum is the row position with the maximum cumulative sum of the reflected noise intensity and the gray value of the light bar;

[0009] The length interval [y1-row, y1+row] is set by the row position y1 of the maximum sum. This length interval should include the weld position. The value of row is generally one-fourth of the number of image rows. The sum is compared within the length interval. i The size of , find the sum i Minimum sum min And the corresponding row position, which is the rough positioning row position of the fillet weld, and is recorded as y2;

[0010] Set a 5×5 vertical template and search from left to right along row y2 and column 0 of the image to find the two boundary points of the light strip in the triangular area where the light strip and the reflection noise overlap. The triangular area is located at the fillet weld.

[0011] When the average grayscale value of pixels in the template is greater than the set threshold of 200 for the first time, the left boundary of the weld position light bar is found; let the template continue to search along the y2 line to find the right boundary of the light bar. When the average grayscale value of pixels in the template is less than the set threshold of 200 for the first time, the right boundary of the weld position light bar is found. The average value of the coordinates of the left and right boundary points is the rough position coordinate of the weld (x0, y2).

[0012] By sum i Construct a histogram with sum at the weld i Minimum value, sum on both sides of the weld i There is a local maximum for each of them, one of which is the global maximum sum max ; By comparing the size of y1 and y2, determine the sum on the other side of the search weld i Local maximum sum max2 Search direction of the row, find sum max2 and the corresponding row position, which is denoted as y3;

[0013] Along y1 and y3, starting from points (x0, y1) and (x0, y3), search for the position of the light bar from left to right; in this process, set a 5×5 horizontal template. When the average grayscale value of the pixels in the template is greater than the set threshold of 200 for the first time, the positions of the two light bars on both sides of the weld are found, and the light bar centers P1 (x st , y1) and P2(x st , y3) points, these two points are the starting center points of the light strip segments on both sides of the weld;

[0014] Starting from the starting center points P1 and P2, the light bar is internally advanced to extract the center, and the light bar centers of the light bar segments on both sides of the weld are extracted respectively;

[0015] The center points of the two light strips are fitted by the least squares straight line fitting method to find the intersection of the two straight lines, which is the precise position of the fillet weld.

[0016] In this manual, sum i The formula is as follows:

[0017]

[0018] Among them, f(x i ,y j ) is the pixel (x i ,y j ) grayscale value, the minimum row and column coordinates are (x s ,y s ), large row and column coordinates (x n ,y m ).

[0019] In this specification, the methods for extracting the center of the light strip by internal propulsion include positive (forward) internal propulsion and reverse (backward) internal propulsion.

[0020] In this specification, forward internal push center extraction and backward internal push center extraction are performed with P1 and P2 as the starting center points, respectively; when forward internal push is performed, the forward push is performed from the starting center, and only one pixel is pushed forward each time; when backward internal push is performed, the backward push is performed from the starting center, and only one pixel is pushed backward each time. Forward push means pushing along the positive direction of the image y-axis, and backward push means pushing along the negative direction of the image y-axis. When y1 is greater than y3, forward internal push center extraction starts from P1, and backward internal push center extraction starts from P2; when y1 is less than y3, forward internal push center extraction starts from P2, and backward internal push center extraction starts from P1.

[0021] In this manual, it is assumed that the current center point is (x t ,y t ), then the forward point (prediction point) is (x t ,y t +1), then the backward push point (prediction point) is (x t ,y t -1); then, with the advancing point as the center, the grayscale values ​​and coordinates of w pixels to the left and right are counted, and the center is calculated using the grayscale centroid method. Suppose the coordinates of the advancing point (prediction point) are (x, y), then the grayscale value of the pixel is I(x, y), then these grayscale values ​​of w pixels to the left and right in the horizontal direction with (x, y) as the center are counted I(x+k, y), the grayscale value I(x+k, y) is the median of the grayscale value of the pixels in the 3×3 window centered on the pixel point (x+k, y), f(x+k, y) is the grayscale value of the pixel (x+k, y), where k∈[-w, w], and the median formula is:

[0022]

[0023] In this specification, w is 2 / 3 of the light strip width w1. Since the width of the fillet weld light strip is uneven, wide at both ends and thin at the weld, the light strip width changes linearly from the weld to the ends of the light strip, and the weld width gradually increases. This linear change is expressed as:

[0024] w1=k b y+b;

[0025] y is the image row coordinate, k b is the linear variation slope of each line of light strip width, and b is the linear variation intercept. Since the relationship between the weld width on both sides of the weld and the line is opposite, the width of each line of light strip in the non-overlapping area of ​​the reflection noise and light strip on both sides of the weld is counted separately, and the k of the linear variation equation of different light strip widths on both sides of the weld is obtained by least squares fitting. b and b.

