A method and system for rapid weld identification based on vision guidance

Abstract

The present application provides a method and system for rapid weld identification based on visual guidance. Among them, the welding area image is captured in real time by a visual sensor, and the image quality of the welding area image is improved by image enhancement technology; the weld edge of the welding area image is tracked to obtain weld edge information; based on the weld edge information, a multi-scale representation of the weld edge is generated; the multi-scale representation of the weld edge is processed to obtain a geometric consistency verification result; the angle and focal length of the visual sensor are adjusted through a real-time feedback mechanism to optimize the recognition accuracy of the weld edge; according to the multi-scale representation of the weld edge and the geometric consistency verification result, a guidance signal of the weld identification path is output to guide the welding equipment to move according to the weld identification path. The technical solution provided by the present application greatly improves the accuracy and reliability of rapid weld identification.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of visual recognition technology, and in particular to a method and system for rapid weld recognition based on vision guidance. Background Art

[0002] With the development of industrial manufacturing technology, automated welding technology has been widely used. However, in actual welding production, the position and size of the weld often change due to factors such as material thickness and thermal deformation, which requires welding equipment to be able to accurately identify the weld position in real time. Especially in the welding process of large structural parts, such as the automotive manufacturing industry, aerospace field, and bridge construction, the accuracy of weld recognition directly affects the welding quality and production efficiency. Therefore, a technical solution that can achieve high-precision weld recognition under complex working conditions is needed.

[0003] At present, the weld recognition methods on the market mainly include technical means based on laser sensing, ultrasonic sensing and visual sensing. Among them, the method based on visual sensing is becoming more and more popular in industrial applications due to its non-contact measurement characteristics. The existing vision-guided weld recognition technology mainly relies on fixed image acquisition equipment and static image processing algorithms to complete the weld recognition task. Although the existing vision-guided weld recognition technology has realized the weld recognition function to a certain extent, there are still some problems in practical applications. For example, the high temperature generated during the welding process will cause a large amount of smoke and spatter to be generated in the welding area. These factors will seriously interfere with the normal operation of the visual sensor, resulting in a decrease in the recognition accuracy of the weld edge.

[0004] In addition, the existing image processing algorithms perform poorly when dealing with dynamically changing welding environments and are difficult to adjust in real time to adapt to welding requirements under different conditions, thus affecting the reliability and consistency of welding. Therefore, it is urgent to develop a more intelligent, flexible and anti-interference weld recognition method. Summary of the invention

[0005] The embodiments of the present application provide a method and system for rapid weld identification based on vision guidance, so as to solve the problem in the prior art that it is difficult to adjust in real time to adapt to welding requirements under different conditions.

[0006] In a first aspect, an embodiment of the present application provides a method for rapid weld identification based on visual guidance, comprising:

[0007] Capturing a welding area image in real time through a visual sensor, and improving the image quality of the welding area image through image enhancement technology to increase the contrast of the weld edge;

[0008] Dynamically tracking the weld edge of the welding area image using an adaptive contour tracking algorithm to obtain weld edge information, wherein the adaptive contour tracking algorithm is used to automatically adjust the tracking strategy according to smoke and spatter generated during the welding process;

[0009] Based on the weld edge information, a multi-scale analysis method is applied, combined with the acquired weld geometric characteristics, to generate a multi-scale representation of the weld edge;

[0010] Using a preset geometric consistency verification algorithm to process the multi-scale representation of the weld edge to obtain a geometric consistency verification result, wherein the geometric consistency verification algorithm is used to ensure the consistency and continuity of the weld edge;

[0011] adjusting the angle and focal length of the visual sensor through a real-time feedback mechanism to optimize the recognition accuracy of the weld edge;

[0012] On the basis of optimizing the recognition accuracy of the weld edge, according to the multi-scale representation of the weld edge and the geometric consistency verification result, a guidance signal of the weld recognition path is output to guide the welding equipment to move according to the weld recognition path.

[0013] Optionally, the dynamically tracking the weld edge of the welding area image using an adaptive contour tracking algorithm includes:

[0014] According to the brightness distribution and texture characteristics in the welding area image, dynamically adjust the threshold of contour detection to track the edge of the weld;

[0015] The position of the weld edge in the welding area image of the next frame is predicted by evaluating the change rate of the weld edge in real time, and when the weld edge is broken or discontinuous, the gap of the weld edge is automatically filled using a pre-established prediction model to obtain a complete weld edge of the welding area image.

[0016] Optionally, based on the weld edge information, applying a multi-scale analysis method and combining the acquired weld geometric characteristics to generate a multi-scale representation of the weld edge includes:

[0017] Extracting features of the weld edge information at different scales to capture feature information of the weld at different scales;

[0018] By performing multi-resolution decomposition on the welding area image, weld edge images at different resolutions are obtained, and based on the weld edge images at different resolutions, a scale space representation of the weld edge is constructed;

[0019] The feature information at different scales is fused to form the weld edge descriptor;

[0020] analyzing geometrical properties of the weld edge;

[0021] A multi-scale representation of the weld edge is generated according to the feature information at different scales, the weld edge descriptor and the geometric characteristics, and the multi-scale representation is used to provide a detailed description of the weld edge at different scales.

[0022] Optionally, the using a preset geometric consistency verification algorithm to process the multi-scale representation of the weld edge to obtain a geometric consistency verification result includes:

[0023] Based on the multi-scale representation of the weld edge, a correspondence relationship between feature points at different scales is established, wherein the correspondence relationship between feature points includes feature point detection and matching of feature descriptors, wherein the matching is used to find feature points of the weld edge representing the same physical position at different scales, so as to evaluate the geometric consistency of the weld edge at different scales;

[0024] By calculating the relative distance and angle change between the characteristic points on the edge of the weld, determining whether the characteristic points meet the preset geometric consistency requirements;

[0025] After geometric consistency evaluation is performed on the characteristic points on the weld edge, a geometric consistency verification result of the weld edge is generated.

[0026] Optionally, the calculating the relative distance and angle change between the feature points on the weld edge includes:

[0027] The relative distance and angle change between the characteristic points on the weld edge are calculated by the following formula:

[0028] ;

[0029] ;

[0030] in, Indicates the relative distance, Indicates the angle change, and Respectively represent and The local curvature at the location of the feature point, and Indicates The horizontal and vertical coordinates of the feature points, and Indicates The horizontal and vertical coordinates of the feature points.

