Visual inspection method and system for automobile welding production line

By applying frame segmentation and image processing technology to the video data of automobile welding production line, weld boundaries and abnormal temperature distribution are identified, combined with timestamp index analysis, the problem of insufficient identification and delay in complex scenarios in the existing technology is solved, and the problem of improvement of the problem of improvement in the identification of complex scenarios in the existing technology is improved. Welding quality and production efficiency.

CN119035863AActive Publication Date: 2024-11-29GUANGZHOU FUJI ASSEMBLY LINE AUTO MFG CO LTD

Patent Information

Application Number
CN202411537129.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-31
Publication Date
2024-11-29
Estimated Expiration
2044-10-31

AI Technical Summary

Technical Problem

The existing visual inspection technology is not robust enough when dealing with complex or varied production scenarios, especially under light changes or background interference, which leads to frequent error identification or missed detection, and is inspected on the automotive welding production line with extremely high degree of automation. Delays lead to reduced productivity and increased costs, with early identification and precise positioning of welding defects limitations.

Method used

By segmenting the video data of the automobile welding production line, and extracting key visual information frame by frame, adjusting contrast and brightness using image processing technology, identifying weld boundaries and performing image binarization processing, analyzing pixel intensity to identify abnormal temperature distribution, combining time Stamp index analysis of key time points, classify welding quality and adjust welding parameters.

Benefits of technology

It improves the accuracy of weld boundary recognition and the recognition accuracy of abnormal temperature distribution, captures the abnormal occurrence time, optimizes welding quality classification and parameter adjustment, and improves production efficiency and welding quality standards.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119035863A_ABST
    Figure CN119035863A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of visual inspection, in particular to a visual inspection method and system for an automobile welding production line, and the method comprises the following steps: carrying out the frame segmentation of an original image sequence based on the video data of the automobile welding production line, extracting key visual information frame by frame, adjusting the contrast and brightness through an image processing technology, and generating a serialized image frame. According to the invention, by using an image processing technology, such as contrast and brightness adjustment, a clearer serialized image frame can be generated, the recognition accuracy of the weld joint boundary is improved, image binarization processing is executed, the boundary of the weld joint and the background is defined, the visualization of the weld joint boundary in the image is enhanced, the recognition of abnormal temperature distribution is more accurate, and the recognition accuracy of the abnormal temperature distribution is improved. The internal non-uniformity is identified through pixel intensity analysis, the system can more effectively mark and track abnormal points, the positioning precision of abnormal temperature welding seams is improved, the production efficiency is optimized, and the overall standard of the welding quality is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of visual inspection technology, and in particular to a visual inspection method and system for an automobile welding production line. Background Art

[0002] Visual inspection technology is a key technology in industrial automation, widely used for quality control and inspection in the manufacturing process. It primarily relies on high-resolution cameras, image processing software, and machine learning algorithms to capture images of products or components and then analyze them in real time to identify defects, measure dimensions, or confirm assembly integrity. Visual inspection systems can improve inspection speed and accuracy, reduce labor costs, and enhance overall production efficiency. With advances in artificial intelligence and machine learning, the intelligence level of visual inspection systems is constantly increasing, enabling them to handle more complex inspection tasks and provide more precise analysis results.

[0003] The automotive welding production line visual inspection method uses visual inspection technology to ensure product quality and safety during the automotive manufacturing process, particularly during the welding and assembly stages. Specially designed cameras and image processing systems monitor weld points and assembly quality, detecting weld defects such as cracks, holes, and uneven welds in real time. Its primary purpose is to improve production line automation, reduce the flow of defective products, ensure the structural strength and durability of automotive components, and enhance product safety and reliability. This allows manufacturers to identify problems early in the production process, avoiding costly rework or recalls.

[0004] Although existing visual inspection technology is widely used in the field of industrial automation, especially playing a key role in quality control and inspection, it still shows some shortcomings when dealing with complex or changing production scenarios. For example, conventional visual inspection systems are not robust enough in image processing under lighting changes or background interference. In other words, they are not stable and reliable in the face of uncertainty and changes, resulting in frequent problems of misidentification or missed detection. Traditional systems require high-performance processor support when analyzing high-speed moving objects in real time, which increases costs. On highly automated automotive welding production lines, even minor inspection delays can lead to reduced production efficiency and increased costs. Existing technologies are still limited in the early identification and precise positioning of welding defects, which to some extent slows down the fault response speed and maintenance efficiency of the production line. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a visual inspection method and system for an automobile welding production line.

[0006] In order to achieve the above object, the present invention adopts the following technical solution, a method for visual inspection of an automobile welding production line, comprising the following steps: S1: Based on the video data of the automobile welding production line, the original image sequence is segmented, key visual information is extracted frame by frame, and contrast and brightness are adjusted using image processing technology to generate serialized image frames; S2: using the serialized image frames, identifying the weld boundary of the automobile welding assembly, performing image binarization processing to distinguish the weld from the background, and obtaining a weld boundary image; S3: Analyzing the unevenness of the interior of the automobile weldment using pixel intensity through the weld boundary image, identifying abnormal temperature distribution, collecting and marking abnormal features, and tracking and locating abnormal points to generate a calibrated abnormal weld; S4: Using the above-described calibration of abnormal welds, analyzing key time points in the automobile welding production line using a timestamp index, comparing weld quality changes in different time periods, recording the time when the abnormality occurred, and obtaining a time positioning result; S5: Analyze the quality of the automobile welding assembly based on the time positioning result, classify the welding quality according to the shape and heat distribution of the weld, grade the automobile welding assembly according to the quality index, and generate a welding quality grade; S6: Using the welding quality level, adjust and optimize welding parameters, update the current and speed of the welding machines on the production line, match differentiated welding requirements, and obtain a production parameter adjustment plan.