[0026] In this specification, the improved Niblack adaptive threshold algorithm calculates the central calculation threshold of the 2w pixel grayscale values, and the threshold is recorded as A. Taking the advancement point as the center, within the 2w pixel length, the pixels (x a ,y) and gray value I i (x a ,y), and calculate the light bar centers of these pixels; I i (x a ,y) is the pixel (x a ,y) is the median grayscale value of the pixels in the 3×3 window centered at;

[0027] The central formula is:

[0028]

[0029] Among them, (U, y) is the new center position of the light bar.

[0030] In this specification, a 1×2w vertical window is set with the predicted center (x, y) of the light bar as the center, and then the average value and mean square error of the pixel grayscale value in the window are calculated; u(x, y) is the average value and t(x, y) is the standard deviation, which are defined as:

[0031]

[0032] The improved Nibrak center threshold A calculation formula is as follows:

[0033] A=(Max+u(x,y)) / 2-0.2t(x,y);

[0034] Where Max is the maximum grayscale value in the window.

[0035] The embodiments of this specification can at least achieve the following beneficial effects:

[0036] The present invention proposes a weld recognition method based on row grayscale and statistical analysis, which can quickly recognize fillet welds with reflective noise. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0038] Figure 1 It is a schematic diagram of the fillet weld recognition method for robot vision welding against reflective noise involved in the present invention.

[0039] Figure 2 It is a schematic diagram of the fillet weld reflection image involved in the present invention.

[0040] Figure 3 It is a schematic diagram of the accumulation and statistics of the grayscale values ​​of pixels in each row of the image involved in the present invention. DETAILED DESCRIPTION

[0041] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, 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 descriptions are considered to be exemplary and non-restrictive in nature.

[0042] The disclosure below provides many different embodiments or examples to implement different structures of the embodiments of the present invention. In order to simplify the disclosure of the embodiments of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the embodiments of the present invention. In addition, the embodiments of the present invention can repeat reference numbers and / or reference letters in different examples, and this repetition is for the purpose of simplification and clarity, which itself does not indicate the relationship between the various embodiments and / or settings discussed.

[0043] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0044] like Figure 1 As shown, this embodiment provides a fillet weld recognition method for robot vision welding with resistance to reflective noise, comprising:

[0045] Collect weld images through line structured light vision sensor;

[0046] According to prior knowledge, a rectangular region of interest (ROI) is set in the weld image, and the region should include the weld and the light strip;

[0047] Sum the grayscale values ​​of each row of pixels in the ROI area, and compare the cumulative sum of the grayscale values ​​of each row of pixels. i The size of , find the cumulative sum i The maximum value of sum max and the corresponding row position, which is recorded as y1. The row with the largest cumulative sum is the row position with the maximum cumulative sum of the reflected noise intensity and the gray value of the light bar;

[0048] The interval [y1-row, y1+row] is set by the row position y1 of the maximum sum above. This interval should include the weld position. The value of row is generally one-fourth of the number of image rows. The sum is compared in this interval. i The size of , find the sum i Minimum sum min And the corresponding row position, which is the rough positioning row position of the fillet weld, and this row is recorded as y2.

[0049] Set a 5×5 vertical template and search from left to right along the image y2 row and 0 column to find the two boundary points of the triangular area where the light strip and the reflective noise overlap. This area is located at the fillet weld. When the average grayscale value of the pixels in the template is greater than the set threshold of 200 for the first time, the left boundary of the light strip at the weld position is found; after finding the left boundary point, let the template continue to search along the y2 row to find the right boundary of the light strip. When the average grayscale value of the pixels in the template is less than the set threshold of 200 for the first time, the right boundary of the light strip at the weld position is found. The average coordinates of the left and right boundary points are the rough position coordinates of the weld (x0, y2);

[0050] By sum i The constructed histogram shows that the weld is sum i Minimum value, sum on both sides of the weld i There is a local maximum for each of them, one of which is the global maximum sum max By comparing the size of y1 and y2, determine the sum of the other side of the search weld i Local maximum sum max2 Search direction of the row, find sum max2 and the corresponding row position, which is denoted as y3;

[0051] Along y1 and y3, starting from (x0, y1) and (x0, y3), search for the position of the light bar from left to right. In this process, set a 5×5 horizontal template. When the average grayscale value of the pixels in the template is greater than the set threshold of 200 for the first time, the positions of the two light bars on both sides of the weld are found. The light bar centers P1 and P2 at these two positions are calculated. These two points are the starting center points of the light bar segments on both sides of the weld.