[0031] Optionally, the method further comprises: correcting the multi-scale representation of the weld edge according to the geometric consistency verification result of the weld edge;

[0032] The step of correcting the multi-scale representation of the weld edge according to the geometric consistency verification result of the weld edge comprises:

[0033] The geometric consistency error is calculated by the following formula, and the multi-scale representation of the weld edge is corrected based on the geometric consistency error:

[0034] ;

[0035] in, represents the geometric consistency error between the weld edge representations at different scales, is the standard deviation of the Gaussian kernel function, used to control the relative distance and angle changes Geometric consistency error The scope of influence, It is a double-layer summation symbol, which means that for all feature points and The sum of the pairs between It is part of the Gaussian kernel function, which is used to calculate the feature points and The relative distance difference between and They are feature points and The original relative distance and the corrected relative distance between and They are feature points and The original angle change and the corrected angle change between is a weight factor used to balance the geometric consistency errors of different feature points Contribution is another weighting factor used to balance the impact of relative distance and angle changes on geometric consistency error.

[0036] Optionally, it also includes:

[0037] In the process of calculating the geometric consistency error, dynamically adjusting the learning rate parameter, and calculating the correction amount of the relative distance between the characteristic points on the edge of the weld and the correction amount of the angle change according to the learning rate parameter;

[0038] Correcting the multi-scale of the weld edge according to the correction amount of the relative distance and the correction amount of the angle change;

[0039] The dynamic adjustment of the learning rate parameter and the calculation of the correction amount of the relative distance between the characteristic points on the edge of the weld and the correction amount of the angle change according to the learning rate parameter are calculated by the following formula:

[0040] ;

[0041] in, At the number of iterations The learning rate parameter at , decreases as the number of iterations increases, and the initial value is , represents the number of iterations, is the decay rate, is the initial learning rate, It is a feature point With feature points The correction amount for the relative distance between is adjusted in each iteration to improve the multi-scale representation of the weld edge. It is a feature point With feature points The correction amount for the angle change between the two is adjusted in each iteration to improve the multi-scale representation of the weld edge.

[0042] In a second aspect, an embodiment of the present application provides a weld rapid identification system based on vision guidance, comprising:

[0043] A capture module, used to capture a welding area image in real time through a visual sensor, and improve the image quality of the welding area image through an image enhancement technology to increase the contrast of the weld edge;

[0044] an acquisition module, used for dynamically tracking the weld edge of the welding area image using an adaptive contour tracking algorithm to obtain weld edge information, wherein the adaptive contour tracking algorithm is used for automatically adjusting the tracking strategy according to the smoke and spatter generated during the welding process;

[0045] A generation module, for generating a multi-scale representation of the weld edge based on the weld edge information by applying a multi-scale analysis method and combining the acquired weld geometric characteristics;

[0046] A processing module, used to process the multi-scale representation of the weld edge using a preset geometric consistency verification algorithm to obtain a geometric consistency verification result, wherein the geometric consistency verification algorithm is used to ensure the consistency and continuity of the weld edge;

[0047] An optimization module, for adjusting the angle and focal length of the visual sensor through a real-time feedback mechanism to optimize the recognition accuracy of the weld edge;

[0048] The output module is used to output a guidance signal of a weld identification path based on optimizing the recognition accuracy of the weld edge and according to the multi-scale representation of the weld edge and the geometric consistency verification result, so as to guide the welding equipment to move according to the weld identification path.

[0049] In a third aspect, an embodiment of the present application provides a computing device, comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a visually guided weld rapid identification method as described in any one of the first aspects.

[0050] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, a method for rapid weld identification based on visual guidance as described in any one of the first aspects is implemented.

[0051] In the embodiment of the present application, the welding area image is captured in real time by a visual sensor, and the image quality of the welding area image is improved by image enhancement technology to improve the contrast of the weld edge; the weld edge of the welding area image is dynamically tracked by an adaptive contour tracking algorithm to obtain weld edge information, and the adaptive contour tracking algorithm is used to automatically adjust the tracking strategy according to the smoke and spatter generated during the welding process; based on the weld edge information, a multi-scale analysis method is applied, combined with the acquired weld geometric characteristics, to generate a multi-scale representation of the weld edge; the multi-scale representation of the weld edge is processed using a preset geometric consistency verification algorithm to obtain a geometric consistency verification result, and the geometric consistency verification algorithm is used to ensure the consistency and continuity of the weld edge; the angle and focal length of the visual sensor are adjusted by a real-time feedback mechanism to optimize the recognition accuracy of the weld edge; on the basis of optimizing the recognition accuracy of the weld edge, according to the multi-scale representation of the weld edge and the geometric consistency verification result, a guidance signal of the weld recognition path is output to guide the welding equipment to move according to the weld recognition path. The technical solution provided by the present application greatly improves the accuracy and reliability of rapid weld recognition.

[0052] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0054] Figure 1 A flowchart of a method for rapid weld identification based on visual guidance provided in an embodiment of the present application;

[0055] Figure 2 A structural schematic diagram of a weld rapid identification system based on vision guidance provided in an embodiment of the present application;

[0056] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0058] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0059] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0060] Figure 1 A flowchart of a method for rapid weld identification based on visual guidance is provided for an embodiment of the present application, such as Figure 1 As shown, the method includes:

[0061] 101. Capturing a welding area image in real time through a visual sensor, and improving the image quality of the welding area image through image enhancement technology to increase the contrast of the weld edge;

[0062] In this step, the image of the welding area is captured in real time by a visual sensor, and the image quality of the image of the welding area is improved by image enhancement technology to improve the contrast of the weld edge. The "visual sensor" referred to here usually refers to a device that can capture and convert into an electrical signal, such as a CCD (Charge-Coupled Device) or CMOS (Complementary Metal-Oxide-Semiconductor) camera. This type of sensor can be used to capture real-time images of the welding area for subsequent processing.

[0063] Image enhancement technology refers to improving the quality of an image through an algorithm to make it more suitable for observation or further processing. Common image enhancement techniques include but are not limited to brightness and contrast adjustment, histogram equalization, sharpening filters, etc. The purpose is to make the key features of the image more prominent, which is convenient for subsequent feature extraction and recognition.