[0007] As a further solution of the present invention, the serialized image frame includes an image with contrast adjustment, an image with frame brightness adjustment, and key visual information of each frame; the weld boundary image includes an image distinguishing the weld boundary and an image comparing the background and the weld; the calibrated abnormal weld includes a weld image marked with abnormal temperature distribution, an analysis image of inhomogeneity within the weld, and an abnormal point tracking and positioning map; the time positioning result includes a timestamp index of a key time point and a time period record of weld quality changes; the welding quality grade includes a weld shape classification result map, a thermal distribution analysis result map, and a welding quality grading standard; the production parameter adjustment plan includes an updated current setting parameter, an adjusted welding speed parameter, and a differentiated welding demand matching plan.

[0008] As a further solution of the present invention, based on the video data of the automobile welding production line, the original image sequence is frame segmented, key visual information is extracted frame by frame, and contrast and brightness are adjusted using image processing technology to generate serialized image frames. The specific steps are as follows: S101: Perform frame segmentation processing on the video data of the automobile welding production line, extract a target number of frames per second at a fixed time interval, verify the continuity and integrity of the data, and generate a frame sequence image; S102: using the frame sequence images, extracting coordinates of the welding points and the contact points of the automobile body in each frame, distinguishing key visual areas through color filtering, and marking visual information to obtain a key visual frame sequence; S103: adjusting the contrast and brightness parameters of each frame of the image through the key visual frame sequence, respectively increasing the brightness and contrast by target amounts, optimizing the image quality and visual effects, and generating a sequenced image frame.

[0009] As a further solution of the present invention, the steps of using the serialized image frames to identify the weld boundary of the automobile welding assembly, performing image binarization processing to distinguish the weld from the background, and obtaining the weld boundary image are specifically as follows: S201: using the serialized image frames, adjusting the image to a uniform grayscale range based on the contrast difference between the weld and non-weld areas of the automobile welding production line, verifying the visualization of the weld features in the image, and obtaining a weld highlight image; S202: Based on the weld protrusion image, a grayscale threshold is set to perform image binarization, and the weld area is separated from the non-weld area by optimizing the contrast between the weld boundary and the background to obtain a background-differentiated image; S203: Using the background to distinguish the image, identify and extract the position of the weld boundary, optimize the continuity and clarity of the boundary line, evaluate and verify the visual recognition effect of the boundary line, and generate a weld boundary image.

[0010] As a further solution of the present invention, the steps of analyzing the unevenness of the interior of the automobile weldment using pixel intensity through the weld boundary image, identifying abnormal temperature distribution, collecting and marking abnormal features, and tracking and locating abnormal points to generate a calibrated abnormal weld are specifically as follows: S301: Using the weld boundary image, evaluating the internal non-uniformity of the weld by analyzing pixel intensities, recording areas where pixel intensities vary, and marking potential non-uniform areas to obtain a non-uniform weld analysis diagram; S302: Based on the uneven weld analysis diagram, edge detection technology is used to identify abnormal temperature areas in the automobile welding production line, record areas that deviate from the standard weld temperature distribution, and process the abnormal temperature distribution to obtain an abnormal temperature weld diagram; S303: Using the abnormal temperature weld map, mark the location of abnormal temperature distribution, track and locate the abnormal point, associate the mark with the real-time location of the abnormal area, and generate a calibrated abnormal weld.

[0011] As a further solution of the present invention, the formula of the edge detection technology is as follows: in, is the gradient amplitude, and are the gradients of the image in the horizontal and vertical directions, and are the weighting coefficients in the horizontal and vertical directions, is the weighting coefficient of the diagonal gradient, is the gradient in the diagonal direction of the image, is the Gaussian blur coefficient, is the standard deviation, is the base of the natural exponential.

[0012] As a further solution of the present invention, the steps of using the above-described abnormal weld calibration to analyze key time points in the automobile welding production line using a timestamp index, comparing weld quality changes in different time periods, and recording the time when the abnormality occurred to obtain a time positioning result are as follows: S401: using the calibration method for abnormal welds, assigning a timestamp to each abnormal weld, synchronizing the timestamp with the log of the welding production line, identifying and extracting the time when the abnormality occurred, and obtaining an abnormal timestamp index map; S402: Analyze the weld quality near the key time point based on the abnormal time stamp index map, compare the weld image recognition quality change trend in the differentiated time period, analyze the time interval of the weld quality change, and obtain a time-differentiated weld quality map; S403: Record the time of each quality abnormality through the time-differentiated weld quality map, compare it with the history of the automobile welding production line, mark the time when the abnormality occurred, and generate a time positioning result.

[0013] As a further embodiment of the present invention, based on the time positioning results, the quality of the automobile welding assembly is analyzed, the welding quality is classified according to the shape and heat distribution of the weld, and the automobile welding assembly is graded according to the quality indicators. The steps of generating the welding quality grade are specifically as follows: S501: Analyze the welding quality at the target time point using the time positioning result, evaluate the consistency with the standard quality by monitoring the shape and heat distribution of the weld, and obtain a weld shape and heat distribution analysis diagram; S502: Based on the weld shape and heat distribution analysis diagram, the weld is classified using preset quality evaluation indicators, including weld integrity, uniformity, and heat distribution regularity, to obtain a classified welding quality diagram; S503: Using the classified welding quality map, the automobile welding assembly is graded according to quality indicators, and quality priorities are sorted. It is verified that each level meets the associated quality standards, and a welding quality grade is generated.