[0052] Take P1(x st , y1) and P2(x st , y3) The light bar is extracted from the starting center point by internal advancement, and the light bar centers of the light bar segments on both sides of the weld are extracted respectively. The light bar center uses the internal advancement center extraction algorithm, which adopts two types: positive internal advancement and reverse internal advancement. Forward advancement means advancing along the positive direction of the image y axis, and backward advancement means advancing along the negative direction of the image y axis. When y1 is greater than y3, the forward internal advancement center extraction starts from P1, and the backward internal advancement center extraction starts from P2; when y1 is less than y3, the forward internal advancement center extraction starts from P2, and the backward internal advancement center extraction starts from P1.

[0053] The center points of the two light strips are fitted by the least square straight line, and the intersection of the two straight lines is obtained. The intersection is the exact position of the fillet weld.

[0054] In a specific embodiment, the present invention provides a method for identifying fillet welds with anti-reflective noise based on robot vision welding with grayscale and statistical analysis, which can quickly identify fillet welds with reflective noise. Figure 2 As shown, the method includes:

[0055] Region of Interest (ROI) Setting:

[0056] In order to improve the calculation speed, a rectangular region of interest (ROI) is set in the weld image according to prior knowledge. The region should include the weld and the light strip. The ROI is defined as the following equation, where x s ,y s are the minimum row and column coordinates of ROI, respectively. n ,y m are the maximum row and column coordinates of ROI respectively.

[0057]

[0058] Calculate the sum of the grayscale values ​​of each row of pixels in the ROI and find the row with the largest cumulative sum:

[0059] Sum the grayscale values ​​of each row of pixels in the ROI area, and compare the cumulative sum of the grayscale values ​​of each row of pixels. i The size of , find the cumulative sum i The maximum value of summax And the corresponding row position, which is recorded as y1. The row with the largest cumulative sum is the row position with the largest cumulative sum of the reflection noise intensity and the gray value of the light bar. According to the statistical experimental data of hundreds of images, the position of this row in the image is a certain distance away from the row where the weld position is located, just far away from the area with the most serious noise pollution;

[0060]

[0061] Among them, f(x i ,y j ) is the pixel (x i ,y j ) grayscale value, the minimum row and column coordinates are (x s ,y s ), large row and column coordinates (x n ,y m ).

[0062] Find the row with the smallest cumulative grayscale value:

[0063] In order to reduce the amount of calculation and accuracy, a certain length interval is set by the row position y1 of the maximum sum above, and sum is compared within this interval. i The size of , find the sum i Minimum sum min And the corresponding row position, which is the approximate row position of the fillet weld, is recorded as y2. Figure 3 shown.

[0064] Find the coarse coordinate points of the weld:

[0065] Set a 5×5 vertical template, and search from left to right along the image y2 row and 0 column to find the two boundary points of the triangular area where the light strip and the reflection noise overlap. This area is located at the fillet weld. When the average grayscale value of the pixels in the template is greater than the set threshold of 200 for the first time, the left boundary (x1, y2) of the light strip at the weld position is found; after finding the left boundary point, let the template continue to search along the y2 row to find the right boundary of the light strip. When the average grayscale value of the pixels in the template is less than the set threshold of 200 for the first time, the right boundary (x2, y2) of the light strip at the weld position is found. The average value of the coordinates of the left and right boundary points is the rough position coordinate of the weld (x0, y2), where x0 = (x1 + x2) / 2;

[0066] Find the rows with the largest cumulative grayscale values ​​on both sides of the weld:

[0067] By sum i The constructed histogram shows that the weld is sum i Minimum value, sum on both sides of the weld iThere is a local maximum for each of them, one of which is the global maximum sum max By comparing the size of y1 and y2, determine the sum of the other side of the search weld i Local maximum sum max2 Search direction of the row, find sum max2 and the corresponding row position, which is denoted as y3;

[0068] Find the starting position of the light bar respectively:

[0069] Along y1 and y3, starting from (x0, y1) and (x0, y3), search for the position of the light bar from left to right. In this process, a 5×5 horizontal template is set. When the average grayscale value of the pixels in the template is greater than the set threshold of 200 for the first time, the positions of the two light bars on both sides of the weld are found. The light bar centers P1 (x st , y1) and P2(x st , y3) points, these two points are the starting center points of the light bar segments on both sides of the weld.