[0064] In the embodiment of the present application, the visual sensor is installed on a welding robot or other automated welding equipment and pointed at the welding area. When welding starts, the visual sensor starts to capture images of the welding area. Then, the captured image is processed using image enhancement technology, such as using histogram equalization technology to increase the overall contrast of the image, making the weld edge more clearly visible. Assuming that the brightness distribution of the original image is relatively uniform, but the weld part is not prominent enough, it can be processed by a histogram equalization algorithm. Specifically, for a grayscale image , and its grayscale range is The histogram equalization step may include calculating a cumulative distribution function (CDF) and mapping the grayscale values ​​of the original image to new grayscale values.

[0065] For example, given an image containing a weld with pixel values ​​ranging from 0 to 255, histogram equalization can be performed using the following steps:

[0066] Calculate the histogram of the original image ,in Indicates grayscale Number of occurrences.

[0067] Calculates the cumulative distribution function , , is a constant to ensure that the grayscale value range after mapping is still between 0 and L-1.

[0068] Apply the cumulative distribution function to each pixel value of the original image to obtain the equalized image .

[0069] In this way, the contrast of the weld edge can be significantly improved, thereby improving the accuracy of subsequent weld identification.

[0070] 102. Dynamically track the weld edge of the welding area image using an adaptive contour tracking algorithm to obtain weld edge information, wherein the adaptive contour tracking algorithm is used to automatically adjust the tracking strategy according to smoke and spatter generated during the welding process;

[0071] Optionally, dynamically tracking the weld edge of the welding area image using an adaptive contour tracking algorithm according to 102 includes: dynamically adjusting the threshold of contour detection to track the weld edge based on the brightness distribution and texture features in the welding area image; predicting the position of the weld edge in the welding area image of the next frame by real-time evaluation of the rate of change of the weld edge, and automatically filling the gap of the weld edge using a pre-established prediction model in the event that the weld edge is broken or discontinuous, so as to obtain a complete weld edge of the welding area image.

[0072] In this step, the weld edge of the welding area image is dynamically tracked using an adaptive contour tracking algorithm to obtain weld edge information. The "adaptive contour tracking algorithm" here is an algorithm that can automatically adjust parameters according to the image content, especially for smoke and spatter interference that may be encountered during welding. This algorithm can dynamically adjust the threshold of contour detection according to the brightness distribution and texture features in the welding area image to ensure that the weld edge can be accurately tracked even in complex environments. In addition, in order to deal with possible breaks or discontinuities at the weld edge, the algorithm will also predict the position of the weld in the next frame of the image by evaluating the rate of change of the weld edge in real time, and use the prediction model to automatically fill the gap at the weld edge.

[0073] In the embodiment of the present application, first, the image of the welding area captured by the visual sensor is input into the adaptive contour tracking algorithm. The algorithm adjusts the threshold of contour detection according to the brightness distribution and texture features in the current frame image. For example, if the brightness of the weld edge in the image is high, the algorithm may lower the threshold to prevent missing part of the weld edge; conversely, if the edge is darker, the threshold is increased to avoid noise interference.

[0074] Secondly, the algorithm tracks the edge of the weld and estimates its rate of change. The rate of change can be estimated by calculating the change in the position of the weld edge between two consecutive frames. Assume that the position of the weld edge in a certain frame is , the predicted position of the next frame is , then the rate of change It can be expressed as:

[0075] ;

[0076] If the edge of the weld is found to be broken or discontinuous during tracking, the algorithm will use the pre-trained prediction model to fill in the gap. For example, suppose in a frame of image, the edge of the weld is at position A fracture has occurred, and the algorithm uses historical data and current image information to predict the fracture location. possible direction and fill the gap to make the weld edge complete.

[0077] In a specific embodiment, suppose that during a welding process, the weld edge in a certain frame image becomes blurred due to the smoke generated by welding. At this time, the adaptive contour tracking algorithm dynamically adjusts the threshold according to the brightness distribution and texture characteristics of other parts of the image to ensure that the main part of the weld edge can still be accurately tracked. At the same time, if the algorithm detects that the weld edge is interrupted in certain areas, it will use the previously established prediction model to fill in these broken areas and generate a complete weld edge contour.

[0078] Through the above steps, even if there are interference factors such as smoke and spatter during the welding process, the system can still effectively track the weld edge and generate accurate weld edge information, thereby providing reliable guidance for subsequent welding operations.

[0079] 103. Based on the weld edge information, a multi-scale analysis method is applied in combination with the acquired weld geometric characteristics to generate a multi-scale representation of the weld edge;

[0080] Optionally, based on the weld edge information in 103, a multi-scale analysis method is applied, combined with the acquired weld geometric characteristics, to generate a multi-scale representation of the weld edge, including: extracting features of the weld edge information at different scales to capture feature information of the weld at different scales; obtaining weld edge images at different resolutions by performing multi-resolution decomposition on the welding area image, and constructing a scale space representation of the weld edge based on the weld edge images at different resolutions; fusing feature information at different scales to form a weld edge descriptor; analyzing the geometric characteristics of the weld edge; generating a multi-scale representation of the weld edge based on the feature information at different scales, the weld edge descriptor and the geometric characteristics, wherein the multi-scale representation is used to provide a detailed description of the weld edge at different scales.

[0081] In this step, the concepts involved include multi-scale analysis methods, feature extraction, multi-resolution decomposition, scale space representation, feature fusion, and generation of weld edge descriptors. These concepts together constitute a systematic process for capturing the feature information of the weld edge at different scale levels and ultimately generating a multi-scale representation of the weld edge.

[0082] Multi-scale analysis method: This is an analysis technique that can capture information about the target object at different levels of detail by extracting features at different scales (or resolutions).

[0083] Feature extraction: Feature extraction of weld edge information at different scales means identifying and extracting feature points or feature areas of the weld edge on images of different resolutions.

[0084] Multi-resolution decomposition: Decompose the original image into images of different resolutions so that the characteristics of the weld edge can be observed at different levels.

[0085] Scale-space representation: The representation constructed by the weld edge images at different resolutions can be understood as a multi-level description that helps capture the local and global features of the weld edge.

[0086] Feature fusion: The feature information at different scales is integrated to form a unified weld edge descriptor to better understand and process the weld edge information.

[0087] Weld edge descriptor: This is a structured representation that integrates feature information at different scales and is used to describe the characteristics of the weld edge.

[0088] Geometric characteristics: refers to the geometric properties of the weld edge, such as length, curvature, direction, etc. This information helps to understand the shape and structure of the weld edge.