[0014] As a further solution of the present invention, the welding quality level is used to adjust and optimize welding parameters, and the current and speed of welding machines on the production line are updated to match differentiated welding requirements. The steps of obtaining a production parameter adjustment solution are specifically as follows: S601: Analyzing welding quality data under differentiated levels using the welding quality levels, identifying welding parameters that need to be adjusted, including current and speed, formulating adjustment requirements for each quality level, and obtaining a welding parameter requirement analysis table; S602: Based on the welding parameter demand analysis table, set a current and speed adjustment plan, with the adjustment range based on the welding quality level requirements, verify that the welding parameters of each level match the quality standards, and obtain a draft welding parameter adjustment plan; S603: The current and speed of the welding machines on the production line are updated through the draft welding parameter adjustment plan, and the production parameter adjustment plan is generated by adjusting the settings of the differentiated welding sites to match the real-time changing welding requirements.

[0015] A visual inspection system for an automobile welding production line, the visual inspection system for an automobile welding production line being used to execute the above-mentioned visual inspection method for an automobile welding production line, the system comprising: The frame segmentation module segments the original image sequence based on the video data of the automobile welding production line, extracts visual information frame by frame, and adjusts the contrast and brightness of each frame to generate serialized image frames; The boundary recognition module identifies the weld and the background based on the serialized image frames, performs image binarization processing, and generates a weld boundary image; The abnormal temperature analysis module uses the weld boundary image to analyze pixel intensity, identify abnormal temperature distribution, collect and mark abnormal features, and track and locate abnormal points to generate a calibrated abnormal weld; The timestamp index module analyzes the key time points in the automobile welding production line by calibrating the abnormal welds, compares the changes in weld quality, records the time when the abnormality occurs, and obtains the time positioning result; The welding quality classification module analyzes the weld shape and heat distribution according to the time positioning results, classifies the welding quality, adjusts and optimizes the welding parameters according to differentiated welding requirements, and generates a production parameter adjustment plan.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by performing frame segmentation and frame-by-frame key visual information extraction on the video data of the automobile welding production line, and utilizing image processing techniques such as contrast and brightness adjustment, it is possible to generate clearer serialized image frames and improve the recognition accuracy of the weld boundary. Performing image binarization processing not only clarifies the boundary between the weld and the background, but also enhances the visualization of the weld boundary in the image, making the identification of abnormal temperature distribution more accurate. By identifying internal inhomogeneities through pixel intensity analysis, the system can more effectively mark and track abnormal points and improve the positioning accuracy of abnormal temperature welds. Timestamp index analysis helps capture key time points, compares changes in weld quality within differentiated time periods, and increases the accurate record of the time when the anomaly occurs. Quality classification based on weld shape and heat distribution, as well as adjustment of welding parameters, not only optimizes production efficiency, but also improves the overall standard of welding quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the workflow of the present invention; Figure 2 This is a detailed flow chart of S1 of the present invention; Figure 3 This is a detailed flow chart of S2 of the present invention; Figure 4 This is a detailed flow chart of S3 of the present invention; Figure 5 This is a detailed flow chart of S4 of the present invention; Figure 6 This is a detailed flow chart of S5 of the present invention; Figure 7 This is a detailed flow chart of S6 of the present invention; Figure 8 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0019] See also Figure 1 The present invention provides a technical solution, a visual inspection method for an automobile welding production line, comprising the following steps: S1: Based on the video data of the automobile welding production line, the original image sequence is frame-segmented, key visual information is extracted frame by frame, and image processing technology is used to adjust the contrast and brightness, optimize the image quality, and generate serialized image frames; S2: Use serialized image frames to identify the weld boundary of the automobile welding assembly, perform image binarization processing to distinguish the weld from the background, and obtain the weld boundary image; S3: Through the weld boundary image, the pixel intensity is used to analyze the internal non-uniformity of the automobile weldment, identify abnormal temperature distribution, collect abnormal features and mark them, and track and locate abnormal points to generate calibrated abnormal welds; S4: Use the timestamp index to analyze the key time points in the automobile welding production line by calibrating abnormal welds. By comparing the changes in weld quality in different time periods, the time when the abnormality occurred is recorded and the time positioning results are obtained; S5: Based on the time positioning results, the quality of the automobile welding is analyzed, the welding quality is classified according to the shape and heat distribution of the weld, and the automobile welding is graded according to the quality indicators to generate the welding quality grade; S6: Use welding quality grades to adjust and optimize welding parameters, update the current and speed of welding machines on the production line, match differentiated welding requirements, optimize production efficiency and quality, and obtain a production parameter adjustment plan.

[0020] The serialized image frames include contrast-adjusted images, frame brightness-adjusted images, and key visual information of each frame. The weld boundary images include images distinguishing the weld boundary and images comparing the background and the weld. The calibrated abnormal welds include weld images marked with abnormal temperature distribution, weld inhomogeneity analysis images, and abnormal point tracking and positioning maps. The time positioning results include timestamp indexes of key time points and time period records of weld quality changes. The welding quality grade includes weld shape classification result maps, thermal distribution analysis result maps, and welding quality grading standards. The production parameter adjustment plan includes updated current setting parameters, adjusted welding speed parameters, and differentiated welding demand matching plans.