[0070] The inner boost algorithm finds the center point:

[0071] The forward and backward internal advancement center extraction algorithms are started with the starting center points P1 and P2 respectively. The forward and backward internal advancement center extraction methods have the same principle, but the advancement direction is different. Taking the forward internal advancement as an example, starting from the starting center, advance forward, and advance only one pixel (one row or one column) each time; assuming that the current center point is (x t ,y t ), then the forward point (prediction point) is (x t ,y t +1), then the backward push point (prediction point) is (x t ,y t -1); then, with the advancing point as the center, the grayscale values ​​and coordinates of w pixels to the left and right are counted, and the center is calculated using the grayscale centroid method. Suppose the coordinates of the advancing point (prediction point) are (x, y), then the grayscale value of the advancing point pixel is I(x, y), then these grayscale values ​​of w pixels to the left and right in the horizontal direction with (x, y) as the center are counted I(x+k, y), the grayscale value I(x+k, y) is the median of the grayscale value of the pixels in the 3×3 window centered on the pixel point (x+k, y), f(x+k, y) is the grayscale value of the pixel (x+k, y), where k∈[-w, w], and the median formula is:

[0072]

[0073] This step is to miniaturize and localize the median filter to improve the computational efficiency and noise resistance. w is 2 / 3 of the light strip width w1. Since the width of the fillet weld light strip is uneven, wide at both ends and thin at the weld, the light strip width changes linearly from the weld to the ends of the light strip, and the weld width gradually increases. This linear change can be expressed as:

[0074] w1=k b y+b;

[0075] y is the image row coordinate, k b is the linear variation slope of each line of light strip width, and b is the linear variation intercept. Since the relationship between the weld width on both sides of the weld and the line is opposite, the width of each line of light strip in the area where the reflected noise and light strips do not overlap on both sides of the weld is counted separately, and the k of the linear variation equation of different light strip widths on both sides of the weld is obtained by least squares fitting. b and b.

[0076] The improved Niblack adaptive threshold algorithm is used to calculate the central threshold of the grayscale values ​​of the 2w pixels. The threshold is recorded as A. Taking the advancing point as the center, within the length of 2w pixels, the pixels (x a ,y) and gray value I i (x a ,y),I i (x a ,y) is the pixel (x a ,y) and calculate the center of the light stripe of these pixels; the center formula is:

[0077]

[0078] Among them, (U, y) is the new center position of the light bar.

[0079] The threshold (Nd) calculation formula of the traditional Nibrak method is as follows:

[0080] Nd=u(x,y)+Kt(x,y);

[0081] Among them, u(x,y) is the average grayscale value of a certain number of pixels around the pixel point (x,y), t(x,y) is the local standard deviation of the pixel grayscale value, and K is a coefficient related to the image, which needs to be manually set according to the foreground and background of the image. K is usually set to -0.2 to 0.2.

[0082] In order to improve the robustness and accuracy of light stripe center extraction, more laser stripe details must be considered during center extraction, especially when there are uneven and weak light areas in the light stripe. Here, the traditional Nibrak method is improved to calculate the center extraction threshold so that there is a suitable threshold for calculating the center of each column of the light stripe. This method first sets a 1×2w vertical window centered on the predicted center (x, y) of the light stripe, and then calculates the mean and mean square error of the pixel grayscale values ​​in the window. u(x, y) is the mean and t(x, y) is the standard deviation, which are defined as:

[0083]

[0084]

[0085] The improved Nibrak center threshold A calculation formula is as follows:

[0086] A=(Max+u(x,y)) / 2-0.2t(x,y);

[0087] Where Max is the maximum grayscale value in the window.

[0088] Center fitting and weld identification algorithm:

[0089] The center points of the two segments are fitted by the least square straight line, and the intersection of the two straight lines is obtained. The intersection is the exact position of the fillet weld.