[0089] In the embodiment of the present application, it is assumed that we have captured an image of the welding area through a visual sensor and improved the image quality through image enhancement technology. Now we need to apply a multi-scale analysis method to generate a multi-scale representation of the weld edge.

[0090] Example steps:

[0091] Get welding area image: Assume that we have a welding area image with a resolution of 1024x768.

[0092] Multi-resolution decomposition: Decompose the image into multiple resolutions, such as using Wavelet Transform (WT) or other multi-scale analysis techniques, to generate images of different resolutions. For example, we generated 4 images of different resolutions, namely 1024x768, 512x384, 256x192, and 128x96.

[0093] Feature extraction: Feature extraction is performed on the image at each resolution. The SIFT (Scale-Invariant Feature Transform, SIFT) algorithm can be used to detect and describe the key points on the weld edge. Suppose we detect the feature point set on images of different resolutions , , , .

[0094] Scale space representation: Based on the weld edge images at different resolutions, a scale space representation of the weld edge is constructed. This step can be understood as a collection of feature points at different resolutions. , , , Combined into a multi-level representation.

[0095] Feature fusion: Fuse feature information at different scales to form a weld edge descriptor. For example, feature fusion technology (such as feature fusion methods in deep learning) can be used to combine feature information at different scales to form a weld edge descriptor. .

[0096] Analyze geometric properties: Analyze the geometric properties of the weld edge, such as curvature, orientation, etc. This information can be extracted from the feature point sets at different resolutions and used to further optimize the multi-scale representation of the weld edge.

[0097] Generate multi-scale representation: Finally, based on the feature information at different scales, the weld edge descriptor The multi-scale representation of the weld edge is generated based on the geometric characteristics. This multi-scale representation can provide a detailed description of the weld edge at different scales, which is helpful for subsequent weld edge identification and welding path planning.

[0098] Example calculation:

[0099] Assume that the feature point sets at different resolutions are:

[0100] The feature point coordinates set in : ;

[0101] The feature point coordinates set in : ;

[0102] The feature point coordinates set in : ;

[0103] The feature point coordinates set in : ;

[0104] These feature point sets are fused into a weld edge descriptor through feature fusion technology. , and after analyzing the geometric characteristics of the weld edge, a multi-scale representation of the weld edge is generated.

[0105] Through the above steps, we are able to generate a multi-scale representation of the weld edge, which provides detailed information support for subsequent weld edge identification and welding path planning.

[0106] 104. Processing the multi-scale representation of the weld edge using a preset geometric consistency verification algorithm to obtain a geometric consistency verification result, wherein the geometric consistency verification algorithm is used to ensure the consistency and continuity of the weld edge;

[0107] Optionally, the multi-scale representation of the weld edge is processed using a preset geometric consistency verification algorithm according to 104 to obtain a geometric consistency verification result, including: based on the multi-scale representation of the weld edge, establishing a correspondence relationship between feature points at different scales, the establishing of the correspondence relationship between feature points includes feature point detection and matching of feature descriptors, the matching is used to find feature points of the weld edge representing the same physical position at different scales to evaluate the geometric consistency of the weld edge at different scales; by calculating the relative distance and angle change between the feature points on the weld edge to determine whether the feature points meet the preset geometric consistency requirements; after performing geometric consistency evaluation on the feature points on the weld edge, generating a geometric consistency verification result of the weld edge.

[0108] In the embodiment of the present application, the specific steps are as follows:

[0109] The weld edge information is feature extracted at different scales to capture the feature information of the weld at different scales. This means that the weld edge needs to be decomposed at multiple resolutions to extract features at different scales. For example, the weld area image can be decomposed at multiple resolutions using methods such as Wavelet Transform (WT) or Multi-scale Pyramid (MP) to obtain weld edge images at different resolutions.

[0110] Example: Assuming there is a welding area image with a resolution of 1024x768, first perform a four-layer wavelet transform on it to obtain four weld edge images with different resolutions, namely 512x384, 256x192, 128x96 and 64x48. Each layer of decomposition retains the main features of the weld edge, but as the resolution decreases, the features become more abstract.

[0111] The feature information at different scales is fused to form a weld edge descriptor. This step involves how to integrate the feature information at different resolutions to form a comprehensive description of the weld edge. For example, feature fusion (FF) technology can be used to combine feature descriptors at different scales.

[0112] Embodiment: For each weld edge image at each resolution, SIFT (Scale-Invariant Feature Transform, SIFT) feature points can be extracted as feature descriptors, and then SIFT feature descriptors of different scales are fused by weighted averaging to form a comprehensive weld edge descriptor.

[0113] Analyze the geometric characteristics of the weld edge. This step is to evaluate the geometric properties of the weld edge, such as shape and size, to help understand the morphological characteristics of the weld.

[0114] Embodiment: The shape of the weld edge can be described by analyzing the geometric properties of the weld edge such as the curvature change and length at different scales. For example, the curvature of the weld edge at each scale can be calculated and used as a component of the geometric properties.

[0115] A multi-scale representation of the weld edge is generated based on the feature information at different scales, the weld edge descriptor and the geometric characteristics. This step combines the results of the previous steps to form a multi-scale representation that fully describes the weld edge.

[0116] Embodiment: Integrate feature descriptors and geometric characteristics at different scales to form a multi-scale representation. For example, SIFT feature points at different scales and corresponding curvature information can be stored together in a structured data as a multi-scale representation of the weld edge.

[0117] Next, the multi-scale representation of the weld edge is processed using a preset geometric consistency verification algorithm to obtain a geometric consistency verification result. The purpose of the geometric consistency verification algorithm is to ensure the consistency and continuity of the weld edge.

[0118] In the embodiment of the present application, based on the multi-scale representation of the weld edge, a correspondence relationship between feature points at different scales is established. This step involves feature point detection and feature descriptor matching, the purpose of which is to find feature points of the weld edge representing the same physical position at different scales.

[0119] Embodiment: The SIFT feature point detection algorithm is used to detect feature points in the weld edge image at each scale, and then a feature matching algorithm (such as FLANN Fast Library for Approximate Nearest Neighbors, FLANN) is used to match the feature points at different scales to find feature point pairs representing the same physical location.

[0120] By calculating the relative distance and angle change between the characteristic points on the weld edge, it is determined whether the characteristic points meet the preset geometric consistency requirements. This step is to evaluate the consistency of the characteristic points on the weld edge to ensure the continuity and consistency of the weld edge.