[0021] See also Figure 2 Based on the video data of the automobile welding production line, the original image sequence is segmented, key visual information is extracted frame by frame, and the contrast and brightness are adjusted using image processing technology. The specific steps to generate serialized image frames are as follows: S101: Frame segmentation is performed on the video data of the automobile welding production line. A target number of frames are extracted at a fixed time interval of one second. The continuity and integrity of the data are verified. The execution process of generating a frame sequence image is as follows; Frame segmentation of video data from an automotive welding production line involves dividing the video into fixed time intervals to ensure that a predetermined number of frames are extracted per second, capturing the continuous action and key moments on the production line. Frame segmentation is crucial for ensuring data continuity and integrity, avoiding the loss of important information during the extraction process. This allows for the efficient extraction of static images from dynamic video, providing a foundation for subsequent analysis. Furthermore, the timestamp and sequence of each frame must be verified to generate a frame sequence image.

[0022] S102: Using the frame sequence image, the coordinates of the welding points and the contact points of the vehicle body in each frame are extracted, key visual areas are distinguished by color filtering, and visual information is marked to obtain the key visual frame sequence. The execution process is as follows; Using frame sequence images, key elements are extracted for identification and coordinate extraction. This requires identifying the contact points between the weld and the car body in each frame and distinguishing key visual areas using color filtering technology. Color filtering helps identify welds and important visual markers for further analysis and processing. Coordinate extraction ensures accurate tracking of the weld's trajectory and operating points on the production line. The extracted visual information will be used for labeling, providing data support for automated monitoring and quality control systems. This is key to implementing automated visual monitoring systems and is of great significance for improving production line efficiency and welding quality. The key visual frame sequence is obtained using the following formula: in, are the welding point coordinates, and are the X and Y coordinates of the point where the weld contacts the car body, and Processing correction values, and are the adjustment parameters for the X and Y coordinates, and is the weight factor for coordinate adjustment.

[0023] S103: Adjust the contrast and brightness parameters of each frame through the key visual frame sequence, increase the brightness and contrast of the target amount, optimize the image quality and visual effect, and generate the sequenced image frame. The execution process is as follows; By adjusting the contrast and brightness parameters of the image through the key visual frame sequence, the image quality of the extracted key visual frame sequence is optimized. By increasing the brightness and contrast of the target number, important features in the image can be made more prominent, facilitating subsequent analysis and recognition. Adjusting the parameters not only improves the visual effect of the image, but also helps the automation system perform tasks such as solder joint detection and alignment more accurately. The optimized image is also more user-friendly for operators to monitor the status of the production line, allowing problems and anomalies to be observed more clearly. By enhancing the quality of each frame of the image, a serialized image frame is generated using the formula: in, is the adjusted image brightness, is the original image brightness, and To adjust the contrast and brightness parameters, is the sensitivity parameter for contrast adjustment, is the average brightness of the image.

[0024] See also Figure 3 , using serialized image frames, identifying the weld boundary of automobile welding, performing image binarization processing to distinguish the weld from the background, and obtaining the weld boundary image in the following steps: S201: Using serialized image frames, adjusting the image to a uniform grayscale range based on the contrast difference between the weld and non-weld areas of the automobile welding production line, verifying the visualization of the weld features in the image, and obtaining a weld highlight image. The execution process is as follows; The serialized image frames are used to highlight the weld area on the automobile welding production line by adjusting the grayscale range of the image. The focus is on using the difference in contrast between the weld and non-weld areas to adjust the image to a unified grayscale range so that the weld position can be more clearly identified. The adjustment helps to enhance the visualization of the weld features, making the weld more obvious in the image, which is convenient for subsequent analysis and processing. The unified grayscale processing of the image not only improves the image quality, but also provides clearer visual input for the automated inspection system, improves recognition accuracy and processing speed, and generates a weld highlight image. The formula used is: in, Represents the grayscale value of the adjusted image, and are the maximum and minimum values ​​of the target grayscale range, is the grayscale value of the original image, is the minimum grayscale value of the original image, and are the maximum and minimum grayscale values ​​in the original image, respectively.

[0025] S202: Based on the weld highlight image, a grayscale threshold is set to perform image binarization. By optimizing the contrast between the weld boundary and the background, the weld area is separated from the non-weld area to obtain a background-separated image. The execution process is as follows; Based on the weld highlight image, binarization processing is performed. The weld area and the non-weld area are separated by setting an appropriate grayscale threshold. The contrast between the weld boundary and the background is optimized by adjusting the threshold to ensure that the weld area is visually clearly separated from the surrounding non-weld area. Binarization is a commonly used technology in image processing. The image is simplified into black and white, with white representing the weld area and black representing the background. This not only simplifies the complexity of the image, but also facilitates subsequent image analysis and feature extraction, such as automatic detection of weld quality, to obtain a background-distinguished image.

[0026] S203: The execution process of distinguishing the image from the background, identifying and extracting the position of the weld boundary, optimizing the continuity and clarity of the boundary line, evaluating and verifying the visual recognition effect of the boundary line, and generating the weld boundary image is as follows; By distinguishing the image from the background, the position of the weld boundary is identified and extracted. Not only is it necessary to identify the boundary, but it is also necessary to optimize the continuity and clarity of the boundary line to ensure that the representation of the weld boundary in the image is accurate. The process of optimizing the boundary line includes adjusting the image processing parameters to improve the visual recognition effect of the boundary, making the boundary more obvious and facilitating accurate interpretation by the machine vision system. This is the key to ensuring the accuracy of welding quality control. Only by clearly identifying and correctly analyzing the weld boundary can the automated equipment accurately perform subsequent welding tasks and generate a weld boundary image. The formula used is: in, represents the weld boundary image, Representative The image of the weld boundary images, Indicates the clarity of the boundary line, is the weight coefficient for adjusting the clarity effect, Indicates the discontinuity of the boundary line, Represents the total number of images.