[0090] The above-described embodiments are used to illustrate the present invention, not to limit the present invention, so changes in the exemplified values ​​or replacement of equivalent elements should still fall within the scope of the present invention.

[0091] From the above detailed description, it can be understood by 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.

[0092] Although the 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 are aware of the basic creative concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. It should be pointed out that any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0093] It should be noted that the above description of the relevant process is only for example and explanation, 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 process under the guidance of this specification. However, these modifications and changes are still within the scope of this specification.

[0094] The basic concepts have been described above. Obviously, for those of ordinary skill in the art who have read this application, the above invention disclosure is only for example and does not constitute a limitation of this application. Although not explicitly stated here, those of ordinary skill in the art may make various modifications, improvements and amendments to this application. Such modifications, improvements and amendments are suggested in this application, so such modifications, improvements and amendments still belong to the spirit and scope of the exemplary embodiments of this application.

[0095] At the same time, the present application uses specific words to describe the embodiments of the present 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 the present application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, some features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.

[0096] In addition, it will be appreciated by those skilled in the art that various aspects of the present application may 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 may 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 may all be referred to as "units", "modules" or "systems". In addition, various aspects of the present application may take the form of a computer program product embodied in one or more computer-readable media, wherein computer-readable program code is contained therein.

[0097] 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, etc., such as C programming language, VisualBasic, Fortran2103, Perl, COBOL2102, PHP, ABAP, such as Python, Ruby and Groovy dynamic programming language 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 the remote computer, or run completely on the 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 connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as software as a service (SaaS).

[0098] In addition, unless explicitly 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 invention embodiments 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.

[0099] Similarly, it should be noted that in order to simplify the description disclosed in this application and thus help understand one or more embodiments of the invention, in the above description of the embodiments of the application, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this method of the application should not be interpreted as reflecting the intention that the claimed object requires more features than those explicitly stated in each claim. On the contrary, the subject of the invention should have fewer features than the above single embodiment.

Claims

1. A fillet weld recognition method for robot vision welding with anti-reflective noise, characterized in that: include: Collect weld images through line structured light vision sensor; Set a rectangular region of interest in the weld image. The minimum row and column coordinates of the rectangular region of interest are (x s ,y s ); Sum the grayscale values ​​of each row of pixels in the rectangular area of ​​interest and save each row y i The corresponding pixel gray value accumulation sum i , by comparing the cumulative sum of the grayscale values ​​of each row of pixels i The size of , find the cumulative sum i The maximum value of sum max and the corresponding row position, which is recorded as y1. The row with the largest cumulative sum is the row position with the maximum cumulative sum of the reflected noise intensity and the gray value of the light bar; Set the length interval [y1-row, y1+row] by the row position y1 of the maximum sum. This length interval should include the weld position. The value of row is one-fourth of the number of image rows. Compare sum within the length interval. i The size of , find the sum i Minimum sum min And the corresponding row position, which is the rough positioning row position of the fillet weld, and is recorded as y2; Set a 5×5 vertical template and search from left to right along row y2 and column 0 of the image to find the two boundary points of the light strip in the triangular area where the light strip and the reflection noise overlap. The triangular area is located at the fillet weld. When the average grayscale value of pixels in the template is greater than the set threshold of 200 for the first time, the left boundary of the weld position light bar is found; let the template continue to search along the y2 line to find the right boundary of the light bar. When the average grayscale value of pixels in the template is less than the set threshold of 200 for the first time, the right boundary of the weld position light bar is found. The average value of the coordinates of the left and right boundary points is the rough position coordinate of the weld (x0, y2). By sum i Construct a histogram with sum at the weld i Minimum value, sum on both sides of the weld i There is a local maximum for each of them, one of which is the global maximum sum max ; By comparing the size of y1 and y2, determine the sum on the other side of the search weld i Local maximum sum max2 Search direction of the row, find sum max2 And the corresponding row position coordinates, the row is recorded as y3; Along y1 and y3, starting from (x0, y1) and (x0, y3), search for the position of the light bar from left to right; in this process, set a 5×5 horizontal template. When the average grayscale value of the template pixels is greater than the set threshold of 200 for the first time, the positions of the two light bars on both sides of the weld are found, and the light bar centers P1 (x st , y1) and P2(x st , y3) points, these two points are the starting center points of the light strip segments on both sides of the weld; Starting from the starting center points P1 and P2, the light bar is internally advanced to extract the center, and the light bar centers of the light bar segments on both sides of the weld are extracted respectively; The center points of the two light strips are fitted by the least squares straight line fitting method to find the intersection of the two straight lines, which is the precise position of the fillet weld.