[0121] Example: For a set of matching feature point pairs , calculate the relative distance between them and angle changes If these values ​​exceed a preset threshold, the weld edge is considered inconsistent or discontinuous at that location.

[0122] After the geometric consistency evaluation of the characteristic points on the weld edge is performed, the geometric consistency verification result of the weld edge is generated. This step is to generate a final consistency verification report based on the previous evaluation results.

[0123] Example: Assuming that the relative distance and angle change of all feature point pairs have been obtained, the consistency of the weld edge can now be determined based on these data. If the relative distance and angle change of most feature point pairs are within the allowable range, the weld edge can be considered to be consistent and continuous, otherwise further correction is required.

[0124] Through the above steps, the system is able to generate a multi-scale representation of the weld edge and ensure its geometric consistency, thereby providing more accurate guidance for subsequent weld identification and welding operations.

[0125] Optionally, calculating the relative distance and angle change between the feature points on the weld edge according to step 104 includes:

[0126] The relative distance and angle change between the characteristic points on the weld edge are calculated by the following formula:

[0127] ; ;

[0128] in, Indicates the relative distance, Indicates the angle change, and Respectively represent and The local curvature at the location of the feature point, and Indicates The horizontal and vertical coordinates of the feature points, and Indicates The horizontal and vertical coordinates of the feature points.

[0129] In this step, the relative distance and angle change between the feature points on the weld edge are calculated to evaluate the geometric consistency of the weld edge at different scales. Refers to two feature points and The Euclidean distance between them, plus the sum of the squared local curvatures at their respective locations. Describes the change in direction between two points, taking into account the effect of local curvature. Here the local curvature is and It reflects the degree of curvature of the weld edge at that point and can help understand the local shape of the weld edge.

[0130] In the formula and Indicates The horizontal and vertical coordinates of the feature points, and Indicates The horizontal and vertical coordinates of the feature points. These coordinate values ​​reflect the location of the feature points in the image. Formula

[0131] ;

[0132] The straight-line distance between feature points is calculated plus the additional difference caused by curvature, and the formula

[0133] ;

[0134] The angle change between the two feature points is taken into account after the influence of curvature.

[0135] In the embodiment of the present application, it is assumed that several feature points on the edge of the weld have been found by a feature detection algorithm (such as SIFT), and the local curvature of each feature point has been calculated. In order to illustrate how to use the above formula, two feature points are used as an example:

[0136] Given feature points and The coordinates of and , and their local curvatures are and . I hope to calculate the relative distance between these two feature points through the formula and angle changes .

[0137] First calculate the relative distance :

[0138] ;

[0139] Next, calculate the angle change :

[0140] ;

[0141] Through the above calculation, the feature points can be obtained and The relative distance between them is about 36.06 unit lengths, and the angle change is about 54.2 degrees. Such calculations help to evaluate the consistency and continuity of the weld edge, thereby ensuring the quality of the weld.

[0142] 105. Adjusting the angle and focal length of the visual sensor through a real-time feedback mechanism to optimize the recognition accuracy of the weld edge;

[0143] In this step, the angle and focal length of the visual sensor are adjusted through a real-time feedback mechanism to optimize the recognition accuracy of the weld edge. This process involves two key technical points: one is the real-time feedback mechanism, and the other is the angle and focal length adjustment of the visual sensor. The real-time feedback mechanism means that the system can dynamically adjust the parameters of the visual sensor according to the current image quality and the effect of weld edge recognition to achieve the best recognition effect. The angle and focal length adjustment of the visual sensor refers to changing the position (angle) and imaging clarity (focal length) of the sensor according to changes in the welding environment and the needs of weld edge recognition to ensure the optimization of image quality.

[0144] In the embodiments of the present application, an application example of this process can be demonstrated through a specific scenario.

[0145] Assume that there is a visual sensor (Camera) installed on a welding robot, which is equipped with a mechanical structure with adjustable angle and focal length. During the welding process, the visual sensor continuously captures images of the welding area and improves the image quality through image enhancement technology. However, due to the smoke and spatter generated during the welding process, the image quality may be affected, resulting in inaccurate recognition of the weld edge.

[0146] To overcome this problem, a closed-loop control system was designed that can adjust the angle and focal length of the visual sensor in real time according to the effect of weld edge recognition. The specific steps are as follows:

[0147] Image capture and processing: Vision sensors continuously capture images of the welding area and improve image quality through image enhancement technology.

[0148] Weld edge recognition: Use the adaptive contour tracking algorithm mentioned above to dynamically track the weld edge, and apply a multi-scale analysis method based on the weld edge information, combined with the weld geometric characteristics to generate a multi-scale representation of the weld edge.

[0149] Real-time feedback mechanism: During the weld edge recognition process, the system will evaluate the accuracy of the recognition result. If the recognition result is not ideal (for example, the weld edge is not clear enough or has breakpoints), the system will trigger the feedback mechanism to adjust the angle and focal length of the visual sensor.

[0150] Adjust vision sensor parameters: The system adjusts the angle and focus of the vision sensor based on the feedback information. For example, if the edge of the weld appears blurry in the image, the system may adjust the focus to get a clearer image; if the edge of the weld is offset in the image, the system may adjust the angle of the sensor to realign the weld.

[0151] Repeat the above process until the accuracy requirement is met: the system will capture the image again after adjustment and perform weld edge recognition until the recognition result reaches the predetermined accuracy requirement. Example

[0152] Assume that during a welding process, the weld edge appears blurred in the image due to the smoke generated by welding. After the system detects that the accuracy of weld edge recognition is lower than expected, it triggers the feedback mechanism to automatically adjust the focal length of the visual sensor so that the lens focuses on the welding area, thereby improving the image clarity. The focal length change before and after adjustment can be expressed as:

[0153] ;

[0154] in is the focal length before adjustment, is the adjusted focal length, The focal length adjustment amount. Assuming the initial focal length , the system decides to adjust the focal length to .

[0155] At the same time, if the position of the weld edge in the image is offset, the system will also adjust the angle of the visual sensor. Assume that the initial angle of the visual sensor is After the system detects the weld edge deviation, it decides to adjust the sensor angle to .

[0156] Through the above steps, the system can dynamically adjust the angle and focal length of the visual sensor according to the actual situation during the welding process, thereby optimizing the recognition accuracy of the weld edge and ensuring a smooth welding process.