[0027] See also Figure 4 ,Through the weld boundary image, pixel intensity is used to analyze the internal non-uniformity of the automobile weld, identify the abnormal temperature distribution, collect and mark the abnormal features, and track and locate the abnormal points. The steps for generating ,calibrated abnormal welds are as follows: S301: Using the weld boundary image, the inhomogeneity within the weld is evaluated by analyzing the intensity of the pixels, recording areas where pixel intensity changes, and marking potential inhomogeneous areas to obtain an inhomogeneous weld analysis diagram. The execution process is as follows; The weld boundary image is used to analyze the internal non-uniformity of the weld. Through detailed evaluation of the pixel intensity in the image, it is possible to identify the non-uniform areas that appear during the welding process. These areas are manifested as significant changes in pixel intensity, which is crucial for detecting potential welding defects because non-uniformity indicates insufficient material strength or welding technology problems. By recording and marking the changed areas, further quantitative analysis and quality control can be carried out to ensure the consistency of the welding process and the structural integrity of the product, and to generate non-uniform weld analysis diagrams.

[0028] S302: Based on the uneven weld analysis diagram, edge detection technology is used to identify abnormal temperature areas in the automobile welding production line, record areas that deviate from the standard weld temperature distribution, and process the abnormal temperature distribution to obtain an abnormal temperature weld diagram. The execution process is as follows; Based on the uneven weld analysis diagram, edge detection technology is used to identify abnormal temperature areas from the uneven weld analysis diagram. This can accurately locate areas with uneven temperature distribution during the welding process, which is crucial for ensuring welding quality and avoiding welding defects. Abnormal temperature areas are related to potential welding problems, such as overheating or failure to reach the required welding temperature. By recording the deviation area and comparing it with the standard weld temperature distribution, the welding parameters can be more accurately processed and adjusted to correct the problem and generate an abnormal temperature weld diagram.

[0029] The formula for edge detection technology is as follows: in, is the gradient amplitude, and are the gradients of the image in the horizontal and vertical directions, and are the weighting coefficients in the horizontal and vertical directions, is the weighting coefficient of the diagonal gradient, is the gradient in the diagonal direction of the image, is the Gaussian blur coefficient, is the standard deviation, is the base of the natural exponential.

[0030] The execution process is as follows: Calculate the horizontal and vertical gradients of each pixel in the image and , applying a weight coefficient to the gradient value and , the weights are automatically adjusted by analyzing the image content to enhance important edge features and introduce diagonal gradients and apply weights , to identify edge directions that are ignored in standard edge detection, apply a Gaussian blur coefficient and standard deviation , the influence of the gradient value is adjusted by the Gaussian function to smooth the noise and optimize the edge detection results. The optimal value of the parameter is determined by the preliminary analysis of the image, or learned from batch samples using machine learning technology to complete the entire edge detection process, obtain a more accurate edge intensity map, and enhance the ability to identify abnormal temperature areas.

[0031] S303: Using the abnormal temperature weld map, mark the location of the abnormal temperature distribution, track and locate the abnormal point, associate the mark with the real-time location of the abnormal area, and generate the execution process of calibrating the abnormal weld as follows; Through the abnormal temperature weld map, the location of abnormal temperature distribution is marked, the abnormal point is tracked and located, and the mark is associated with the real-time location of the abnormal area, allowing operators to quickly respond and adjust welding parameters to solve potential quality problems. By accurately calibrating the abnormal area, the welding quality control on the production line can be ensured to be more accurate, effectively reducing the scrap rate and improving product consistency. The formula for generating calibrated abnormal welds is: in, represents the normalized difference measure of abnormal temperature distribution in the image, For The observed point Temperature values, Normal temperature value for reference, is the number of measurement points, is the standard deviation of the temperature values, The number of measurement points minus 1.

[0032] See also Figure 5 , by calibrating abnormal welds, using timestamp index to analyze key time points in the automobile welding production line, by comparing the changes in weld quality in different time periods, recording the time when the abnormality occurred, and obtaining the time positioning results, the specific steps are as follows: S401: calibrate abnormal welds, assign a timestamp to each abnormal weld, synchronize the timestamp with the log of the welding production line, identify and extract the time when the abnormality occurred, and obtain the abnormal timestamp index map. The execution process is as follows; By calibrating abnormal welds, assigning a timestamp to each identified abnormal weld and synchronizing it with the operation log of the welding production line, the specific time point of each abnormality can be accurately tracked, allowing for more in-depth analysis and root cause exploration. By synchronizing the timestamp with the production line log, the time of each welding abnormality can be accurately identified and extracted, which not only helps the operation team quickly locate the problem, but also helps to perform preventive maintenance and operation optimization. The abnormal timestamp index map is generated using the formula: in, Represents the abnormal timestamp index map, Representative timestamp of each abnormal weld, is the weight coefficient for adjusting the time synchronization impact, Indicates time synchronization, Represents the time record of the production log, Indicates the time deviation, Represents the total number of abnormal welds.