2. The method for identifying fillet welds with robot vision welding and anti-reflective noise according to claim 1 is characterized in that: sum i The formula is as follows: Among them, f(x i ,y j ) is the pixel (x i ,y j ) grayscale value, the minimum row and column coordinates are (x s ,y s ), maximum row and column coordinates (x n ,y m ).

3. The method for identifying fillet welds with robot vision welding and anti-reflective noise according to claim 1 is characterized in that: The methods for extracting the inner center of the light strip are forward pushing and backward inner pushing.

4. The method for identifying fillet welds with robot vision welding and anti-reflective noise according to claim 3 is characterized in that: The forward internal pushing center extraction and the backward internal pushing center extraction are performed with P1 and P2 as the starting center points respectively; when pushing forward internally, the extraction is started from the starting center and pushed forward by only one pixel at a time; when pushing backward internally, the extraction is started from the starting center and pushed backward by only one pixel at a time; pushing forward means pushing along the positive direction of the image y-axis, and pushing backward means pushing along the negative direction of the image y-axis; when y1 is greater than y3, the forward internal pushing center extraction is started from P1, and the backward internal pushing center extraction is started from P2; when y1 is less than y3, the forward internal pushing center extraction is started from P2, and the backward internal pushing center extraction is started from P1.

5. The method for identifying fillet welds with robot vision welding and anti-reflective noise according to claim 4 is characterized in that: Assume that the current center point is (x t ,y t ), then the forward point, i.e. the predicted point, is (x t ,y t +1), then the backward point, i.e. the predicted point, is (x t ,y t -1); then, with the advancing point as the center, the grayscale values ​​and coordinates of w pixels to the left and right are counted respectively, and the center is calculated using the grayscale centroid method; assuming that the coordinates of the advancing point, i.e. the prediction point, are (x, y), the grayscale value of the pixel is I(x, y), then these grayscale values ​​of w pixels to the left and right in the horizontal direction with (x, y) as the center are counted respectively I(x+k, y), and the grayscale value I(x+k, y) is the median of the grayscale values ​​of the pixels in the 3×3 window centered on the pixel point (x+k, y), f(x+k, y) is the grayscale value of the pixel (x+k, y), where k∈[-w, w], and the median formula is:

6. The method for identifying fillet welds with robot vision welding and anti-reflective noise according to claim 5 is characterized in that: w is 2 / 3 of the light strip width w1. Since the width of the fillet weld light strip is uneven, wide at both ends and thin at the weld, the light strip width changes linearly from the weld to the two ends of the light strip, and the weld width gradually increases. This linear change is expressed as: w1=k b y+b; y is the image row coordinate, k b is the linear variation slope of each line of light strip width, and b is the linear variation intercept. Since the relationship between the weld width on both sides of the weld and the line is opposite, the width of each line of light strip in the non-overlapping area of ​​the reflection noise and light strip on both sides of the weld is counted separately, and the k of the linear variation equation of different light strip widths on both sides of the weld is obtained by least squares fitting. b and b.

7. The method for identifying fillet welds with robot vision welding and anti-reflective noise according to claim 6, characterized in that: The improved Niblack adaptive threshold algorithm is used to calculate the central threshold of the grayscale values ​​of the 2w pixels. The threshold is recorded as A. Taking the advancing point as the center, within the length of 2w pixels, the pixels (x a ,y) and gray value I i (x a ,y), and calculate the light bar centers of these pixels; I i (x a ,y) is the pixel (x a ,y) is the median of the grayscale values ​​of the pixels in a 3×3 window centered at; the center formula is: Among them, (U, y) is the new center position of the light bar.

8. The method for identifying fillet welds with robot vision welding and anti-reflective noise according to claim 7, characterized in that: A 1×2w vertical window is set with the predicted center (x, y) of the light bar as the center, and then the mean and mean square error of the pixel grayscale values ​​in the window are calculated; u(x, y) is the mean and t(x, y) is the standard deviation, which are defined as: The improved Nibrak center threshold A calculation formula is as follows: A=(Max+u(x,y)) / 2-0.2t(x,y); Where Max is the maximum grayscale value in the window.

Images