[0157] 106. On the basis of optimizing the recognition accuracy of the weld edge, according to the multi-scale representation of the weld edge and the geometric consistency verification result, a guidance signal of the weld recognition path is output to guide the welding equipment to move according to the weld recognition path.

[0158] In this step, the goal is to optimize the weld edge recognition accuracy and output the guidance signal of the weld recognition path based on the multi-scale representation and geometric consistency verification results of the weld edge. The key to this step is how to convert the previously obtained multi-scale representation and geometric consistency verification results of the weld edge into a specific weld recognition path, and pass this path information to the welding equipment to guide it to move according to these paths.

[0159] The guidance signal of the weld seam identification path is actually a control signal or instruction, which tells the welding equipment how to move to correctly perform the welding operation along the weld seam. These signals usually include but are not limited to the specific path, speed, angle and other information that the welding equipment should follow.

[0160] In the embodiment of the present application, it is assumed that we have completed the multi-scale representation of the weld edge and verified the consistency and continuity of the weld edge through the geometric consistency verification algorithm. Now, we will generate a guidance signal for the weld identification path based on this information.

[0161] The specific implementation steps are as follows:

[0162] Obtaining multi-scale representation of weld edge: We have obtained the representation of weld edge at different scales through multi-scale analysis method, and ensured the consistency and continuity of weld edge through geometric consistency verification algorithm.

[0163] Generate weld identification path: Based on the multi-scale representation of the weld edge, we generate a weld identification path. This path includes a series of coordinate points that the welding equipment needs to follow, and each coordinate point corresponds to a feature point on the weld edge.

[0164] For example, suppose the feature points on the edge of the weld are , then the generated weld identification path can be expressed as .

[0165] Output guidance signals for weld seam identification path: Based on the generated weld seam identification path, the system needs to output corresponding guidance signals to the welding equipment. These signals include path points that the welding equipment needs to follow, speed of each point, angle and other parameters.

[0166] For example, if the welding device is a robotic arm, the output guidance signal may be a data packet containing multiple instructions, each corresponding to a waypoint, specifying how the robotic arm should move to ensure that the welding gun follows the correct path to weld.

[0167] Assume that the guidance signal data packet received by the welding equipment is as follows:

[0168] ;

[0169] in is the first coordinate points, is the speed at which the welding equipment reaches that point, It is the angle or attitude of the welding equipment at that point.

[0170] For example, for the first waypoint , we might have:

[0171] ;

[0172] in is the coordinate point location, is the speed at that point, It is the attitude angle of the welding equipment at this point.

[0173] In this way, the welding equipment can perform precise welding operations along the weld identification path based on the received guidance signal data packet, thereby improving welding quality and efficiency.

[0174] Optionally, the method further comprises: correcting the multi-scale representation of the weld edge according to the geometric consistency verification result of the weld edge;

[0175] The step of correcting the multi-scale representation of the weld edge according to the geometric consistency verification result of the weld edge comprises: calculating a geometric consistency error by the following formula, and correcting the multi-scale representation of the weld edge based on the geometric consistency error:

[0176] ;

[0177] in, represents the geometric consistency error between the weld edge representations at different scales, is the standard deviation of the Gaussian kernel function, used to control the relative distance and angle changes Geometric consistency error The scope of influence, It is a double-layer summation symbol, which means that for all feature points and The sum of the pairs between It is part of the Gaussian kernel function, which is used to calculate the feature points and The relative distance difference between and They are feature points and The original relative distance and the corrected relative distance between and They are feature points and The original angle change and the corrected angle change between is a weight factor used to balance the geometric consistency errors of different feature points Contribution is another weighting factor used to balance the impact of relative distance and angle changes on geometric consistency error.

[0178] In this step, the concepts involved include geometric consistency error Calculation and correction of geometric consistency error It is used to evaluate the consistency between the weld edge representations at different scales. The smaller the error value, the more consistent the representation of the weld edge at different scales.

[0179] In the formula is the standard deviation of the Gaussian kernel function, which determines the relative distance between feature points and angle changes right The scope of influence. Double summation symbol For all feature points and The sum is taken between the pairs. Part of the Gaussian kernel function Based on feature points and The relative distance difference between the two weights their contributions. and They are feature points and The original relative distance and the corrected relative distance between and They are feature points and The original angle change and the corrected angle change between them.

[0180] Weight Factor Used to balance different feature point pairs 's contribution, and is another weighting factor used to balance the relative distance and angle changes impact.

[0181] In the embodiment of the present application, it is assumed that a multi-scale representation of the weld edge has been obtained, and the relative distance and angle change between the feature points have been calculated by a geometric consistency verification algorithm. Now, the multi-scale representation of the weld edge needs to be corrected according to the geometric consistency verification result.

[0182] Assume that at a certain scale, there are two feature points and , their original relative distance Unit length, original angle change Corrected relative distance Unit length, corrected angle change . Set the standard deviation of the Gaussian kernel function , weight factor , weight factor .

[0183] Calculate the geometric consistency error according to the formula :

[0184] ;

[0185] Substitute specific values:

[0186] ;

[0187] Obtained by calculation The multi-scale representation of the weld edge can be corrected based on the error value. It is larger, indicating that there are inconsistencies in the representation of the weld edge at different scales, and the multi-scale representation needs to be adjusted so that For example, it can be gradually adjusted through an iterative optimization algorithm and ,until reach a satisfactory level.

[0188] Through the above steps, the system can verify the results according to the geometric consistency of the weld edge and optimize the multi-scale representation of the weld edge, thereby improving the accuracy and reliability of the weld identification path.

[0189] Optionally, the method further comprises: dynamically adjusting a learning rate parameter in the process of calculating the geometric consistency error, and calculating a correction amount of a relative distance between feature points on the edge of the weld and a correction amount of an angle change according to the learning rate parameter;

[0190] Correcting the multi-scale of the weld edge according to the correction amount of the relative distance and the correction amount of the angle change;

[0191] The dynamic adjustment of the learning rate parameter and the calculation of the correction amount of the relative distance between the characteristic points on the edge of the weld and the correction amount of the angle change according to the learning rate parameter are calculated by the following formula:

[0192] ;

[0193] in, At the number of iterations The learning rate parameter at , decreases as the number of iterations increases, and the initial value is , represents the number of iterations, is the decay rate, is the initial learning rate, It is a feature point With feature points The correction amount for the relative distance between is adjusted in each iteration to improve the multi-scale representation of the weld edge. It is a feature point With feature points The correction amount for the angle change between the two is adjusted in each iteration to improve the multi-scale representation of the weld edge.