[0033] S402: Based on the abnormal timestamp index map, analyze the weld quality near the key time point, compare the weld image recognition quality change trend in the differentiated time period, analyze the time interval of the weld quality change, and obtain the time-differentiated weld quality map. The execution process is as follows; Based on the abnormal timestamp index map, the weld quality near the key time points is analyzed. By comparing the weld images in different time periods, the quality change trend can be identified to ensure the continuous optimization of welding quality. By analyzing the time interval of weld quality changes, the problems and fluctuations in the production process can be better understood, helping the team predict potential risk points and take measures to improve or adjust them, thereby generating a time-differentiated weld quality map.

[0034] S403: Using the time-differentiated weld quality map, the time of each quality anomaly is recorded and compared with the history of the automobile welding production line. The time when the anomaly occurred is marked, and the execution process of generating the time positioning result is as follows; By using a time-differentiated weld quality map, the specific time of each quality anomaly is recorded and compared with the historical data of the automobile welding production line. By marking the time when the anomaly occurred, the root cause of the problem can be tracked and located more accurately, which is crucial for understanding periodic problems and random failures that occur on the production line. By correlating the anomaly time with the actual operating conditions, the production process can be optimized to ensure that the problem is identified and resolved in a timely manner. The time location result is generated using the formula: in, For time positioning results, is the current quality abnormality time point, is the historical data time point, For abnormal quality record values, is the historical data weight coefficient.

[0035] See also Figure 6 Based on the time positioning results, the quality of automobile welding is analyzed, the welding quality is classified according to the shape and heat distribution of the weld, and the automobile welding is graded according to the quality indicators. The specific steps for generating welding quality grades are as follows: S501: Using the time positioning results, analyze the welding quality at the target time point, monitor the shape and heat distribution of the weld, evaluate the consistency with the standard quality, and obtain the weld shape and heat distribution analysis diagram. The execution process is as follows; Time positioning results are used to analyze welding quality at specific points in time, including detailed monitoring of weld shape and heat distribution to assess consistency with standard quality. This can identify quality issues that affect product structural integrity and performance, such as irregular weld shape or uneven heat distribution. This is crucial to ensuring that every welding operation meets high quality standards and can help production teams adjust welding parameters or processes in a timely manner to prevent quality deviations from affecting product performance. The weld shape and heat distribution analysis diagram is generated using the following formula: in, is the weld quality analysis value, and are the weld shape and heat distribution coefficient, and is the shape and heat distribution coefficient of the standard mass, is the standard mass deviation.

[0036] S502: Based on the weld shape and heat distribution analysis diagram, the weld is classified using preset quality evaluation indicators, including weld integrity, uniformity, and heat distribution regularity. The execution process for obtaining the classified weld quality diagram is as follows; Based on the weld shape and heat distribution analysis diagram, the preset quality evaluation indicators are applied to classify the welds, including evaluating the integrity, uniformity and regularity of the heat distribution of the welds, to improve the quality control of the welding process. Through quantitative analysis of various aspects of the welds, it can be ensured that the welding process meets the predetermined quality standards. This not only helps the operation team understand the quality status of each weld, but also provides a basis for subsequent process improvements. The classified welding quality diagram is generated using the formula: in, is the classification welding quality value, is the weld integrity value, is the weld uniformity value, is the regularity value of heat distribution, is the evaluation weight coefficient.

[0037] S503: By classifying the welding quality map, the automobile welding assembly is graded according to the quality indicators, and the quality priority is sorted. It is verified that each level meets the associated quality standards. The execution process of generating the welding quality grade is as follows; By classifying the welding quality map, grading the automotive welding assembly according to the quality indicators and prioritizing the quality, it is possible to effectively verify whether each level meets the associated quality standards, which is crucial to ensuring the quality consistency of the entire production line. Through accurate quality grading, the welding process can be better managed, quality problems can be prevented, and each welded part can be ensured to meet or exceed the expected performance standards. The welding quality grade is generated using the formula: in, is the welding quality grade, is the welding quality grading value, is the priority sorting coefficient, is the compliance standard threshold, is the consistency coefficient of historical data.

[0038] See also Figure 7 , using welding quality levels, adjust and optimize welding parameters, update the current and speed of welding machines on the production line to match differentiated welding requirements, and obtain the production parameter adjustment plan in the following steps: S601: Using welding quality levels, analyze welding quality data under differentiated levels, identify welding parameters that need to be adjusted, including current and speed, formulate adjustment requirements for each quality level, and obtain a welding parameter requirement analysis table. The execution process is as follows; Using welding quality grades, the team analyzes welding quality data for each grade, including identifying welding parameters that require adjustment, such as current and speed. By analyzing the welding data for each quality grade in detail, key factors influencing welding quality can be identified and specific adjustments can be made for each grade, improving welding quality and production efficiency. This allows the operator to adjust the welding machine settings to meet the specific requirements of each quality grade and generate a welding parameter requirements analysis table.

[0039] S602: Based on the welding parameter demand analysis table, set the current and speed adjustment plan. The adjustment range is based on the welding quality level requirements. Verify that the welding parameters of each level match the quality standards. The execution process of the draft welding parameter adjustment plan is as follows; Based on the welding parameter requirements analysis table, set the current and speed adjustment plan. The range of the adjustment plan is determined by the specific requirements of the welding quality level. It is key to ensure that the welding parameters match the quality standards. Through accurate parameter setting, the welding quality and the overall efficiency of the production line can be significantly improved. Setting the right current and speed not only affects the welding quality, but also affects the production cost and equipment life. Generate a draft welding parameter adjustment plan.

[0040] S603: Update the current and speed of the welding machines on the production line through the draft welding parameter adjustment plan. By adjusting the settings of the differentiated welding stations to match the real-time changing welding requirements, the execution process of generating the production parameter adjustment plan is as follows; Draft welding parameter adjustment plans are used to update the current and speed of welding machines on the production line, adjusting the settings of differentiated welding stations to match real-time welding requirements. This is crucial for ensuring that welding processes adapt to changing production demands. Effective parameter adjustments can reduce quality issues and improve production efficiency. By implementing these adjustments, the production line can more flexibly respond to different production challenges and generate production parameter adjustment plans.

[0041] See also Figure 8 A visual inspection system for an automobile welding production line is provided. The visual inspection system for an automobile welding production line is used to perform the above-mentioned visual inspection method for an automobile welding production line. The system comprises: The frame segmentation module segments the original image sequence based on the video data of the automobile welding production line, extracts visual information frame by frame, and adjusts the contrast and brightness of each frame to generate serialized image frames; The boundary recognition module identifies the weld and the background based on the serialized image frames, performs image binarization processing, and generates a weld boundary image; The abnormal temperature analysis module uses the weld boundary image to analyze pixel intensity, identify abnormal temperature distribution, collect and mark abnormal features, and track and locate abnormal points to generate calibrated abnormal welds; The timestamp indexing module calibrates abnormal welds and analyzes key time points in the automotive welding production line. By comparing changes in weld quality and recording the time when the abnormality occurred, the time positioning result is obtained. The welding quality classification module analyzes the weld shape and heat distribution based on the time positioning results, classifies the welding quality, adjusts and optimizes the welding parameters according to differentiated welding requirements, and generates a production parameter adjustment plan.

[0042] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A visual inspection method for an automobile welding production line, characterized in that: The following steps are involved: Based on the video data of the automobile welding production line, the original image sequence is segmented, key visual information is extracted frame by frame, and the contrast and brightness are adjusted using image processing technology to generate serialized image frames; The serialized image frames are used to identify the weld boundary of the automobile welding assembly, and image binarization processing is performed to distinguish the weld from the background to obtain a weld boundary image; By using the weld boundary image, the pixel intensity is used to analyze the non-uniformity inside the automobile welding assembly, the abnormal temperature distribution is identified, the abnormal features are collected and marked, and the abnormal points are tracked and located to generate a calibrated abnormal weld; The abnormal welds are calibrated as described above, and the key time points in the automobile welding production line are analyzed using the timestamp index. By comparing the changes in weld quality in different time periods, the time when the abnormality occurs is recorded, and the time positioning result is obtained; Based on the time positioning result, the quality of the automobile welding is analyzed, the welding quality is classified according to the shape and heat distribution of the weld, the automobile welding is graded according to the quality index, and the welding quality grade is generated; The welding quality grade is used to adjust and optimize welding parameters, update the current and speed of welding machines on the production line, match differentiated welding requirements, and obtain a production parameter adjustment plan.

2. The visual inspection method for automobile welding production line according to claim 1 is characterized in that: The serialized image frames include contrast-adjusted images, frame brightness-adjusted images, and key visual information of each frame; the weld boundary image includes an image distinguishing the weld boundary and an image comparing the background and the weld; the calibrated abnormal weld includes a weld image marked with abnormal temperature distribution, an image for analyzing the inhomogeneity within the weld, and an abnormal point tracking and positioning map; the time positioning result includes a timestamp index of a key time point and a time period record of weld quality changes; the welding quality grade includes a weld shape classification result map, a heat distribution analysis result map, and a welding quality grading standard; the production parameter adjustment plan includes an updated current setting parameter, an adjusted welding speed parameter, and a differentiated welding demand matching plan.

3. The visual inspection method for automobile welding production line according to claim 1 is characterized in that: Based on the video data of the automobile welding production line, the original image sequence is segmented, key visual information is extracted frame by frame, and the contrast and brightness are adjusted using image processing technology. The specific steps to generate serialized image frames are as follows: Perform frame segmentation processing on the video data of the automobile welding production line, extract the target number of frames per second with a fixed time interval, verify the continuity and integrity of the data, and generate frame sequence images; The frame sequence images are used to extract coordinates of the welding points and the contact points of the automobile body in each frame, distinguish key visual areas by color filtering, and mark visual information to obtain a key visual frame sequence; Through the key visual frame sequence, the contrast and brightness parameters of each frame of the image are adjusted to increase the brightness and contrast of the target amount respectively, optimize the image quality and visual effect, and generate a serialized image frame.

4. The automobile welding production line visual inspection method according to claim 1 is characterized in that: The steps of using the serialized image frame to identify the weld boundary of the automobile welding assembly, performing image binarization processing to distinguish the weld from the background, and obtaining the weld boundary image are specifically as follows: The serialized image frames are used to adjust the image to a uniform grayscale range according to the contrast difference between the weld and non-weld areas of the automobile welding production line, verify the visualization of the weld features in the image, and obtain a weld highlight image; Based on the weld protrusion image, a grayscale threshold is set to binarize the image, and the weld area is separated from the non-weld area by optimizing the contrast between the weld boundary and the background to obtain a background-distinguished image; The background is distinguished by the image, the position of the weld boundary is identified and extracted, the continuity and clarity of the boundary line are optimized, the visual recognition effect of the boundary line is evaluated and verified, and the weld boundary image is generated.