[0194] In this step, in the process of calculating the geometric consistency error, the learning rate parameter is dynamically adjusted, and the correction amount of the relative distance between the feature points on the edge of the weld and the correction amount of the angle change are calculated according to the learning rate parameter. This process involves the concept of optimization algorithm in machine learning, especially the gradient descent method, which is a commonly used learning rate adjustment mechanism that minimizes the loss function or error function by iteratively updating parameters.

[0195] Specifically, the learning rate parameter As the number of iterations This helps to converge quickly in the early stage of the optimization process, and slow down the convergence speed when approaching the optimal solution to prevent overfitting. is the learning rate at the beginning of the iteration, Is a decay rate parameter that controls the decay rate of the learning rate.

[0196] In the formula and Represents the feature points and The corrections for the relative distance and angle changes between Relative distance and angle changes By this method, the relative distance and angle changes between feature points can be gradually adjusted in each iteration to improve the multi-scale representation of the weld edge.

[0197] In the embodiment of the present application, it is assumed that there is a welding scene, the system has obtained a multi-scale representation of the weld edge according to the previous steps, and calculated the geometric consistency error Now we need to correct the multi-scale representation of the weld edge by dynamically adjusting the learning rate parameter.

[0198] Given an initial learning rate , attenuation rate , and the current iteration number , start the correction. Assume that in the first iteration, the feature point and The relative distance between Unit length, angle change , the corrected relative distance Unit length, corrected angle change .

[0199] Calculate the learning rate parameter according to the formula :

[0200] ;

[0201] for ,have:

[0202] , ,

[0203] ;

[0204] Next, according to the learning rate parameter Calculate relative distance and angle changes The correction amount and :

[0205] ;

[0206] Assumptions and , then:

[0207] ;

[0208] This means that in the first iteration, the feature points and The relative distance between them needs to be corrected by 0.05 unit length, and the angle change needs to be corrected by 0.02 radians.

[0209] In the next iteration, the learning rate According to the attenuation rate Make adjustments and continue to update the relative distance and angle changes according to the correction amount until the geometric consistency error Reaching a predetermined threshold or convergence.

[0210] Through the above steps, the system can dynamically adjust the learning rate parameters according to the geometric consistency verification results of the weld edge, and calculate the correction amount of the relative distance between the feature points on the weld edge and the correction amount of the angle change according to the learning rate parameters, thereby gradually optimizing the multi-scale representation of the weld edge and improving the accuracy and reliability of the weld identification path.

[0211] Figure 2 A structural schematic diagram of a weld rapid identification system based on visual guidance is provided for an embodiment of the present application, such as Figure 2 As shown, the device comprises:

[0212] A capture module 21 is used to capture the welding area image in real time through a visual sensor, and improve the image quality of the welding area image through image enhancement technology to improve the contrast of the weld edge;

[0213] An acquisition module 22, configured to dynamically track the weld edge of the welding area image using an adaptive contour tracking algorithm to obtain weld edge information, wherein the adaptive contour tracking algorithm is configured to automatically adjust the tracking strategy according to smoke and spatter generated during the welding process;

[0214] A generating module 23, for generating a multi-scale representation of the weld edge based on the weld edge information by applying a multi-scale analysis method in combination with the acquired weld geometric characteristics;

[0215] A processing module 24 is used to process the multi-scale representation of the weld edge using a preset geometric consistency verification algorithm to obtain a geometric consistency verification result, wherein the geometric consistency verification algorithm is used to ensure the consistency and continuity of the weld edge;

[0216] An optimization module 25, for adjusting the angle and focal length of the visual sensor through a real-time feedback mechanism to optimize the recognition accuracy of the weld edge;

[0217] The output module 26 is used to output a guidance signal of the weld identification path based on optimizing the recognition accuracy of the weld edge and according to the multi-scale representation of the weld edge and the geometric consistency verification result, so as to guide the welding equipment to move according to the weld identification path.

[0218] Figure 2 The vision-guided weld rapid identification system can be implemented Figure 1 The implementation principle and technical effect of the method for rapid weld identification based on visual guidance described in the embodiment shown are not repeated here. The specific manner in which each module and unit performs operations in the system for rapid weld identification based on visual guidance in the above embodiment has been described in detail in the embodiment of the method, and will not be described in detail here.

[0219] In one possible design, Figure 2 A vision-guided weld rapid identification system of the illustrated embodiment can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0220] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0221] The processing component 32 is used to: capture the welding area image in real time through a visual sensor, and improve the image quality of the welding area image through image enhancement technology to improve the contrast of the weld edge; dynamically track the weld edge of the welding area image using an adaptive contour tracking algorithm to obtain weld edge information, and the adaptive contour tracking algorithm is used to automatically adjust the tracking strategy according to the smoke and spatter generated during the welding process; based on the weld edge information, a multi-scale analysis method is applied, combined with the acquired weld geometric characteristics, to generate a multi-scale representation of the weld edge; use a preset geometric consistency verification algorithm to process the multi-scale representation of the weld edge to obtain a geometric consistency verification result, and the geometric consistency verification algorithm is used to ensure the consistency and continuity of the weld edge; adjust the angle and focal length of the visual sensor through a real-time feedback mechanism to optimize the recognition accuracy of the weld edge; on the basis of optimizing the recognition accuracy of the weld edge, output a guidance signal for the weld recognition path according to the multi-scale representation of the weld edge and the geometric consistency verification result, so as to guide the welding equipment to move according to the weld recognition path.

[0222] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0223] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0224] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0225] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0226] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0227] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0228] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment is a method for rapid weld identification based on vision guidance.