5. The visual inspection method for automobile welding production line according to claim 1 is characterized in that: The steps of analyzing the unevenness of the interior of the automobile welding assembly by using pixel intensity through the weld boundary image, identifying abnormal temperature distribution, collecting and marking abnormal features, and tracking and locating abnormal points to generate and calibrate abnormal welds are as follows: Using the weld boundary image, evaluating the internal non-uniformity of the weld by analyzing the intensity of pixels, recording areas where pixel intensity changes, and marking potential non-uniform areas to obtain a non-uniform weld analysis diagram; Based on the uneven weld analysis diagram, edge detection technology is used to identify abnormal temperature areas in the automobile welding production line, record areas that deviate from the standard weld temperature distribution, and process the abnormal temperature distribution to obtain an abnormal temperature weld diagram; Through the abnormal temperature weld map, the position of abnormal temperature distribution is marked, the abnormal point is tracked and located, the mark is associated with the real-time position of the abnormal area, and a calibrated abnormal weld is generated.

6. The visual inspection method for automobile welding production line according to claim 5 is characterized in that: The formula for the edge detection technique is as follows: in, is the gradient amplitude, and are the gradients of the image in the horizontal and vertical directions, respectively. and are the weighting coefficients in the horizontal and vertical directions, is the weighting coefficient of the diagonal gradient, is the gradient in the diagonal direction of the image, is the Gaussian blur coefficient, is the standard deviation, is the base of the natural exponent.

7. The visual inspection method for automobile welding production line according to claim 1 is characterized in that: The steps of using the calibrated abnormal welds, analyzing the key time points in the automobile welding production line using the timestamp index, comparing the changes in weld quality in differentiated time periods, recording the time when the abnormality occurred, and obtaining the time positioning result are as follows: Using the calibration method for abnormal welds, assigning a timestamp to each abnormal weld, synchronizing the timestamp with the log of the welding production line, identifying and extracting the time when the abnormality occurred, and obtaining an abnormal timestamp index map; Based on the abnormal timestamp index map, the weld quality near the key time point is analyzed, the quality change trend of the weld image recognition in the differentiated time period is compared, the time interval of the weld quality change is analyzed, and a time-differentiated weld quality map is obtained; The time of each quality anomaly is recorded through the time-differentiated weld quality map, and compared with the history of the automobile welding production line, the time when the anomaly occurred is marked, and a time positioning result is generated.

8. The automobile welding production line visual inspection method according to claim 1 is characterized in that: Based on the time positioning results, the quality of automobile welding is analyzed, the welding quality is classified according to the shape and heat distribution of the weld, and the automobile welding is graded according to the quality indicators. The specific steps of generating the welding quality grade are as follows: The time positioning result is used to analyze the welding quality at the target time point, and the consistency with the standard quality is evaluated by monitoring the shape and heat distribution of the weld, so as to obtain a weld shape and heat distribution analysis diagram; Based on the weld shape and heat distribution analysis diagram, the welds are classified using preset quality evaluation indicators, including the integrity, uniformity and regularity of heat distribution of the welds, to obtain a classified welding quality diagram; Through the classified welding quality map, the automobile welding is graded according to the quality index, and the quality priority is sorted, and each level is verified to meet the associated quality standards to generate the welding quality grade.

9. The automobile welding production line visual inspection method according to claim 1, characterized in that: The welding quality level is used to adjust and optimize welding parameters, update the current and speed of welding machines on the production line, match differentiated welding requirements, and obtain the production parameter adjustment plan in the following steps: Using the welding quality grades, analyzing welding quality data under differentiated grades, identifying welding parameters that need to be adjusted, including current and speed, formulating adjustment requirements for each quality grade, and obtaining a welding parameter requirement analysis table; Based on the welding parameter demand analysis table, set the current and speed adjustment plan, the adjustment range is based on the requirements of the welding quality level, verify that the welding parameters of each level match the quality standards, and obtain a draft welding parameter adjustment plan; Through the draft welding parameter adjustment plan, the current and speed of the welding machines on the production line are updated, and by adjusting the settings of differentiated welding sites to match the real-time changing welding requirements, a production parameter adjustment plan is generated.

10. A visual inspection system for an automobile welding production line, characterized in that: According to the automobile welding production line visual inspection method according to any one of claims 1 to 9, the system comprises: The frame segmentation module segments the original image sequence based on the video data of the automobile welding production line, extracts visual information frame by frame, and adjusts the contrast and brightness of each frame to generate serialized image frames; The boundary recognition module recognizes the weld and the background based on the serialized image frame, performs image binarization processing, and generates a weld boundary image; The abnormal temperature analysis module uses the weld boundary image to analyze pixel intensity, identify abnormal temperature distribution, collect and mark abnormal features, and track and locate abnormal points to generate calibrated abnormal welds; The timestamp index module analyzes the key time points in the automobile welding production line by calibrating the abnormal welds, and obtains the time positioning result by comparing the changes in weld quality and recording the time when the abnormality occurs; The welding quality classification module analyzes the weld shape and heat distribution according to the time positioning result, classifies the welding quality, adjusts and optimizes the welding parameters according to differentiated welding requirements, and generates a production parameter adjustment plan.

Citation Information

Patent Citations

  • Circular weld defect detection method based on visual identification

    CN115100127A

  • Weld joint identification and track detection system and method based on visual technology

    CN118744296A

  • Appearance inspecting device, welding system, shape data correcting method, and method for appearance inspection of a weld

    US20240337604A1

  • Automatic seam detection for a welding process

    WO2023280679A1

Cited By

  • Automobile body welding seam detection method and system based on machine vision

    CN119574571A

  • Positioning weld robot system based on active and passive visual sensing

    CN120306762A

  • A seam-finding robot system based on active and passive vision

    CN120306762B

  • Automatic production line control system and method

    CN120335417A