[0229] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0230] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0231] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0232] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for rapid weld identification based on visual guidance, characterized in that: include: Capturing a welding area image in real time through a visual sensor, and improving the image quality of the welding area image through image enhancement technology to increase the contrast of the weld edge; Dynamically tracking the weld edge of the welding area image using an adaptive contour tracking algorithm to obtain weld edge information, wherein the adaptive contour tracking algorithm is used to automatically adjust the tracking strategy according to smoke and spatter generated during the welding process; Based on the weld edge information, a multi-scale analysis method is applied, combined with the acquired weld geometric characteristics, to generate a multi-scale representation of the weld edge; Using a preset geometric consistency verification algorithm to process the multi-scale representation of the weld edge to obtain a geometric consistency verification result, wherein the geometric consistency verification algorithm is used to ensure the consistency and continuity of the weld edge; adjusting the angle and focal length of the visual sensor through a real-time feedback mechanism to optimize the recognition accuracy of the weld edge; On the basis of optimizing the recognition accuracy of the weld edge, outputting a guidance signal of a weld recognition path according to the multi-scale representation of the weld edge and the geometric consistency verification result, so as to guide the welding equipment to move according to the weld recognition path; The method of using a preset geometric consistency verification algorithm to process the multi-scale representation of the weld edge to obtain a geometric consistency verification result includes: Based on the multi-scale representation of the weld edge, a correspondence relationship between feature points at different scales is established, wherein the correspondence relationship between feature points includes feature point detection and matching of feature descriptors, wherein the matching is used to find feature points of the weld edge representing the same physical position at different scales, so as to evaluate the geometric consistency of the weld edge at different scales; By calculating the relative distance and angle change between the characteristic points on the edge of the weld, determining whether the characteristic points meet the preset geometric consistency requirements; After performing geometric consistency evaluation on the characteristic points on the weld edge, a geometric consistency verification result of the weld edge is generated; The step of calculating the relative distance and angle change between the characteristic points on the edge of the weld includes: The relative distance and angle change between the characteristic points on the weld edge are calculated by the following formula: ； in, Indicates the relative distance, Indicates the angle change, and Respectively represent and The local curvature at the location of the feature point, and Indicates The horizontal and vertical coordinates of the feature points, and Indicates The horizontal and vertical coordinates of the feature points.

2. The method according to claim 1, characterized in that The method of dynamically tracking the weld edge of the welding area image using an adaptive contour tracking algorithm includes: According to the brightness distribution and texture characteristics in the welding area image, dynamically adjust the threshold of contour detection to track the edge of the weld; The position of the weld edge in the welding area image of the next frame is predicted by evaluating the change rate of the weld edge in real time, and when the weld edge is broken or discontinuous, the gap of the weld edge is automatically filled using a pre-established prediction model to obtain a complete weld edge of the welding area image.

3. The method according to claim 1, characterized in that The method of applying a multi-scale analysis method based on the weld edge information and combining the acquired weld geometric characteristics to generate a multi-scale representation of the weld edge includes: Extracting features of the weld edge information at different scales to capture feature information of the weld at different scales; By performing multi-resolution decomposition on the welding area image, weld edge images at different resolutions are obtained, and based on the weld edge images at different resolutions, a scale space representation of the weld edge is constructed; The feature information at different scales is fused to form the weld edge descriptor; analyzing geometrical properties of the weld edge; A multi-scale representation of the weld edge is generated according to the feature information at different scales, the weld edge descriptor and the geometric characteristics, and the multi-scale representation is used to provide a detailed description of the weld edge at different scales.

4. The method according to claim 1, characterized in that: Also includes: According to the geometric consistency verification result of the weld edge, the multi-scale representation of the weld edge is corrected; The step of correcting the multi-scale representation of the weld edge according to the geometric consistency verification result of the weld edge comprises: The geometric consistency error is calculated by the following formula, and the multi-scale representation of the weld edge is corrected based on the geometric consistency error: ； in, represents the geometric consistency error between the weld edge representations at different scales, is the standard deviation of the Gaussian kernel function, used to control the relative distance and angle changes Geometric consistency error The scope of influence, It is a double-layer summation symbol, which means that for all feature points and The sum of the pairs between It is part of the Gaussian kernel function, which is used to calculate the feature points and The relative distance difference between and The feature points are and The original relative distance and the corrected relative distance between and They are feature points and The original angle change and the corrected angle change between is a weight factor used to balance the geometric consistency errors of different feature points Contribution is another weighting factor used to balance the impact of relative distance and angle changes on geometric consistency error.

5. The method according to claim 4, characterized in that Also includes: In the process of calculating the geometric consistency error, dynamically adjusting the learning rate parameter, and calculating the correction amount of the relative distance between the characteristic points on the edge of the weld and the correction amount of the angle change according to the learning rate parameter; Correcting the multi-scale of the weld edge according to the correction amount of the relative distance and the correction amount of the angle change; The dynamic adjustment of the learning rate parameter and the calculation of the correction amount of the relative distance between the characteristic points on the edge of the weld and the correction amount of the angle change according to the learning rate parameter are calculated by the following formula: ； in, At the number of iterations The learning rate parameter at , decreases as the number of iterations increases, and the initial value is , represents the number of iterations, is the decay rate, is the initial learning rate, It is a feature point With feature points The correction amount for the relative distance between is adjusted in each iteration to improve the multi-scale representation of the weld edge. It is a feature point With feature points The correction amount for the angle change between the two is adjusted in each iteration to improve the multi-scale representation of the weld edge.

6. A vision-guided weld rapid identification system, used to execute the vision-guided weld rapid identification method according to any one of claims 1 to 5, characterized in that: include: A capture module, used to capture a welding area image in real time through a visual sensor, and improve the image quality of the welding area image through an image enhancement technology to increase the contrast of the weld edge; an acquisition module, used for dynamically tracking the weld edge of the welding area image using an adaptive contour tracking algorithm to obtain weld edge information, wherein the adaptive contour tracking algorithm is used for automatically adjusting the tracking strategy according to the smoke and spatter generated during the welding process; A generation module, for generating a multi-scale representation of the weld edge based on the weld edge information by applying a multi-scale analysis method and combining the acquired weld geometric characteristics; A processing module, used to process the multi-scale representation of the weld edge using a preset geometric consistency verification algorithm to obtain a geometric consistency verification result, wherein the geometric consistency verification algorithm is used to ensure the consistency and continuity of the weld edge; An optimization module, for adjusting the angle and focal length of the visual sensor through a real-time feedback mechanism to optimize the recognition accuracy of the weld edge; The output module is used to output a guidance signal of a weld identification path based on optimizing the recognition accuracy of the weld edge and according to the multi-scale representation of the weld edge and the geometric consistency verification result, so as to guide the welding equipment to move according to the weld identification path.

7. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a vision-guided weld rapid identification method as described in any one of claims 1 to 5.

8. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, a method for rapid weld identification based on vision guidance as described in any one of claims 1 to 5 is implemented.

Images